Analytics glossaryAnalytics, growth and support terms, defined the way TapCub uses them
Forty-eight terms from PV to two-way translation. Each one gets a plain definition, the place in TapCub where you see it, and the terms it is usually read with.
Definitions, not marketing
Where tools disagree — unique visitors, bounce rate, sessions — we say which definition TapCub uses and why.
Linked to the product
Every entry points to the report, setting or help article where the term shows up, so you can go from word to screen.
Search or jump
Type to filter, or pick a letter. Links from the blog and help centre land on individual entries.
A 500 error (HTTP 500 Internal Server Error) is the status code a server returns when an unexpected condition on its side stopped it from completing the request. It says the server failed but not why, so the cause is found in server logs, not in the browser.
The 75th percentile (p75, upper quartile or Q3) is the value that 75% of measurements are at or below; the other 25% are higher. Google's Core Web Vitals judge pages on p75: at least 75% of page views must meet each metric's good threshold.
A percentile is the value below which a given share of measurements falls. The 95th percentile (P95) is the value that 95% of measurements are at or below, so only the slowest or largest 5% are above it.
The 95th percentile (P95) is the value that 95% of measurements fall at or below, so only 5% are higher. It describes the slow tail an average hides: on 20 sample page loads the mean is 2.51 s but P95 is 6.8 to 9.4 s, depending on the method.
A growth framework with five stages: Acquisition, Activation, Retention, Revenue and Referral. It turns "is the business growing" into five measurable questions and shows which stage leaks most. TapCub builds the five stages from your events once you map one event to each.
ABX (account-based experience) is a B2B approach in which sales, marketing and customer success give each target account one connected, personalised experience across the whole buying journey. It extends ABM past the deal and measures by account, not by individual person.
Account intelligence is the practice of merging what you know about a company, such as visits, chats, product events and CRM data, into one account-level view and score. The unit is the organization, not the lead, so you can see whether it is engaged and what to do next.
Account-based marketing (ABM) is a B2B strategy that treats each target company as its own market: sales and marketing choose the accounts together, plan outreach to the several people who buy, and measure progress by account (engaged, opportunity, won) instead of by individual lead.
Account-level analytics, also called group analytics, measures behavior by the organization a person belongs to (a company, team, workspace or store) instead of by the individual. In a B2B product the customer is the account, so twenty people each using a little differs from one person using a lot.
Activation rate is the percentage of new users who complete an activation event, the action that marks their first real value from the product, within a defined window after signing up.
An active user is a unique user who performed at least one qualifying action in a product during a set window, such as a day, week or month. Who counts as a user, what counts as an action and how long the window is all change the number, so the same people can be 1,900 or 5,600 monthly active users.
An ad blocker is software, usually a browser extension, that cancels requests matching filter-list rules for ads and often trackers. A blocked analytics script never records the visit. The useful question is how large your own gap is and whether it differs between audience groups.
An ad exchange is a marketplace that auctions each ad impression in real time: sellers offer the slot through supply-side platforms, buyers bid through demand-side platforms, and the exchange applies the floor, picks the winner and sets the price, all within a bid deadline of about 100 ms.
Ad fraud is the deliberate creation of fake ad activity, such as clicks, impressions or installs, to collect money from advertisers. It is the intentional part of invalid traffic. In your analytics it shows as clicks that never arrive, near-zero engagement and no conversions from one source.
An ad network is a company that collects ad space from many websites or apps and sells it to advertisers as one package, keeping part of the price. For a site owner the key question is revenue against the effect on visitors.
An ad server is the system that decides which ad fills each slot on a page or app, delivers it and counts impressions and clicks. Publishers run one to manage inventory; advertisers use a third-party one to verify delivery across publishers.
Ad tech is the software and services that buy, sell, deliver, target and measure digital advertising: advertiser, agency, DSP, exchange, SSP, ad server and verification. Each layer takes a fee, and the ANA found only 43.9% of programmatic spend reached consumers in 2024.
Ad hoc analysis is a one-off, question-driven investigation of data that answers a specific business question no existing report covers. The expensive part is starting from zero every time; good ad hoc work leaves a reusable metric or segment definition behind.
Addressable TV is television advertising in which different households or audience segments watching the same programme see different ads, chosen from household data such as demographics, viewing behaviour and location. It works on set-top boxes and smart TVs, not only streaming.
An advertisement is a paid or sponsored message that tells an audience about a product, service or idea, and is meant to be recognized as promotion. Digital ads are bought per impression (CPM), per click (CPC) or per action (CPA), so compare formats on cost per conversion.
An advertiser is the company, organisation or person that pays for an advertisement; in online advertising it is the buyer of inventory, while the publisher is the seller. Advertisers judge a campaign by CPM, CPC, CPA and ROAS against a break-even of 1 ÷ gross margin.
A member role that opens the chat inbox and nothing else in the console. Agents reply to visitors, see visit context and ratings, but cannot read analytics, billing or settings. Use it for support staff and outsourced teams.
In product growth, the aha moment is the point at which a new user first experiences the core value of a product, usually tied to a specific action that predicts whether they keep using it.
A bot operated by an AI company to collect training data or answer queries in real time, such as fetchers behind chat assistants and search answers. TapCub keeps them on a separate ledger from human traffic and compares what they fetched with what your robots.txt and llms.txt allowed.
Alpha testing is the first round of testing a build gets from people outside the development team, run in the company's own environment before external users see it. It answers whether the product can be used at all, and should end on exit criteria written as numbers, not on a date.
Analytics is the systematic, computer-assisted analysis of data to find patterns and support decisions. It is commonly split into four types: descriptive (what happened), diagnostic (why), predictive (what will happen) and prescriptive (what to do).
Annual growth rate is the percentage change in a metric over one year, but the name covers at least three formulas: year-over-year change, an annualized short-period rate, and the compound annual growth rate (CAGR). Always state which formula produced the number.
Annual recurring revenue (ARR) is the annualised value of the recurring subscription contracts a company has at a point in time, excluding one-off fees and variable usage that is not committed.
Annual run rate is a short period's revenue extrapolated to a full year, such as last month × 12 or last quarter × 4. It is a straight-line projection, so it overstates after peak seasons or one-off deals, and it is not the same as annual recurring revenue (ARR).
Anomaly detection is the process of finding observations, such as a day of traffic or a single transaction, that differ so much from the expected pattern that they are unlikely to be normal variation and are worth investigating.
Anonymization means processing data so people can no longer be identified by anyone using means reasonably likely to be used. Removing names is not enough, and a salted hash that stays stable is usually only pseudonymous. The test is whether a record can still be singled out, linked or inferred.
Anscombe's quartet is a set of four data sets, published by Francis Anscombe in 1973, that share nearly identical means, variances, correlation and linear regression line but have very different distributions, demonstrating the importance of plotting data.
APAC is short for Asia-Pacific: the countries of Asia and Oceania. It has no single official list; by the UN M49 geoscheme that is 79 countries and areas (Asia 50, Oceania 29), or 61 without Western Asia and 56 without Central Asia as well.
An Apdex score (Application Performance Index) is an open-standard 0–1 measure of user satisfaction with response time: samples at or below threshold T count 1, up to 4T count 0.5, slower ones and errors count 0; the score is (satisfied + tolerating ÷ 2) ÷ total samples.
API rate limiting is a cap, enforced by the server, on how many requests a client may send in a given period. Requests over the cap are refused, usually with HTTP 429 Too Many Requests, and the client is expected to wait before trying again.
ARPDAU (average revenue per daily active user) is a day's revenue divided by that day's active users. It shows the immediate effect of events and releases, but must be read with DAU: when daily actives fall, ARPDAU can rise while total revenue falls.
ARPPU (average revenue per paying user) is total revenue in a period divided by the number of users who made a payment in that period. Multiplied by the payer conversion rate, it equals ARPU.
ARR (annual recurring revenue) is the yearly value of the recurring subscription revenue a business expects to keep from current customers, usually MRR × 12. It is a run rate on one day, not GAAP revenue, and excludes one-time fees, trials and variable usage.
ASP (average selling price) is the average price at which a product or service actually sold in a period: net revenue divided by units sold. It is net of discounts, a per-unit figure, and it moves with the product mix even when no price changes.
An assisted conversion is a conversion in which a channel or touchpoint appeared on the path before the final interaction. The channel helped but did not close. Counting assists shows channels that last-click attribution undervalues because they open or sustain journeys rather than finish them.
Audience segmentation is the practice of dividing your audience into smaller groups that share traits — such as demographics, location, behavior or acquisition source — so you can measure, target and message each group separately instead of treating all visitors as one average.
Revenue divided by number of orders over a period. It tells you whether growth comes from more orders or bigger ones, and is the first thing to check when revenue moves while traffic does not. Requires an order event with a value.
Revenue over a period divided by the number of active users in that period. Subscription businesses watch it monthly; commerce teams compare it by cohort or channel. TapCub computes it from revenue events and the active-user definition you choose.
The rule that decides which channel gets credit for a conversion. First-touch credits the channel that brought the person the first time; last-touch credits the one right before the conversion. The two answer different questions, so TapCub shows both side by side rather than picking one.
Recording clicks, form submits and page changes automatically, without writing tracking code for each one. TapCub's AI event inbox then proposes names for the patterns it finds, so the tracking plan is built from what users actually do.
A backlink is a link on another website that points to a page on yours. Backlinks bring visitors and are one of the signals search engines use to judge which pages other sites consider worth citing.
A banner ad is a rectangular graphic advertisement, shown as an image, animation or HTML5 creative on a web page or in an app, that links to the advertiser's landing page when clicked. Standard sizes include 300×250, 728×90 and 300×600.
The base rate fallacy is the error of judging the probability of an event from specific evidence while ignoring how common the event is overall (its base rate). With rare events, it leads people to badly overestimate how often a positive signal is correct.
Basic authentication (HTTP Basic) is the simplest HTTP login scheme: the client sends Authorization: Basic followed by the Base64 of username:password with every request. Base64 is encoding, not encryption, so it is only acceptable over HTTPS and is a poor fit for user-facing sign-in.
A behavioral cohort is a group of users defined by an action they did or did not take within a time window, used to compare retention or conversion and find behaviors linked to long-term success.
Behavioral data is a record of what people did: pages viewed, buttons clicked, items bought, features used, and when. It differs from attribute data (who they are) and attitudinal data (what they say). It is usually stored as timestamped events and answers what happened and in what order, not why.
Behavioral segmentation groups customers by what they do, such as purchases, usage frequency, features used and lifecycle stage, rather than who they are. It predicts next actions better than demographics once each segment is a rule: did or did not, in a time window, a number of times.
Benchmarking in business is the ongoing process of comparing your metrics or processes against a reference point — your own past results, competitors, or leaders in other industries — to find performance gaps and decide what to improve.
Benefit segmentation groups customers by the benefit they seek from a product, such as saving time, control or low cost, rather than by who they are. Because wants are hidden, teams usually approximate them with behavior rules, such as inviting teammates for collaboration.
Beta testing is the stage where a nearly finished product or feature is released to a limited group of real users outside the development team, to find bugs, usability problems and missing pieces in real conditions before the general release.
The Bonferroni correction is a method for multiple comparisons: when you run m tests, you call a result significant only if its p-value is below α divided by m, which keeps the chance of any false positive across all tests at or below α.
A bot farm is a coordinated collection of automated accounts, scripts or devices run by one operator to imitate real users at scale, usually to fake clicks, visits, views, installs or followers. It differs from a botnet because the operator owns or rents the machines.
Bot filtering is the step in web analytics that detects automated traffic, such as crawlers, monitors and scrapers, and keeps it out of human reports so visitors, sessions and conversion rates describe real people. It works best when bots are split into their own report instead of silently deleted.
Bot mitigation is identifying automated traffic and applying a policy to it: allow, count separately, slow down, challenge or block. The aim is to reduce harm (fraud, scraping, load, skewed numbers), not to reach zero bots, because every strict rule also stops some real people.
A botnet is a group of internet-connected devices that malware has taken over and that one operator controls remotely. To a website it looks like many unrelated visitors from many countries, so it shows up as a cluster of requests with a shared signature, not as one bad address.
Bottom of funnel (BOFU) is the last stage of the marketing funnel, where a person has already identified the problem and compared options and is deciding whether to buy, sign up or book. BOFU content and metrics focus on that decision: pricing, demos, trials, proof and purchase rates.
The share of visits that end after a single page with no further interaction. A high bounce rate on a blog post can be fine; on a landing page it usually means the promise and the page disagree. In TapCub any tracked event — a click, a form submit, a scroll-depth event — rescues a visit from being a bounce.
Automated traffic: search engine spiders, monitoring tools, scrapers, AI fetchers and abusive scripts. TapCub identifies them with five layers — user agent, headers, client signals, network ownership and request rate — and keeps them out of your human numbers by default.
Browser fingerprinting identifies a browser by combining many attributes it exposes, such as user agent, screen size, time zone, languages, installed fonts and how it renders canvas graphics, into an identifier that survives clearing cookies.
A brute force attack uses software to try many passwords or keys until one works. It needs no vulnerability, only unlimited guesses, so defence limits guesses, slows them down or makes a correct guess insufficient on its own.
Burn multiple is a capital-efficiency metric equal to net cash burn divided by net new annual recurring revenue over the same period. It shows how much a company spends to add each dollar of ARR; lower is better.
Marketing and sales spend divided by the number of new customers it produced. Compare it with LTV to know whether a channel pays back. TapCub computes CAC per channel once you import spend and define a "became customer" event.
CAC payback period is the number of months it takes for the gross profit from a new customer to cover the cost of acquiring that customer. It is calculated as customer acquisition cost divided by monthly revenue per account times gross margin.
A callback is a function passed into other code so that the other code can call it later, when an event happens or a task finishes. It is the basic building block of event handlers and asynchronous code, including SDK error and consent hooks.
Canary testing (a canary release) is a release practice in which a new version is first served to a small share of real production users or traffic, its health is compared with the current version, and it is rolled out further only if it performs as well, otherwise rolled back.
Cardinality is the number of distinct elements in a set. In analytics it usually means how many distinct values a field or property holds (low for device type, high for user IDs or raw URLs); in data modeling it describes one-to-one, one-to-many or many-to-many relationships.
Cart abandonment rate is the percentage of shopping carts, or sessions with an add-to-cart, that end without a completed purchase. It is 1 minus completed orders divided by carts created.
Categorical data is data that places each record in a group or label, such as channel, device, country or plan, instead of measuring an amount. You can count and compare groups, but you cannot meaningfully add or average the values themselves.
A categorical variable is a variable whose values are labels for groups, such as country, plan or error code, not measured amounts. Test it by asking whether adding or averaging the values means anything: ZIP codes and status codes are digits but still categories.
Causation means that changing A makes B change, and B would have been different had A not changed. It needs more than two metrics moving together: A must come first, other explanations must be ruled out, and the strongest proof is a randomised experiment.
The central limit theorem states that the means of sufficiently large independent random samples are approximately normally distributed around the population mean, with standard deviation σ/√n, regardless of the shape of the original data, provided its variance is finite.
Chat analytics is the measurement of live chat conversations: how many start, how fast someone replies, how many end resolved, how visitors rate them, and when they begin. Read the metrics together, because any one of them can look healthy while another fails.
A chat box is the window where a conversation happens: a header, a scrolling list of messages and an input field. On a website it opens from a small button, called the chat bubble; the whole embedded component is the chat widget.
A chat widget is the small chat window embedded in a web page: a launcher button in a corner that opens a conversation with a live agent or bot. Whether it is used depends on three choices: where it sits, what makes it open (nothing, a teaser, a timed invite), and the first line it shows.
The share of users or customers who stop coming back, or cancel, in a period. It is the inverse of retention and the fastest way to see whether growth is real or a leaky bucket. TapCub shows churn alongside the retention grid and lets you segment who left.
CIAM, customer identity and access management, covers how customers register, sign in, manage their profile and control how their data is used. It is IAM aimed at external users, designed for low friction at large scale. Analytics then measures what those identified users do.
A parameter an ad platform appends to a landing URL — gclid, fbclid, ttclid and others — that identifies the click. TapCub reads click IDs from 14 platforms so a visit is attributed to the campaign even when UTM tags are missing.
A click map is a heatmap that shows where visitors click or tap on a page, aggregated across visits, with busier spots in warmer colors. It shows where people clicked, not why they did not, so read it with scroll depth, rage clicks and dead clicks, split by device.
Click-to-install time (CTIT) is the time between a user clicking an ad or tracking link and the app registering the install, usually at first open. A cluster of installs seconds after the click suggests click injection or install hijacking; a flat spread over hours and days suggests click flooding.
Clickstream data is the time-ordered sequence of page views, clicks and other events that a visitor generates while using a website or app, usually stored as one row per event with a timestamp, a session or user ID and properties.
A code freeze is a period when changes to a system's code are stopped or tightly restricted, usually before a major release or a seasonal peak. It does not remove risk, it concentrates it in the release that opens and the one that closes the freeze, so fix your baseline numbers before it starts.
A group of users defined by when they did something — signed up in March, first bought in week 12. Comparing cohorts shows whether a change improved behaviour for new users rather than just shifting the average. Retention grids are cohorts read left to right.
Competitive parity means matching what competitors do or spend, on budgets, prices or features, rather than trying to beat them. It protects the baseline users expect but gives them no new reason to choose you, so it suits defence and not differentiation.
The place in the console where privacy is configured per site: cookieless or identification mode, consent requirements by region, IP truncation, retention, child mode and deletion requests. Presets lock a known-good configuration in one click.
A confidence interval is a range of values that a sample cannot rule out for the true figure. If 120 of 4,000 visitors convert (3.0%), the 95% interval is about 2.51% to 3.58%. The 95% describes the method: repeat the study and about 95 of 100 such intervals cover the truth.
A confounding variable is a third factor that influences both the thing you think is the cause and the outcome you measure, creating a link between them that is partly or wholly not causal.
Connected TV (CTV) is a television connected to the internet, either natively as a smart TV or through a stick, game console or set-top box. Ads can be measured to the impression, but getting results back to a website needs tags and a comparison group.
A consent management platform (CMP) is software that asks visitors for permission to use cookies and tracking, records who consented to which purpose and when, lets them withdraw consent, and passes the consent state to analytics and advertising tags.
Contribution margin is revenue minus variable costs: the amount each sale leaves, after the costs that scale with it, to cover fixed costs and then become profit. Stated in dollars (total or per unit) or as a ratio of revenue.
Conversions divided by the people or visits that could have converted. The denominator matters: per visit, per visitor and per session give different numbers for the same page. TapCub states the denominator on every conversion report.
Conversion rate optimization (CRO) is the systematic practice of increasing the percentage of visitors or users who complete a desired action, such as a purchase or sign-up, by researching what prevents them and testing changes that remove those obstacles.
Measuring visits without storing anything on the visitor's device. TapCub derives a visitor identifier on the server from a secret salt, the site, the IP and the browser, and never stores the IP. No banner is needed for the script itself; the cost is that a new network or device looks like a new visitor.
Correlation measures how closely two variables change together. The correlation coefficient r runs from -1 (one rises as the other falls) through 0 (no straight-line link) to +1 (they rise together). It describes a pattern, not a cause.
Cost per acquisition (CPA) is the average marketing cost of one conversion, calculated as the spend on a campaign or channel divided by the number of conversions it produced, where the conversion is a defined action such as a purchase, sign-up or lead.
CPC (cost per click) is ad spend divided by clicks. Platforms count billed clicks, but your site sees fewer sessions (slow pages, blocked scripts, repeat clicks), so spend ÷ sessions is usually higher than the platform CPC, and that gap is the leak worth checking.
CPM (cost per mille) is what an advertiser pays for 1,000 ad impressions: spend ÷ impressions × 1,000. It prices reach, not results, so a low-CPM channel can still cost more per purchase; compare channels only after converting to cost per session and cost per conversion.
Crawl depth (click depth) is the minimum number of internal links a crawler or user must follow from the homepage to reach a page. Shallow pages are usually discovered and recrawled more often than deep ones.
Credential harvesting is the bulk collection of usernames and passwords by an attacker, through fake login pages, malware, interception or stolen databases, so that the logins can be used or sold.
Cross-device tracking is linking one person's activity across several devices, such as a phone, a laptop and a tablet, so it is analyzed as one person. Deterministic linking uses something the person supplied, usually a login. Without a link, each device counts as a separate visitor.
Cross-site tracking is recognising the same browser or person on websites run by different companies and joining their activity into one record, using third-party cookies, fingerprinting or ids added to links. Browsers block these cross-site mechanisms, not a site's own first-party analytics.
CSAT stands for customer satisfaction score. It is based on one question asked right after a specific experience, such as a support chat, and is usually reported as the share of customers who rated it satisfied (4 or 5 on a 1–5 scale). It measures one interaction, not loyalty.
A call to action (CTA) is the button, link or line of text that tells a person what to do next, such as "Start free trial" or "Download the guide". It has an action, an outcome and sometimes a friction reducer. Measure it by click-through rate and, more importantly, by what happens after the click.
Cumulative Layout Shift (CLS) is a Core Web Vital that scores how much visible content moves unexpectedly while a page is open. It is the largest burst of layout-shift scores in a session window. Good is 0.1 or less.
CUPEDControlled-experiment Using Pre-Experiment Data
CUPED (Controlled-experiment Using Pre-Experiment Data) is a variance-reduction technique for A/B tests that adjusts each user’s outcome using a correlated measurement from before the experiment, which makes the treatment effect easier to detect without biasing it.
A custom dimension is a user-defined attribute, such as plan tier, author or product category, attached to analytics data at event, session, user, account or item scope so reports can be grouped and filtered by it.
Customer acquisition is the process of turning people who have never bought from you into first-time paying customers, a chain from first impression to first purchase where improving results means finding the most expensive stage rather than watching blended cost.
Customer centric means a company starts from what customers are trying to do and makes decisions from evidence of their behavior and conversations, not from internal opinion. It is a decision rule, not a slogan: each decision names the signal behind it.
Data enrichment adds fields to records you already have, such as company, industry or region, using derived, imported or third-party data. Each added field is a claim: store its source, date and confidence, measure usable coverage rather than match rate, and add only what you need and can justify.
A customer data platform (CDP) is software that collects first-party customer data from many sources, merges it into a persistent profile for each person, and makes those profiles and audiences available to other systems such as email, ads and support tools.
Customer effort score (CES) measures how much effort a customer had to put in to get something done, such as solving a problem or completing a purchase. It is based on one question asked right after the interaction, answered on a numbered scale, and it is usually reported as the average score.
A customer health score is a composite number, usually 0–100 or green/amber/red, that combines product usage, customer outcomes, relationship and commercial signals to estimate how likely an account is to renew, expand or churn.
The customer life cycle is the sequence of stages a customer moves through with a brand, from first awareness to purchase, repeat use, loyalty and the eventual end of the relationship; teams use it to decide what to measure and which action fits each stage.
A customer segment is a group of customers who share a definable trait, such as behaviour, location or company size, and who are expected to respond differently from other groups. It is useful only if you would act differently on it.
Customer sentiment is how customers feel about a company, product or interaction: positive, negative, neutral or mixed, and how strongly. It is a process signal that changes during a conversation, unlike satisfaction, which is an outcome rating given after it ends.
Customer value has two directions. Value to the customer is what they get minus what they give up (benefits minus costs, as they see it). Value from the customer is what they are worth to your company over time, usually lifetime value. A business needs both; a big gap between them does not last.
The dark funnel is the part of the buying journey that happens where analytics can't see: private communities, podcasts, word of mouth and AI answers. It cannot be tracked directly, only estimated from self-reported sources, direct traffic and branded-search changes.
Dark social is traffic from links shared in private channels, such as messaging apps, email, Slack or direct messages, that reaches your site without a referrer, so analytics tools usually count it as direct traffic.
A dashboard is a single screen that shows the few most important metrics for a goal side by side, so a person can tell at a glance whether things are on track. It is for noticing problems quickly, not for displaying all of your data.
Data activation is the practice of moving modeled data from a central store such as a data warehouse into the operational tools where teams act on it, including ad platforms, email tools, CRMs and support desks, and keeping it in sync so that a finding becomes a live audience or field.
Data aggregation is the process of combining many raw records into summary values — counts, sums, averages, medians, percentiles — usually by grouping them over a time period or a dimension such as channel or device.
A data catalog is an organized inventory of an organization's data assets, built from metadata, that answers four questions about each asset: what data exists, where it lives, who owns it, and whether it can be trusted for a given use.
A data clean room is a controlled environment where two or more parties can join and analyse their first-party data, usually matched on hashed identifiers, under rules that let each side see only aggregated results rather than the other’s raw, person-level records.
Data collection is the planned, systematic gathering of information about the variables you care about, in order to answer a question. Before collecting, decide what decision each field changes, how it will be recorded and on what basis.
Under the GDPR, a data controller decides the purposes and means of processing personal data, and a data processor processes it on the controller's behalf and instructions. Roles follow what a party actually does, so a processor that sets its own purposes becomes a controller for that processing.
Data conversion changes data from one format or structure to another, such as a file format, character encoding, unit, data type or time zone, without changing its meaning. It succeeds only when both sides agree on meaning: a conversion can write every row and still move numbers onto the wrong day.
A data dictionary is a reference document or table that lists every field in a dataset or system and records, for each one, its name, plain-language meaning, data type, allowed values, source and owner.
Data discovery is finding out what data you have, where it lives, how good it is and who can use it, before you analyze it. In practice it is four steps: inventory, classify, assess quality and record an owner and definition for each item.
Data federation is a way to query data held in several separate systems as if it were one database, without copying it into a central store first. A federation layer splits the query, runs each part where the data lives, and combines the results at query time.
Data filtering in analytics is excluding records, such as your own team's visits, bots and test traffic, according to rules, either before storage (collection-time) or when a report is built (report-time), so that metrics describe the audience you mean to measure.
Data governance is the set of roles, rules and processes that decide who owns each data set, how it is defined, who can access it, how its quality is checked, and how long it is kept, so that data can be trusted and used safely.
Data granularity is the level of detail a dataset stores: what one row represents, from one row per event with timestamp, user and properties (fine) to one row per day of totals (coarse). Fine data rolls up to coarse; coarse never splits back, so grain is set at collection time.
Data integration combines data from different source systems into a unified, consistent form for analysis and decisions. Done well, it gives one answer per question: every team sees the same number because sources share one identity and event model.
Data intelligence is the practice of turning data into conclusions people can trust and act on: understanding what data exists, where it came from and its quality, then using it to make a decision and checking the result.
Data latency is the time between an event happening, such as a page view or purchase, and that event being available to query or see in a report. Data freshness is the related question of how recent the newest data in a report is.
In web analytics, a data layer is a JavaScript object or array on the page, such as window.dataLayer, where the site writes structured facts about the page, the visitor and their actions so tag managers and analytics tools can read them without scraping the HTML.
The data lifecycle, or data cycle, is the path data follows from the decision to collect it to its deletion: plan, collect, process, store, use, share, retain and delete. There is no single standard model; the stages matter because each one needs an owner and a rule, especially the last.
Data lineage is the documented path of data from its sources through every transformation to the reports and systems that use it, at table, column or business-metric level, used for root-cause analysis, impact analysis, audit and compliance.
Data masking replaces sensitive values with altered ones, such as asterisks, a shortened form or a realistic fake, so data stays usable while the real value is hidden. It can be applied at capture, in a prepared copy or only when someone views the data, and each leaves something different stored.
Data maturity is how far an organization has moved from collecting numbers to deciding with them: data that is trusted, findable and actually used. It is usually graded in five levels, and the level is set by the weakest area, not by how many tools you own.
Data minimization is the privacy principle that you should collect, use and keep only the personal data that is adequate, relevant and necessary for a specific purpose, in the least identifying form and for no longer than needed.
A data model describes what a dataset contains and how it connects: the entities you keep records about, the attributes of each, and the relationships between them. In analytics, the core of the model is how events, users and their properties link together.
Data observability is the practice of continuously measuring the health of data, such as its freshness, volume, schema, value distribution and lineage, so that problems are detected early and traced to the step where they started.
Data preparation is the process of collecting, cleaning, transforming and validating raw data so it is accurate, consistent and structured for analysis, reporting or machine learning.
Data processing is the series of steps that turns raw data into usable results: collecting it, cleaning and checking it, transforming it, storing it and presenting the output. In analytics it is what turns raw hits into visitors, sessions and reports.
Data redaction is removing or hiding sensitive values in a record, such as a card number in a chat message, while leaving the rest readable. The value is blacked out, deleted or replaced with a placeholder or partial value, ideally before the record is stored.
Data segmentation is the practice of dividing a dataset into smaller groups by one or more conditions, such as who people are, where they are, what they did or how they arrived, so each group can be examined and compared on its own.
A data silo is data held by one team, tool or system that the rest of the organization cannot easily reach. The usual result is that one person is counted as several, because no shared key links their records in web, app, support and billing systems.
A data subject request (DSR) is a person's request to exercise a right over the personal data an organization holds about them: to access, correct, delete, restrict, move or object. A DSAR is the access-only kind. Finding the person's data across systems is usually the hardest part.
A data table organises data in rows and columns: each row is one record, each column is one field, and each cell holds one value. A table is only reliable when you can say what one row is (one order, one event, one user), because every count and sum depends on that choice.
A data type is a classification that tells a program, database or analyst what kind of value a field holds, such as an integer, a decimal number, text, true/false or a date, and therefore which operations are valid on it.
Data unification is the process of matching, deduplicating and reconciling records about the same person, account or thing from several systems so each appears once, with consistent fields and one stable identifier, for example one profile instead of separate web, app and chat records.
Data validation is checking data against defined rules, such as type, range, format and required fields, before it is stored or used. It shows that data is well formed, not that it is true, and it works best at the source and at intake, before wrong values reach reports.
Data visualization is the graphical representation of data, such as charts, graphs, maps and dashboards, designed so people can see patterns, trends, comparisons and outliers faster and more accurately than in a table of numbers.
Daily, weekly and monthly active users — the number of distinct people who did a qualifying action in that window. The DAU/MAU ratio is a stickiness measure: 0.5 means the average monthly user shows up every other day. TapCub lets you define "active" per product.
A dead click is a click or tap on a web page that produces no visible response, such as no navigation, content change or triggered request, usually because an element looks clickable but is not, or because its code fails or responds too slowly.
A deep link is a link that opens a specific screen inside a mobile app, such as a product page or an order, instead of only launching the app's home screen. It makes the destination of an ad, email or push message match what it promised.
A demand-side platform (DSP) is software that advertisers and agencies use to buy digital ad impressions from many sources through one interface, deciding in real time what to bid. Its reported conversions need reconciling with your own analytics.
Descriptive statistics summarize a data set as it is: counts and frequencies, center (mean, median, mode), spread (range, standard deviation, interquartile range) and shape (skewness, percentiles), without generalizing beyond the data.
A device ID is an identifier used to recognize a specific device over time. It can be an operating-system advertising ID (IDFA, Android advertising ID), a vendor- or app-scoped ID, or a random ID generated by an app or SDK at install.
Diagnostic analytics is the type of analysis that explains why something happened, by breaking a change in a metric down by segments, time and related events to find the factors that caused it.
Digital analytics is the collection and analysis of data from all of an organization’s digital touchpoints (websites, apps, ads, email, social and support) to understand customer behavior across channels and improve business results.
Digital experience (DX) is the overall impression a person forms from all their interactions with your digital channels: site, app, chat, email and checkout. It is judged from behaviour, speed and reliability, and feedback together.
Digital touchpoints are the online interactions between a customer and a brand, such as a search result, an ad click, a product page visit, a live chat conversation or a marketing email, at any stage before, during or after a purchase. In analytics, each is a step on a conversion path.
Direct response marketing is advertising designed to get one specific action right away, such as a click, purchase or signup, and to trace every response back to the ad that caused it. It is judged by arithmetic: conversion rate, cost per acquisition and return on ad spend.
Discrete data is quantitative data obtained by counting, which can take only separate, specific values, usually whole numbers, such as the number of orders, page views or items in a cart. It contrasts with continuous data, which is measured and can take any value in a range.
Dogfooding (eating your own dog food) is the practice of a company using its own product internally for real work, to find defects and usability problems early and build first-hand understanding of the customer experience.
DOMContentLoaded is a browser event fired when the initial HTML document has been fully parsed and deferred scripts have executed, without waiting for images, iframes and other subresources, which complete at the later load event.
Drop-off rate is the percentage of users who leave a process at a given step instead of continuing to the next, calculated as (users at the step − users at the next step) ÷ users at the step × 100.
Dwell time is the length of time a searcher spends on a page after clicking a search result and before returning to the search results. Only search engines can measure it; site analytics can only approximate it with time on page and engagement.
A dynamic URL is an address whose page is generated from a query at request time, with the query carried in parameters after a ?. Dynamic URLs don't hurt SEO by themselves; the problem is parameter combinations exploding into duplicate pages, in search and in analytics reports alike.
Earned media is publicity a brand gets without paying for the placement, such as press coverage, independent reviews, customer posts, backlinks and word of mouth. Unlike paid and owned media, you do not control it, and it is the hardest to attribute.
Effect size quantifies how large a difference or relationship is, independent of sample size. It can be expressed in the metric’s own units (e.g. +0.6 percentage points) or standardized, as with Cohen’s d, Cohen’s h or a correlation coefficient.
Embedded analytics is the practice of putting reports, dashboards and metrics inside the software people already use — a product, customer portal or internal workflow — instead of sending them to a separate BI tool, with access controlled per user or customer.
An engaged session is a session that lasted at least 10 seconds, included at least two pageviews, or completed a key event. Engagement rate is engaged sessions divided by all sessions.
Entity resolution is the process of deciding which records in one or more datasets describe the same real-world person, company, product or address, then linking or merging them. It works with exact rules, similarity scores or learned models, and is also called record linkage or deduplication.
The first and last page of a visit. Entry pages tell you what search engines and campaigns actually land people on; exit pages show where attention runs out. An exit page is not a problem by itself — the order confirmation is everyone's exit page.
Error rate is the share of errors within a chosen total over a time window: errors ÷ denominator × 100. The denominator — pageviews, sessions, visitors or requests — decides what the number means, so name it before comparing rates or setting alerts.
ETL stands for extract, transform, load: a process that copies data out of source systems, cleans and reshapes it, and loads the result into a destination such as a data warehouse. An ETL pipeline is the automated set of jobs that does this repeatedly.
An event is something that happened — "added to cart"; a property is a detail attached to it — product, price, currency. Good tracking uses few events with rich properties rather than hundreds of near-duplicates. TapCub's tracking plan keeps names and properties consistent across platforms.
Event analytics is analysis built on events: records of individual actions, each with a name, a timestamp, the person or device that did it and properties that describe it. It counts, compares and breaks down those actions over time.
An event schema is the written contract for one analytics event: its name, the properties it carries, each property's type, and which ones are required — agreed before any data is sent and checked as data arrives.
Experimenter bias (observer-expectancy effect) is a distortion of research results caused by the researcher’s expectations influencing how an experiment is designed, conducted, measured or interpreted, often unconsciously.
Exponential growth is growth by a constant percentage per period, so the absolute increase gets larger each period. It follows N(t) = N₀ × (1 + r)ᵗ, where r is the growth rate per period.
Exponential smoothing is a time-series method that averages past observations with exponentially decaying weights, so recent data counts most. One constant, α (0–1), sets the trade-off: high α follows new data fast, low α stays steady. The smoothed value doubles as a one-step forecast.
The false discovery rate (FDR) is the expected proportion of false positives among all results declared significant. The Benjamini-Hochberg procedure controls it, so that when many hypotheses are tested at once, on average no more than a chosen share (for example 5%) of the discoveries are false.
Feature flagging wraps a feature in a conditional switch that can be changed at runtime, so code can be deployed first and released later, to a small share of users, and rolled back without a new deploy.
First Contentful Paint (FCP) measures the time from the start of navigation until the browser renders the first piece of content from the page: text, an image, a non-white canvas or an SVG. Good is 1.8 seconds or less.
First input delay (FID) measured the time from a user’s first interaction with a page to when the browser could begin processing event handlers for it. It was a Core Web Vital until 12 March 2024, when Interaction to Next Paint (INP) replaced it.
How long a visitor waits between starting a chat and receiving the first reply from a person. It is the metric visitors feel most. TapCub reports the average and median, per agent, and counts conversations the AI assistant handled on its own separately.
A first-party cookie is a cookie set by the website shown in the address bar, either by its server or by a script running on its pages. It is used for logins, carts, preferences and first-party analytics, and is sent only to that site.
A business flywheel is a self-reinforcing loop in which growth output feeds back into input: customers create referrals, content or supply that acquire the next customers, so each turn makes the next easier. Jim Collins named it in Good to Great (2001).
Frustration signals are click and interaction patterns recorded in a session that suggest a visitor is stuck, such as rage clicks, dead clicks and clicks followed by a script error. They are leads, not conclusions: rank pages by them, then watch sessions to find the cause.
FTUE (first-time user experience) is everything a new user sees and does in their first session with a product or game. It is good when the user reaches the step where the product delivers its core value in that session, which you can only measure once that step is defined as an event.
An ordered set of steps — view product, add to cart, start checkout, pay — with the number of people who reached each one. The drop between two steps is where to look. TapCub funnels can be built from any events, split by segment, and read within one data set so the steps are comparable.
The three privacy regimes most sites meet: the EU's General Data Protection Regulation, California's Consumer Privacy Act and China's Personal Information Protection Law. They differ in detail but share the ideas of lawful basis, notice, access, deletion and limits on transfers. The compliance centre maps settings to each.
Geofencing draws a virtual boundary around a real place on a map and triggers an action when a device enters it, leaves it or stays inside. Geofencing marketing uses that trigger to message people near a store or event. It works at roughly 100 to 150 m precision and needs consent.
The total value of orders placed in a period, before refunds, discounts and fees. Marketplaces and shops report it as the top line; it is not revenue. TapCub sums the value property of your order events and shows refunds separately.
A goal is a visitor action you define in advance and count as a conversion, such as a purchase, a sign-up or a thank-you page view. A key event is the newer name for an important event. Keep goals to a few outcomes you would pay for, with precise rules, or the conversion rate stops meaning anything.
Gross margin is the share of revenue left after the direct cost of what you sold: (revenue − cost of goods sold) ÷ revenue. A 40% margin keeps 40 cents of each dollar before marketing, salaries and tax. For ads, it sets the break-even ROAS: 1 ÷ margin.
Gross revenue retention (GRR) is the percentage of recurring revenue from a starting group of customers that remains after a period, counting losses from cancellations and downgrades but excluding any expansion revenue, so it can never exceed 100%.
A growth loop is a closed cycle in which the output of one group of users, such as invites, public content or revenue, becomes the input that brings in the next group, so growth feeds itself instead of relying only on new spend.
GRP, or gross rating points, measures how much advertising exposure a campaign delivers: reach (%) times average frequency, or impressions as a percentage of the target population. It can exceed 100 and measures exposure, not results.
Guardrail metrics are measures a team commits to protect while it optimises a primary metric. A change that lifts the primary metric but pushes a guardrail past an agreed limit, such as slower pages or more refunds, is not shipped until the guardrail is fixed.
The happy path is the default, error-free sequence of steps a user takes to accomplish a goal in a product, with valid input and no detours. It is the main success scenario that design, testing and analytics start from.
A HAR (HTTP Archive) file is a JSON log of the network requests a browser made while loading a page, including URLs, headers, cookies, bodies, status codes and per-request timings. It is exported from browser developer tools for troubleshooting.
Turning a value into a fixed-length string that cannot be reversed. TapCub turns the IP address and browser into a visitor ID with a keyed hash and keeps only the result; in strict mode the key changes daily, so the same person cannot be followed across days. Offline conversion imports accept emails already hashed with SHA-256.
Header bidding is a way for publishers to sell ad slots by asking several demand partners to bid at the same time, before the ad server is called, so the bids compete with the ad server's own demand instead of being tried one by one in a waterfall.
A page overlay that shows where visitors click and how far they scroll, aggregated across many visits. It answers "do people see the button" faster than a funnel. TapCub heatmaps contain no typed text and no identifiers.
A holdout group is a randomly chosen part of an audience that deliberately does not get a campaign, feature or change, so its results can be compared with everyone else's. The gap is what the change caused; without a holdout you cannot separate it from what would have happened anyway.
The hook model is Nir Eyal's four-phase loop - trigger, action, variable reward, investment - that describes how products become habits. It is a lens for diagnosing repeat use, not a growth recipe: each phase has to be visible in your data before you act on it.
The referrer is the URL of the page that led to the current request, sent by the browser in the HTTP Referer header (spelled with one r) and available to scripts as document.referrer. Analytics tools use it to tell where a visit came from.
Hypothesis testing is a procedure that uses sample data to decide whether a difference is larger than chance alone would plausibly produce. You fix a null hypothesis, a significance level and a sample size first, then compute a p-value and reject the null only if it falls below that level.
The call that tells TapCub "this visitor is user 4821", and the rules that merge the anonymous visits before sign-in with the identified ones after. Identity resolution is what makes retention and LTV per person possible — and it is opt-in, with consent handled in the compliance centre.
An identity graph stores identifiers such as visitor ids, device ids and login ids as nodes, and the rules that link them as edges. Its value depends on whether those edges can be explained: a login link can be audited, a probabilistic link can quietly merge different people.
Identity resolution is the process of recognising that different identifiers, such as anonymous visitor ids, device ids, login ids and email addresses, belong to the same person, and linking them into one profile.
Impact mapping is a planning technique that links a goal to the people who can affect it, the behavior changes you need from them and the deliverables that could cause those changes. Four layers (why, who, how, what) make every feature justify itself against the goal.
Impressions are the number of times an ad, post or search result was shown, whether or not anyone clicked. Each platform counts differently: Search Console can count a result the user never scrolled to, Meta counts an ad first on screen. Check whose rule applies; never add them up.
An in-app message is a message shown inside an app while the person is using it, as a banner, modal, card or image. It needs no notification permission but reaches only people who open the app, so it suits guiding active users rather than bringing back lapsed ones.
Incrementality is the additional outcome (conversions, revenue, sign-ups) caused by a marketing activity, measured as the difference between a group exposed to it and a comparable group that was not.
Index bloat is a condition where a search engine has indexed many more of a site’s URLs than the site has valuable pages, typically from parameter variants, internal search, thin archives and duplicates, wasting crawl resources and diluting quality signals.
Indexability is a page’s ability to be added to a search engine’s index after it is crawled. It depends on a successful response, no noindex directive, a self-referencing canonical, renderable content and sufficient value; crawlability is the earlier step of being fetchable at all.
A pre-built set of metrics, dashboards, funnels and event names for one sector — e-commerce, SaaS, education and seven more. Installing one gives you the reports a sector actually uses before you write a tracking plan.
Inferential statistics use data from a sample to draw conclusions about a larger population or process, quantifying uncertainty with tools such as confidence intervals, hypothesis tests and p-values.
An intelligent virtual assistant (IVA) is software that understands questions in everyday language and responds with answers or actions. In customer service, a good IVA is judged by whether it knows when not to answer and hands off to a human instead.
Interaction to Next Paint (INP) is a Core Web Vital that measures responsiveness: the time from a click, tap or key press until the browser paints the next frame, taken from close to the slowest interaction of the visit. Good is 200 milliseconds or less.
The interquartile range (IQR) is the difference between the third quartile (75th percentile) and the first quartile (25th percentile) of a data set. It measures the spread of the middle 50% of values and is resistant to outliers.
Interstitial ads are full-screen ads shown at a break between content, such as between game levels, articles or screens. They earn well because they take the whole screen, and cost retention because they interrupt; judge them on both.
IP anonymization (IP truncation or masking) zeroes the last part of an IP address, such as the final byte of IPv4 or the last 80 bits of IPv6, before it is stored or looked up. It lowers precision. Whether the result is still personal data depends on what else can identify the person.
A JavaScript error happens when the browser cannot parse or run a script as written, so the code that failed stops and the feature it powered may not work. Visitors usually see no message. What matters is which errors stop a task and how many sessions they touch, not how often they occur.
Key driver analysis ranks which factors, such as early behaviors or survey ratings, are most strongly associated with an outcome like retention or satisfaction. It shows association, not cause, so a top driver should be checked for confounders and confirmed with a holdout before you act on it.
The articles and answers your AI assistant is allowed to use. When a visitor asks a question, TapCub finds the matching snippets, sends only those with the question to the model, and labels the reply as AI. Corrections from agents flow back into the base.
A landing page has two meanings: in marketing, a page built for a campaign or ad; in analytics, simply the first page of a session. Reports use the second meaning, so any page, a blog post, the pricing page or the home page, can be a landing page. Compare them only when the traffic is comparable.
Largest Contentful Paint (LCP) is a Core Web Vital that measures the time from when a page starts loading until the largest image, video poster or text block in the viewport has rendered. Good is 2.5 seconds or less.
Last click attribution is a model that gives all of the credit for a conversion to the last touchpoint a customer clicked before converting, and none to any earlier interaction, so it shows which channel closed the sale but not which channels created the demand.
A linear attribution model is a rules-based multi-touch model that gives every touchpoint on a conversion path the same share of credit: 1 divided by the number of touchpoints, so four touches get 25% each and three get a third each. It shows who took part, not who should get the budget.
Live chat is a customer communication channel that lets website or app visitors exchange real-time text messages with a company's support or sales agents, or an AI assistant, through a chat window embedded in the page.
The revenue a customer brings over their whole relationship with you. Compared with CAC it tells you how much you can afford to spend to win one. TapCub computes LTV from revenue events per identified user, by cohort and by acquisition channel.
The margin of error is the range above and below a sample estimate within which the true value is expected to fall at a given confidence level. For a proportion it is z × √(p(1 − p)/n); it shrinks as the sample size grows.
A marketing qualified lead (MQL) is a prospective customer that marketing has judged worth a sales follow-up, because the lead fits the target market and has shown buying signals such as visiting the pricing page, attending a webinar, or requesting a demo.
Martech (marketing technology) is the software and tools marketers use to plan, execute, measure and optimize marketing, such as analytics, CRM, marketing automation, CMS, customer data platforms and chat. A team’s combined set of tools is its martech stack.
Media mix modeling (MMM) is a statistical method that estimates how much each marketing channel contributed to sales by regressing weekly sales on channel spend, seasonality and other controls. It needs no user-level tracking, but it needs about two years of weekly data with real swings in spend.
A metric is a quantitative measure of something over a defined period, such as visitors, sign-ups, conversion rate or average order value. A KPI is a metric chosen to track a specific goal; a dimension is an attribute, such as country or page, that splits a metric.
Micro interactions are small, single-task moments in an interface, such as a button's loading state or a toggle sliding on: a trigger, rules, feedback, and loops or modes. Judge one by behavior, not looks: missing feedback shows up as repeated clicks, pauses and abandoned steps.
The minimum detectable effect (MDE) is the smallest true difference between variants that an experiment is designed to detect at a chosen significance level and power. A smaller MDE needs a larger sample; with the same traffic, the test has to run longer.
A minimum viable product (MVP) is the version of a new product that delivers just enough value to real users to test the riskiest assumption about it, with the least effort, so the team can learn whether to continue, change direction or stop.
Mobile app analytics is the collection and analysis of data about how people use a mobile app (starts, screens, events, retention, purchases) and how stable and fast it is, usually gathered by an SDK inside the app.
Mobile first is an approach that designs and builds for small screens and constrained conditions first, then enhances for larger screens; in SEO, mobile-first indexing means Google uses the mobile version of a page for indexing and ranking.
A mobile measurement partner (MMP) is an independent company that matches ad clicks and impressions to app installs, credits one ad network per install and filters fraud. In-app behaviour analysis is a separate job, and the two only line up when they use the same event names.
A modal window is a dialog that appears over the page and blocks the content behind it until the user responds, putting the interface into a special mode. Use it for decisions that cannot be skipped, not for nonessential requests such as newsletter prompts.
A rule that watches a metric and notifies you when it crosses a threshold or deviates from its usual pattern — conversions down 30% against the same hour last week, for example. Alerts go to email, WeCom, DingTalk, Feishu or a webhook, with a ready-made template for Slack.
Month over month (MoM) growth is the percentage change in a metric from one calendar month to the previous one: this month minus last month, divided by last month.
Monthly active users (MAU) is the number of distinct users, accounts or devices that did at least one qualifying action in a 30-day window. What counts as active, who is counted and which window is used all change the number.
A moving average (rolling average) is a series of averages, each taken over a fixed number of the most recent values, used to smooth short-term noise and show the underlying trend in time-series data.
Monthly recurring revenue (MRR) is the predictable subscription revenue a business earns per month, with every active plan converted to its monthly value and one-off charges left out.
A multi-armed bandit is a testing approach in which traffic is allocated adaptively: options that perform better receive more traffic as results come in, balancing exploration (learning which option is best) against exploitation (using the best option found so far).
Native advertising is paid advertising designed to match the form and function of the content around it, such as a sponsored article in a feed or a promoted result in search. It is legitimate when readers can tell it is an ad, which US regulators expect a clear label to show.
A native app is a mobile application built for one specific operating system using its own languages, SDKs and UI components (Swift for iOS, Kotlin for Android), installed from an app store with full access to device features.
Net revenue retention (NRR) is the recurring revenue you earn now from a group of existing customers, divided by what the same customers paid at the start of the period, including expansion and subtracting contraction and churn. New customers are excluded.
A north star metric is the single metric a company chooses because it best captures the value customers get from the product and leads long-term revenue. It is paired with a few input metrics that teams can influence directly.
The novelty effect is a temporary lift (or dip) in metrics caused by users reacting to something being new — curiosity-driven trials inflate early usage, then decay toward a steady state. Judge a launch by days since each user's first exposure, not by its opening days.
An opportunity solution tree (OST) is a product-discovery map, popularised by Teresa Torres, linking one desired outcome to the customer opportunities that could drive it, the solutions for each opportunity, and the experiments that test them.
An ordinal variable is a categorical variable whose values have a meaningful order but no equal spacing, such as a 1–5 satisfaction score, plan tier or priority level. Because the gaps are unequal, an average can mislead: summarize it with the distribution, the median and the top-two-box share.
OTT (over-the-top) is video, audio or messaging delivered over the internet instead of through a cable, satellite or broadcast provider. It describes the delivery route, not the screen; in marketing the hard part is measuring results and overlap.
An outbound link (outlink) is a hyperlink on your site that points to a URL on a different domain. It is both an SEO signal and a user action: what matters is which page visitors left from, which link they clicked, and whether the exit was good or bad.
Owned media is content a company publishes on channels it controls, such as its website, blog, app and email list. Its value is control, not size: judge it by how many people return and which goals it produces, and use it to hold the people paid and earned media bring in.
P-hacking is the practice of running many analyses, such as peeking, adding metrics or segments, changing windows or excluding data, and reporting only those that reach statistical significance, which inflates the false positive rate far above the nominal level.
A p-value is the probability of getting a result at least as extreme as the one you observed, assuming there is really no difference or effect (the null hypothesis). A small p-value means the data would be surprising if nothing were going on.
Page experience is Google’s term for aspects of a page that affect how users experience it beyond its content: Core Web Vitals, HTTPS, mobile usability, the absence of intrusive interstitials and an unobtrusive ad experience.
Pages per session is the average number of pageviews in a session, calculated as total pageviews divided by total sessions for the same period and filters.
One load of one page by one browser. A person reading five pages creates five pageviews. It is the oldest web metric and still the right one for content sites and ad inventory, as long as bots are filtered out first.
Paid search is advertising shown in search engine results for keywords you bid on, usually paid per click. Its cost is read in the ad platform and its results on your site, so tag every paid click with UTM parameters or a click ID, or paid visits land in organic search and skew both.
A painted door test (fake door test) is a demand test in which users see a realistic entry point, such as a button, menu item, pricing option or landing page, for a feature that has not been built, and the share who try to use it is measured before any build decision.
The sequences of pages or events people actually follow, drawn as a flow from a starting point. Where funnels test a path you expect, path analysis shows the ones you did not. Use it to find the detour before the drop.
A paywall is a barrier that asks people to pay or subscribe before they can read content or use a feature. A hard paywall blocks everything, a soft one leaves some content free and a metered one allows a set number of free uses first. Judge it by reach, tap rate and completion rate.
PII (personally identifiable information) is any information that can identify a specific person, either on its own (a name, email address, ID number) or when combined with other data (a ZIP code plus birth date plus gender).
Pipeline velocity is the revenue a sales pipeline is expected to generate per day: qualified opportunities × average deal value × win rate, divided by the sales cycle length in days. Improving any one of the four inputs raises it.
PMF stands for product-market fit: the degree to which a product satisfies strong demand in a specific market, so customers keep using it, pay for it and recommend it. It is not a single moment but a set of signals, such as a flattening retention curve and growing organic demand.
Position-based attribution is a multi-touch model that gives 40% of a conversion's credit to the first touch, 40% to the last touch and splits 20% among the touches in between. It is also called U-shaped attribution. With one touch it gives 100%; with two, 50% each.
In ads and app marketing, a postback is a server-to-server request sent automatically after an install, sign-up or purchase, so another system, often an ad network, learns about it. It is not the event in your own analytics; it is a notification about a result an attribution tool computed.
In product analytics, a power user is someone who uses a product much more often or more deeply than typical users, for example active on most days of the month or using the core feature many times a week. Teams usually define them with a threshold on active days or key actions.
Privacy-first analytics is measurement designed to collect only the data needed to answer stated questions, keep it no longer than needed, and avoid linking it to a person unless that is the purpose. It does not mean collecting nothing: the test is that every field has a reason.
A chat message the site opens itself when a rule matches — for example, a visitor who has been on the pricing page for 20 seconds. Done well, it answers the question the visitor was about to leave with; TapCub limits invites so they do not become a nuisance.
Product adoption is the process by which users move from learning a product exists to using it as a regular part of their routine. Measure it with three signals: they used it, they used it again, and it became a habit.
Product analytics is the practice of recording what people do inside a digital product, as events tied to people over time, and analysing it to see where they finish a task, where they stop and who comes back. It answers questions about behaviour, not only how much traffic arrived.
Product differentiation is making your product different from the alternatives in a way a chosen group of buyers can perceive, put in words and actually uses. It gives them a reason to choose you; a longer feature list does not by itself.
A product engagement score (PES) is a composite 0–100 metric that averages adoption (breadth of core feature use), stickiness (how often users return, e.g. DAU/MAU) and growth (new and returning versus lost users) to summarize product usage health.
Product experience (PX) is the sum of a user's interactions with a product itself, including onboarding, core workflows, performance, reliability and in-product help, and how well those interactions deliver the value the user came for.
A product hypothesis is a testable statement that predicts how a specific product change will affect a specific user behavior or metric for a segment, why, and by how much within a time frame.
Product intelligence combines product usage data with revenue, account, support and feedback data to explain behavior and decide what to build, fix and promote.
The product life cycle is the sequence of stages a product passes through in the market, usually introduction, growth, maturity and decline, each with its own sales pattern, competition and priorities.
Product operations (product ops) is a function that supports product teams with reliable data and insights, a structured flow of customer feedback, and shared processes and tools, so that product managers spend their time on decisions rather than on gathering information.
A product qualified lead (PQL) is a free or trial user, or account, that has reached usage signals showing real value and buying intent, such as inviting teammates or hitting a plan limit. It is qualified by what they did in the product, not by form fills.
A product tour is a short guided sequence inside a product that walks a user through a few key steps, using pointers, tooltips or modals. Judge it by how many users complete it and reach the target action, not by how often it is shown.
Product-led growth (PLG) is a go-to-market strategy in which the product itself is the main way customers discover, try, adopt and buy, typically through a free plan or free trial, with sales stepping in later for larger accounts.
Product-led sales (PLS) is a go-to-market model in which a self-serve product is the main way customers start, and a sales team engages the accounts whose product usage shows buying intent, prioritised by signals such as team growth, usage limits and key feature adoption.
Programmatic advertising is the buying and selling of ad space through software instead of manual negotiation. An auction or pre-agreed deal decides who gets each slot as the page loads; the advertiser's result on its own site is measured separately.
Progressive delivery is releasing a change to users in stages, such as 1%, 5%, 25% and 100% of traffic, and moving to the next stage only when measured gates (error rate, latency, a business metric) stay healthy, with traffic returned to the old version if they fail.
Pseudonymization is the processing of personal data so that it can no longer be attributed to a specific person without additional information, which is kept separately and protected (GDPR Art. 4(5)). Pseudonymized data is still personal data, unlike anonymized data.
A push notification is a short message a server sends to a user's phone or browser without the user asking for it at that moment. It shows up on the lock screen or in the notification drawer even when the app is closed, but only if the user has allowed notifications.
A qualified lead is a lead, a person or company that has shown interest, that has been judged against written criteria to be a good fit and ready for the next step in the sales process. Common stages are marketing qualified (MQL), sales qualified (SQL) and product qualified (PQL).
Qualitative data is non-numerical information that describes qualities, opinions, reasons and experiences, recorded as words, images, audio or categories. It answers why and how, where quantitative data answers how many and how much.
Quantitative data is information recorded as numbers that count or measure something, such as visitors, orders, load time or order value. A number needs a definition and a denominator before it can be compared, and discrete counts and continuous measures need different summaries.
A rage click is a burst of rapid, repeated clicks or taps on the same spot, a sign the visitor expected something to happen and it did not. Tools detect it with a click-count, time and distance threshold, such as three clicks within a few hundred milliseconds.
Raw data is data in the form it was originally collected, before cleaning, aggregation or transformation. In analytics it usually means individual event records, from which reports and metrics are calculated.
Real-time bidding (RTB) is a method of buying ad impressions in which each impression is auctioned the moment it becomes available, in under a second. The winning bid is a price for an impression, not for a visit or a conversion.
Who is on the site right now, which pages they are on and where they came from, updated every few seconds. It is the fastest check that a launch, a campaign or a fix is working — and in TapCub it is the same list the chat inbox uses to start a conversation.
Recency bias is the tendency to give recent events or data more weight than older ones when judging a trend or making a decision, even when the recent data is a small or unusual sample.
A redirect chain is a sequence of two or more HTTP redirects between a requested URL and its final destination (A → B → C). Chains add latency, waste crawl resources and can drop query strings; a chain that returns to an earlier URL is a redirect loop.
A referral source is the external page, site or app a visitor was on just before arriving at your site, as reported by the browser's referrer. It is not the same as a channel, which is a bucket your analytics tool assigns, and it is often missing.
Regression to the mean is the statistical tendency for a measurement that is unusually high or low, partly because of chance, to be followed by a measurement closer to the average, even when nothing has changed.
The share of customers who bought more than once in a period. It is the commerce version of retention and the main lever on LTV when acquisition gets expensive. TapCub computes it per cohort and per first product bought.
The share of conversations that were marked as resolved. Read it next to first response time: fast replies that do not solve anything make one look good and the other bad. TapCub shows it site-wide, per agent, and for conversations the AI resolved alone.
Retargeting shows ads to people who already visited your site or app but did not buy, using a list built from what they did. Results depend mostly on list quality: who is on it, who is excluded and how often they see ads. Only a comparison with a held-out group shows whether it added sales.
The share of a cohort that comes back after a given time. Day-1 retention is the share that returned the next day; day-7 and day-30 follow. The retention grid shows every cohort as a row so you can see whether newer cohorts hold on longer.
A retention curve is a chart of the percentage of a user cohort still active after each day, week or month since they started; a curve that flattens shows a stable base of users who keep getting value.
Revenue churn is the percentage of recurring revenue lost in a period from cancellations and downgrades, relative to revenue at the start. Gross revenue churn counts only losses; net revenue churn subtracts expansion and can be negative.
Reverse ETL copies data out of a data warehouse and loads it into operational tools such as a CRM, email platform or support desk, so the calculated results are available where teams work.
Scoring customers on Recency, Frequency and Monetary value, then grouping them — champions, at risk, hibernating. It is the quickest useful segmentation for a shop and the basis for who to invite to chat or email first.
Return on investment is profit divided by cost; return on ad spend is revenue divided by ad cost. ROAS of 3× means every dollar of ads brought three dollars of revenue — not profit. TapCub shows channel ROI against imported spend so Newsletter at 6.2× and TikTok at 0.8× sit on one chart.
Root cause analysis (RCA) is a structured way to work back from a problem to the underlying reason it happened, so the fix removes the source instead of the symptom. For a metric drop, it means finding where and when the change started, then testing the candidate cause.
A ROPA (record of processing activities) is the written inventory GDPR Article 30 asks controllers and processors to keep: why personal data is processed, whose data and which, who receives it, where it goes, how long it is kept and how it is secured. The two roles keep different records.
The Rule of 40 is a benchmark for software companies: revenue growth rate plus profit margin, both as percentages, should add up to 40% or more, showing a healthy balance between growing and making money.
The SaaS magic number measures how much annualized new recurring revenue each dollar of sales and marketing spend generates: (net new ARR in the quarter × 4) ÷ prior-quarter S&M spend. It is sensitive to timing, so quarterly revenue, ARR and spend must sit on one stated convention.
The SaaS quick ratio is (new MRR + expansion MRR) ÷ (churned MRR + contraction MRR) over a period. It measures growth efficiency: how much recurring revenue a subscription business adds for every dollar it loses. A ratio of 4 or more is commonly cited as efficient.
A sales qualified lead (SQL) is a prospect that a salesperson has reviewed against agreed criteria and accepted as ready for a direct sales conversation, because the prospect fits your ideal customer profile and has shown clear intent to buy.
Sampling bias is a systematic error that occurs when the method of collecting a sample makes some members of the population more likely to be included than others, so the sample does not represent the population and estimates are skewed regardless of sample size.
Scroll depth is how far down a page a visitor got, measured as the furthest point reached and usually shown as a percentage of page height. It shows whether content was seen, not whether it was read, and percentages cannot be compared between pages of different length.
A scroll map is a report that shows how far down a page visitors scroll: each band is the share of visits that reached that height, so colors cool as fewer people go deeper, and a sharp drop marks where content stops being seen.
An SDK (software development kit) is a package of libraries, documentation, sample code and tools that lets developers add a platform’s or service’s capability to their own software, often by wrapping that service’s API.
Search analytics is the analysis of data generated by searches, and it covers two separate things: people searching on a search engine and seeing your page (impressions, clicks, click-through rate, position), and visitors typing into your own site search (the terms they use).
Seasonality is a regular, predictable pattern in time-series data that repeats over a fixed calendar period, such as the hour of the day, the day of the week or the month of the year.
Second-party data is another company's first-party data that you receive directly through a partnership or data-sharing agreement, such as a retailer sharing its loyalty audience with a brand. Its provenance and consent terms are known — that agreement separates it from third-party data.
A saved group of users or visits defined by properties and behaviour — paid visitors from Germany who added to cart twice. Segments are reusable across funnels, retention, heatmaps and chat targeting, so one definition answers many questions.
Selection bias is a distortion that happens when the people or events included in an analysis differ systematically from the population you want to draw conclusions about, because of how they were chosen or chose themselves.
In web analytics, a self-referral is a session whose source is recorded as your own domain or subdomain, because a visit was split or lost its original source, so internal navigation appears as referral traffic.
Self-selection bias is a distortion that appears when people put themselves into a group, so the group already differed from those who did not choose it. A gap in retention or conversion then reflects who opted in as well as the effect of the thing you are studying.
Sequential testing is a way of analysing an experiment repeatedly as data arrives, with stopping rules designed so that the overall false positive rate stays at the chosen level (for example 5%), allowing a test to stop early when the evidence is strong.
A group of pageviews and events from one visitor with no gap longer than 30 minutes. Sessions make time-on-site and pages-per-visit possible, but they are a convention; TapCub also reports visitors and pageviews directly so you are not forced through the session lens.
A session ID is an identifier that groups the actions of one visit or login. An authentication session ID is a secret that logs a browser in; an analytics session ID is a label produced by cutting rules such as idle timeout. The same events can yield 2 to 5 sessions depending on those rules.
Session replay reconstructs a single visitor session from recorded page structure changes, mouse movement, clicks, scrolls and inputs, so you can play it back like a video. It is a rebuild of the page, not a screen recording.
In web analytics, a session timeout is the period of inactivity after which the current session ends and further activity starts a new one, commonly 30 minutes. In application security, it is the idle or absolute time after which a user is logged out.
Simpson's paradox is a situation in which a relationship that holds in each of several groups disappears or reverses when the groups are combined, because the groups differ in size and in a third factor that affects the outcome.
A single customer view is one consistent profile and timeline for each person, combining identity, behaviour, transactions, conversations and attributes from every system. It exists only when the same person has one stable ID; joining tables without that ID does not create it.
Smoke testing is a quick, shallow set of checks run on each new build or release to confirm that its most critical functions still work before deeper testing or shipping. It is fast, automatable and go or no-go: if any critical path fails, the build goes back.
A smurf attack is a DDoS attack that sends ICMP echo requests with a forged source address to a network's broadcast address, so every host there replies to the victim and floods it. Modern routers block it by default.
SOC 2 is an independent CPA report on a service organization's controls, measured against the AICPA Trust Services Criteria. A Type II report covers whether those controls operated effectively over a stated period; Type I covers a single date. It is a report, not a certification.
A soft 404 is a URL that returns HTTP 200 OK while showing a “not found” message or almost no content, so search engines cannot rely on the status code. Google also flags thin pages it judges to be errors as soft 404s.
Source is where a visit came from — google.com, newsletter, an ad click; channel is the group it belongs to — organic search, email, paid social. TapCub classifies sources into channels automatically and lets you edit the rules.
Speed Index is a lab performance metric that measures how quickly the visible content of a page is displayed during load, computed as the area above the visual-completeness curve of a recorded page load. Lower values are better.
Statistical power is the probability that a test will find a statistically significant result when a real effect of a specified size exists; it equals 1 minus the probability of a type II error (a false negative).
Stratified sampling is a sampling method that divides a population into non-overlapping groups (strata) based on a shared trait, then draws a random sample from each group, so every group is represented in the sample.
A supply-side platform (SSP) is software publishers use to sell ad space automatically: it offers each impression to many buyers, applies price floors, runs the auction and returns the winning ad, keeping a fee (often a percentage) of the media spend.
Survivorship bias is the error of drawing conclusions from only the cases that passed some selection, such as customers who stayed or campaigns that worked, while the cases that dropped out are not in the data.
SWOT analysis sorts what affects a goal into four boxes: strengths and weaknesses inside the organization, opportunities and threats outside it. It is only as good as its evidence, so every item should carry a metric, a time range and a source.
Synthetic monitoring tests websites, apps and APIs by running scripted requests or browser journeys from controlled locations on a schedule, detecting outages and performance regressions before or without real user traffic.
A t-test is a statistical test that compares a mean, or the difference between two means, with the spread of the data to judge whether the difference is likely to be more than random noise.
A taxonomy is a system for classifying things into named categories, usually arranged in a hierarchy and drawn from a closed, planned list of terms. In analytics it governs event names, traffic sources and content categories.
A tech stack (technology stack) is the combination of programming languages, frameworks, databases, infrastructure and third-party tools that a team uses to build, run and measure a software product.
Technical support helps users solve technical problems with a product, through channels such as chat, email and phone. It is organized in tiers, so simple problems are fixed by generalists and hard ones reach specialists. The goal is to get the user working again, ideally on the first contact.
The technology adoption curve models how a new technology spreads through a population, dividing adopters into innovators (2.5%), early adopters (13.5%), early majority (34%), late majority (34%) and laggards (16%). Cumulative adoption follows an S-shaped curve.
A test statistic is one number computed from sample data that measures how far the data sit from what the null hypothesis predicts. A z of 2 means the observed difference is about two standard errors from zero. It is compared with a critical value or turned into a p-value.
A third-party cookie is a cookie set by, and sent to, a domain other than the site shown in the address bar, typically by an embedded ad, tracking pixel, social button or widget. It lets that other domain recognise the same browser across different sites.
Ticket deflection is when a customer resolves a support need through self-service — an FAQ, knowledge base article or AI answer — before any ticket or conversation is created. The ticket deflection rate measures that share of intents, and a repeat-contact check keeps the number honest.
An escalated ticket is a support conversation or ticket handed to a more senior or specialist agent because the current owner lacks the knowledge, authority, or time to resolve it. Escalation is not a failure; the failure is making the customer explain the problem again after the handoff.
How long a visitor stayed on a page before moving on. Classic tools could not measure it for the last page of a visit; TapCub's script reports the time when the visitor switches page, hides the tab or closes it, so the number includes single-page visits and the last page.
Time to first byte (TTFB) is the time from the start of navigation until the first byte of the page’s response arrives. It includes redirects, DNS lookup, connection and TLS setup, and server processing. Good is 800 milliseconds or less.
Time to value (TTV) is the time between a customer starting with a product, usually sign-up or purchase, and the first moment they get the benefit they came for, such as a first report, first sale or first completed task.
Top of funnel (TOFU) is the first, awareness stage of the marketing funnel, where people who have a problem or interest first discover your brand, usually through search, social, content or referrals, before they are ready to evaluate or buy.
Total blocking time (TBT) is a lab metric that sums, for every main-thread task longer than 50 ms between first contentful paint and Time to Interactive, the time beyond 50 ms. It indicates how unresponsive a page is during load.
A tracking pixel, or web beacon, is a tiny, usually invisible image or script embedded in a web page or email. When it loads, the browser or mail client requests it from a server, which records that the page or email was viewed, plus details such as time, IP address and user agent.
The agreed list of events and properties, their names, when they fire and who owns them. It is what stops "signup", "sign_up" and "SignUp" from becoming three events. TapCub keeps the plan in the console and flags events that arrive outside it.
Traffic share is the percentage of total traffic that one channel, page or website receives: a channel's sessions ÷ all sessions on your site, or your visits ÷ all visits in a market. A share only means something next to the numbers behind it.
A traffic spike is a sudden rise in visits, events, errors or conversions well above the normal level for that day and hour. It is defined against a baseline, and it does not say whether the rise is good: it can be real attention, bots or a tracking error.
A trouble ticket is a record that logs one reported problem, gives it a number and an owner, and tracks it from report to resolution. Its value is accountability: a problem stops being something anyone can mention and becomes something one named person must answer for by a set time.
Troubleshooting is a systematic search for the cause of a problem, followed by fixing it and confirming the fix. It works backwards from a symptom: define it, reproduce it, narrow down where it is, test a cause, fix and verify, then record what was found.
The visitor writes in their language, the agent reads and replies in theirs. Agents see the original next to the translation; visitors see the reply in their own language. Translation only triggers when the two languages differ and can be turned off per site.
A Type 1 error is a false positive: a statistical test rejects a null hypothesis that is actually true, reporting an effect that is not there. Its probability equals the significance level α, chosen before the test, commonly 0.05.
A Type II error is a false negative: the test does not reject the null hypothesis although it is false, so a real difference goes unseen. Its probability is β, and power is 1 − β. Small samples, small effects and a strict significance level all raise it.
UA stands for three things: Universal Analytics (Google's former analytics, which stopped processing data on 1 July 2023 or 2024), user acquisition (getting new users, measured by cost per install or per signup) and user agent (the string a browser or bot sends to identify itself).
UAT (user acceptance testing) is the final test before release, in which the people who will use or own the software check that it does the job the business needs, using scenarios that state who does what and what result they should see, including whether tracking events fire as planned.
UGC stands for user-generated content: content such as reviews, photos, videos, questions and posts that ordinary users create and share, rather than the brand or platform operator. It is persuasive because it comes from real users, and it needs moderation, permissions and disclosure.
A unified data model is a shared schema in which data from every platform — web, app, mini-program, server — uses one definition of a person and one definition of each event, so counts, funnels and retention can be compared in a single view.
The number of distinct people in a period. Every tool estimates it differently — cookies, logins, fingerprints or, in TapCub, a salted server-side hash — which is why two tools never agree. In TapCub's default mode the same browser on the same network is recognised across days; identified users are counted per account.
A unique visitor is one identifier, such as a browser ID, device ID or login, that visited a site at least once in a period and is counted once however often it returns. It is not always one person: one person on a phone and a laptop is usually two.
Unit economics measure the revenue and variable costs associated with a single unit of a business, such as one customer or one order, to show whether each unit is profitable before fixed costs. Common measures are CAC, LTV, LTV:CAC, payback period and contribution margin.
URL parameters (query parameters) are key-value pairs added to the end of a URL after a question mark, separated by ampersands, such as ?color=red&utm_source=newsletter. They pass information to the page or to tracking tools without changing the path.
User acquisition is the process of attracting new users to a product, such as app installs, sign-ups or accounts, through paid, organic, referral, partner and product-led channels, measured by cost, volume and the quality of the users acquired.
A user agent is the software, such as a browser, app or crawler, that makes a request on the web. It identifies itself with the User-Agent header, a string naming the browser, rendering engine, operating system and sometimes the device.
User engagement is how much people take part in a product in ways that matter. Measure it as breadth (share of active users who did a key action), depth (key events per active user) and, for daily products, frequency. One number such as time spent hides what changed.
A user error is a mistake made by the person using a system, such as a wrong value or button. When many people make the same error at the same place, it signals a design problem, which you see in error rates and retries.
User feedback is what people who use a product or site tell you about it: what they like, what confuses them, what is broken and what they want. It can be active or passive, in words or numbers, and should be checked against behavior data before you act on it.
A user ID is a stable, unique identifier your product assigns to a signed-in person and sends to analytics, so their activity on every device and in every session counts as one user. It is not the username they type to sign in.
A user interface (UI) is the layer where a person operates a product: the layout, controls, wording and feedback they see and use. It is not the whole experience (UX) and cannot be judged by looks. Judge it by behavior: click rate among people who saw it, follow-through, and repeated rapid clicks.
A user profile is the record of what you know about one person who uses a product or site: attributes (who), behavior (what they did) and derived labels such as segments. It only works if the same person is recognised across visits and devices, usually through a stable login id.
User properties are attributes that describe the person behind the events, such as plan, country or signup date. They are stored once on the user profile and carry across all of that person's events, unlike event properties, which describe a single action as it happened.
Five tags appended to a link — utm_source, utm_medium, utm_campaign, utm_term, utm_content — that tell analytics where a click came from. They are the manual, platform-neutral complement to click IDs. Build them consistently with the UTM tool.
A UUID (universally unique identifier) is a 128-bit value, usually written as 36 characters like 3f2b8c1e-7d4a-4f6b-9a21-5c0e8d7b1f42, that can be generated independently on any machine with a negligible chance of two ever being the same.
Validation is confirming, with evidence, that something meets its intended use. Verification asks whether it was built to spec; validation asks whether it meets a real need. For a feature, validation needs behavior: how many people tried it, used it again, and came back more than before.
A value metric is the unit a SaaS or usage-based product charges for (seats, contacts, events, projects, GB), chosen so that the price grows as the customer gets more value.
View-through rate (VTR) is the share of video ad impressions that end in a view: views ÷ impressions. Google counts a view at 30 seconds, the full ad if shorter, or a click. The name is also used for a different metric, view-through conversions, so check which one a report means.
The viral coefficient, or K-factor, is the average number of new users that each existing user brings in through invites or sharing. It is the number of invites sent per user multiplied by the share of invites that convert.
A walkthrough is a guided pass through a process one step at a time. In product work it means either a guided walkthrough that teaches a user a task, or an evaluation walkthrough where a team steps through a flow to find where users would fail.
Web analytics is the collection, measurement and analysis of data about how people find and use a website (visits, sources, pages, clicks and conversions) in order to understand and improve the site.
A web beacon is a small, invisible request that a web page or email makes to a server so the server can record that the content was loaded. It is usually a one-pixel image; the browser's Beacon API (navigator.sendBeacon) does the same job for data sent as a page closes.
Google's page-experience metrics: Largest Contentful Paint for loading, Interaction to Next Paint for responsiveness, Cumulative Layout Shift for visual stability. TapCub collects them from real visits and shows them per page next to bounce and conversion, so speed becomes a business number.
A website conversion is any action on a website defined as valuable, such as a purchase, sign-up, demo request or form submission; the conversion rate is conversions divided by sessions or visitors.
A call to action (CTA) is a prompt, usually a button or link, that tells a visitor what to do next, such as start a trial, buy, book a demo or subscribe.
CX (customer experience) is the customer's overall perception of a company, formed by every interaction across the journey: discovering it, buying, using the product, getting help and renewing or leaving.
EMEA stands for Europe, the Middle East and Africa. It is a business grouping used for sales territories and reports, not a country or an official region, so the country list behind any EMEA number has to be written down.
Opt-in means a person actively agrees before something starts, such as tracking or marketing email. Silence, a pre-ticked box or inaction is not opt-in. Opt-out is the reverse: it starts by default and the person has to act to stop it.
Winsorization is a way of limiting the influence of outliers by replacing values beyond a chosen percentile (for example above the 99th) with the value at that percentile, so every data point is kept but extreme values no longer dominate averages.
X-Robots-Tag is an HTTP response header that gives search engine crawlers indexing instructions, such as noindex or nofollow, for the URL being served; unlike the robots meta tag, it works for any file type, including PDFs and images.
Year over year (YoY) is a comparison of a metric in one period with the same period one year earlier, usually expressed as a percentage change, which removes most of the effect of seasonal patterns.
Zero-party data is data a customer intentionally and proactively shares with a brand, such as preferences, purchase intentions, goals and how they want to be contacted, as opposed to behavior a company observes (first-party) or buys (third-party).
A zombie page is a live URL that has had almost no search impressions and no visits for a long period, such as 90 days. Work out whether it is not indexed, not linked, not ranking or not clicked, then improve, merge, remove or keep it. Do not delete them all.