TapCub is live — analytics, insights and live chat, free on one platform

Blog Analytics, growth and chat, explained with real numbers · Page 10

Practical notes from the team building TapCub: how metrics are defined, where funnels break, what AI crawlers take from your site, and how support teams turn conversations into conversions.

All articles

Browse every article

Filter by category or search a keyword. Every article links to the report or setting it talks about.

723 articles · page 10 of 16

Sample data: a 1,248,300-row dataset after its first trial. 1,248,300 − 1,102 − 286 = 1,246,912; the amount and customer differences equal what the rejected rows carry.
AnalyticsNov 17, 2026 · 16 min read

Data migration project: a five-phase plan with reconciliation checks

A data migration project succeeds on the inventory before it and the reconciliation after it. Five phases with gates, a worked reconciliation, rollback conditions, an eight-week timeline and how to run it for analytics.

Read more
Sample data: 30 healthy, 12 watch and 8 at-risk accounts lost 0, 1 and 2 accounts in the next quarter. The at-risk band is 16% of accounts but 2 of the 3 losses.
Growth & MarketingNov 17, 2026 · 13 min read

Customer Success KPIs: Which Ones Warn You Before a Customer Leaves

Customer success KPIs split into outcome numbers (churn, GRR, NRR) and leading numbers (active members, time to first value, health score). A renewal-quarter example, a backtested health score and a six-number dashboard.

Read more
Sample data: 45 orders arrive per hour but the line delivers only 30, the capacity of picking. Raising picking to 40 adds 33.3%; raising packing to 60 adds nothing.
AnalyticsNov 17, 2026 · 14 min read

What Is a Bottleneck in Business? How to Find the Slowest Step With Data

A bottleneck is the one step that caps the output of a whole process. Four types, a worked five-step example, how to find it with conversion and wait-time data, and what fixing it is worth.

Read more
Sample data: health = 40 × share of members active + 30 × share of active days + recency (30 within 7 days, 15 within 30). Birch scores 95 with a single member, Acme 62 with three active people.
Growth & MarketingNov 17, 2026 · 14 min read

B2B Customer Journey: How to Measure Stages at the Account Level

A B2B customer journey belongs to the account, not one person. Five stages with an observable signal each, a four-account health score example, rules to roll members up into a stage, and the mistakes that mislead.

Read more
Sample data: weekdays near 1,100 clicks, weekends near 600, one real drop to 640 on day 11. A rule that compares with the 7-day mean also fires on two weekend days; comparing with the same weekday fires once.
AnalyticsNov 17, 2026 · 13 min read

Automated SEO monitoring: four signals to watch and the alert rules to set

Automated SEO monitoring only helps if you know what to watch and what change should wake someone up. Four signal types with starting thresholds, why weekday and rank noise cause false alarms, and how to test a rule first.

Read more
Sample data: 2 hours × 52 weeks is 104 hours by hand. Automated, year one is 6 setup + 13 weekly check + 4 upkeep = 23 hours, saving 81 hours (about $3,240 at $40 an hour), with setup repaid in about 3.4 weeks.
AnalyticsNov 17, 2026 · 12 min read

Automated Reports: Fix the Definitions First, Then Schedule Them

Automated reports build themselves and arrive on a schedule, which also repeats any wrong definition faster. Four steps, a one-page definitions table, a frequency guide and a recomputed time-saved example.

Read more
Sample data: conversion rose from 28% to 33%, so downloads grew 17.9% and day-1 retained users 11.7%, but day-7 retention fell from 14% to 11% and day-7 retained users dropped 7.4%. The new page brings in people who leave within the week.
Growth & MarketingNov 17, 2026 · 15 min read

ASO Guidelines: Platform Rules, Tests and the Retention Check

ASO guidelines cover what Apple and Google allow in a store listing and how to tell whether a change helped. A before-and-after example shows downloads rising 17.9% while day-7 retained users fall 7.4%.

Read more
Sample data: sessions fell from 10,000 to 8,200 (−18.0%). Paid search explains half of the drop, and within it one prospecting group lost 850 sessions (47.2% of the total). The split narrows the search; it does not say why the group fell.
AnalyticsNov 17, 2026 · 12 min read

AI Marketing Analytics: What It Can Do, and How to Check Its Answers

AI marketing analytics is safest as a way to ask your data, not to decide for you. Four uses and how far to trust each, how to phrase a question, an 18% traffic drop split by channel, and six checks before you act.

Read more
Sample data: the dashboard shows 45% handled without a handoff, but 35% were truly resolved and about 33% resolved correctly. Count the strict figure when you decide.
Live ChatNov 17, 2026 · 15 min read

AI Help Desk: What AI Should Answer and How to Hand Off

An AI help desk answers stable, written-down questions first and hands everything else to people with the context. A worked month shows 45% handled shrinking to 35% truly resolved, plus a quality-check method.

Read more
Sample data: 1,500 tasks a month at 6 minutes and $30 an hour, $300 subscription, 3% of completed outputs wrong. Break-even completion is 31.8% when an error costs $40 and 21.5% when it costs $5.
Live ChatNov 17, 2026 · 14 min read

AI Agents for Business: Where to Start and How to Control Risk

AI agents for business pay off on tasks with a measurable success criterion and a low cost of failure. Three tests for choosing a task, an eight-point risk checklist, a worked break-even and a trial method.

Read more
Sample data: one store month, 12,000 sessions including 3,000 bots, 270 orders. Recomputed on 9,000 human sessions: conversion rate 270/9,000 = 3.0% (+33% vs 2.25%), bounce rate 40% (−27% vs 55%), pages/session 3.0 (+20% vs 2.5), average engagement 90s (+33% vs 67.5s).
AnalyticsNov 16, 2026 · 13 min read

Website KPIs: how to choose the few numbers that matter

Pick website KPIs backward from the job your site must do: one primary KPI, three to five explainers, clean denominators, and targets you can recompute. With KPI trees for four site types and two worked examples.

Read more
Sample data: one site, 30 days, server records. TapCub counts SemrushBot, SiteAuditBot and SplitSignalBot in one SEO row (4,800 + 2,300 + 420). After Disallow, OCOB fell 1,900 → 60 (−96.8%); the SEO row stayed at 7,520.
AI SearchNov 16, 2026 · 10 min read

SemrushBot: The 10-Agent Family and How to Control Each One

SemrushBot is a family of ten Semrush crawlers with independent robots.txt tokens. What each does, how to keep SEO data while opting out of SemrushBot-OCOB, how to verify requests, and how to decide per goal.

Read more
Sample data: PerplexityBot read 8,000 pages and sent back 120 visitors (1.5%, ≈67 reads per visitor); the training crawler read 6,000 pages and sent back ≈0 (0%). In TapCub the ledger shows the same row as about 67 reads : 1 visitor, and AI visibility shows Crawl→visit 1.5%.
AI SearchNov 16, 2026 · 10 min read

PerplexityBot: What It Reads, What It Sends Back, and Whether to Allow It

PerplexityBot reads pages so Perplexity answers can cite your site; Perplexity-User fetches on user request. What each does, robots.txt rules, IP verification, and a reads-vs-visitors rule for allowing or blocking.

Read more
Sample data: step conversion 80% / 70% / 60% / 75%; activation 252/1,000 = 25.2%. The pre-value invite step loses 224 users (−40%), the biggest drop; deferring it projects 560 × 75% = 420 activations (42.0%).
Product NewsNov 16, 2026 · 17 min read

Onboarding flow: formats, design steps, and a worked funnel example

An onboarding flow ends when the user first gets value, not when they finish the tour. Compare four formats, design backward from activation, cut the biggest pre-value drop with a recomputable funnel example, and verify the change with cohorts.

Read more
Sample data: of 1,000 web-first people, 120 ordered in the app within 30 days. By first-touch source that is 60/400 = 15.0% (paid), 42/350 = 12.0% (organic), 18/250 = 7.2% (social). Last-click would credit all 120 to the app and hide social's 18 assists.
AnalyticsNov 16, 2026 · 14 min read

Omnichannel analytics: connect the same person before comparing channels

Omnichannel analytics unifies web, app and mini-program data into one person-level journey. Why per-channel reports double-count people, how to stitch identity and events, a recomputable cross-channel example with four attribution models, and the limits.

Read more
Sample data: mobile converts at half the desktop rate (120/6,000 = 2.0% vs 160/4,000 = 4.0%). After enlarging the CTA and compressing the hero image, mobile reaches 210/6,000 = 3.5% — a 75% relative lift, gap 2.0pp to 0.5pp. Observational before/after, not an A/B test.
Product NewsNov 16, 2026 · 17 min read

Mobile UX: Find the Problems in Your Data Before You Redesign

Mobile UX is the whole phone experience: loading, touch, navigation and error recovery. Find the five common problems in device-split data first — rage clicks, scroll cliffs, slow LCP — then fix them in priority order, with a recomputable mobile-vs-desktop example.

Read more
Sample data: six of 48 scripts account for 780 of 1,200 uses (65.0%) in 30 days, and 16 scripts were never used. The top six deserve the most careful wording; the unused ones can be retired.
Live ChatNov 16, 2026 · 19 min read

Live chat scripts: 21 conversation templates that don't sound scripted

Live chat scripts are multi-turn conversation templates that lock the structure and must-say information while leaving wording to the agent. Here are 21 templates across 7 scenarios, when to go off-script, and a 30-day audit that shows which scripts to keep, rewrite or retire.

Read more
Sample data: in the sample month, 24 of 200 visitors who chatted reached the goal (12%) vs 120 of 4,000 who didn't (3%) — a relative lift of +300% (12 ÷ 3 − 1). Observational, not causal; weeks with fewer than 30 chatters are flagged as too small.
Live ChatNov 16, 2026 · 16 min read

Customer Service Chat: A Team Playbook for Staffing, Routing and Metrics

Run customer service chat as a real service channel: what belongs in chat versus email and phone, a staffing worked example, routing and handoff rules, and four metrics with a recomputable sample month.

Read more
Sample data: each magnet was shown to its own 1,000 landing visitors. The pricing-page calculator drew one third the ebook's downloads (40 vs 120) yet closed four times as many deals (4 vs 1). Judge magnets by qualified leads and closes, not downloads.
Growth & MarketingNov 16, 2026 · 17 min read

Lead Magnet Examples: 15 Ideas Matched to Visitor Intent

Fifteen lead magnet examples grouped by what the visitor is doing on the page — learning, comparing or deciding — each with format, effort, fit, gate and the metric to watch. Plus a recomputable example showing why qualified leads beat download counts.

Read more
Sample data: 20 queries on one content site, 30 days. Weighted CTR falls from 4.57% at positions 1–3 to 0.37% at 21+; the flagged page sits in the 4–10 bucket at 1.2% against a 2.0% median.
AnalyticsNov 16, 2026 · 15 min read

Keyword Performance Report: Template, Worked Example, and Monthly Cadence

A keywords performance report should answer one question: which page do I change next? Here is a nine-column template, a 20-query example with impression-weighted CTR by rank bucket, and a monthly cadence — every number recomputable.

Read more
Sample data: moving security badges + returns + a chat invite beside the payment button lifted checkout completion from 25.0% to 30.0% and scroll-to-trust from 32% to 71%. Observational before/after, not an A/B test.
Growth & MarketingUpdated NovNov 16, 2026 · 15 min read

Ecommerce Trust Signals: Put Them Where Shoppers Hesitate

Ecommerce trust signals only work at the hesitation point — price, shipping, checkout, returns. Which signals matter most, where each goes by buying stage, how to find hesitation points with data, and one recomputable fix.

Read more
Sample data: Acme scores 86 (healthy); Beacon 47 and Cedar 24 fall below the 60 threshold and enter the trigger-based play queue. Cedar, with one active member out of five, is the most fragile.
Growth & MarketingNov 16, 2026 · 14 min read

Customer success strategy for SaaS: stages, health scores, tiers and plays

Find problems before customers do: four lifecycle stages with early-warning signals, a health score you can recompute, ARR-by-health tiering with capacity math, trigger-action-owner playbooks, and the GRR/NRR cohort math that proves it works.

Read more
Sample data: 200 ratings across 400 conversations. Bars are the mean on a 1–5 scale (5.0 = 100). The overall 780/200 = 3.9 hides billing at 192/60 = 3.2 versus 588/140 = 4.2 everywhere else.
Live ChatNov 16, 2026 · 19 min read

Customer service questions: what to ask after a chat, and what to ask during one

Customer service questions work when there are few of them and each maps to a decision you would change. A keep/drop test for post-service surveys, diagnostic sequences for agents, and a month of sample data read two ways.

Read more
Sample data: human handling time for 1,000 intents. Routing, canned replies and KB self-service alone cut 100 h to 45 h (−55%) before any AI answer; adding AI answers and correcting for 21 repeat contacts brings it to 30.8 h (−69.2%).
Live ChatNov 16, 2026 · 15 min read

How to Automate Customer Service: Triage First, Answers Second

Automate customer service in the right order: triage and preparation first (routing, canned replies, offline fallback, knowledge base), then AI answers. A nine-part map, stage gates, three reply templates, hard handoff rules, and a sample month you can recompute.

Read more
Sample data: 30-day return rate of the activated-once dormant segment — 24/240 = 10.0% in January (no win-back) vs 60/250 = 24.0% in February (win-back sent), +14.0 pp and 2.4x. Observational before/after, not an experiment.
Growth & MarketingNov 16, 2026 · 19 min read

Customer Retention Examples: 12 Tactics Grouped by Why People Leave

Twelve customer retention examples grouped by why people leave — never reached value, friction, forgot, pulled away, payment — each with a tactic, when to use it and a metric, plus a recomputable win-back before/after.

Read more
Sample data: case A (stable answers, low error cost) saved 1,200 × 3 min = 60 agent-hours; case B (unstable answers, high error cost) took 952 min versus 800 min human-only — 2.5 hours more. Same store, opposite results.
Live ChatNov 16, 2026 · 17 min read

Chatbot use cases: 12 examples and a two-axis test for what to automate

12 chatbot use cases sorted by answer stability and error cost: what to automate first, where a human must confirm, what stays human-only, and a recomputable one-month example.

Read more
Sample data: 50 chats × 5 items = 250 points. Before 22+18+15+20+30 = 105 (42%); after 40+38+36+41+44 = 199 (79.6%). Bars show each item's pass rate out of 50 chats, before and after the rewrite rules.
Live ChatNov 16, 2026 · 16 min read

Chat Etiquette: Rules for Every Moment of a Live Chat

Chat etiquette is fewer round-trips to the answer, not a politeness list. Rules and good/bad examples for opening, waiting, transferring and closing, a full 11-to-6 rewrite, and a 50-chat audit you can recompute.

Read more
Sample data: 30-day window with server records. After the day-15 rule, CCBot only fetched robots.txt (~2 checks/day × 15 days = 30). TapCub runs no DNS check for CCBot, so trust shows Unverified; verify with reverse DNS or ccbot.json.
AI SearchNov 16, 2026 · 9 min read

CCBot: Common Crawl's Dataset Crawler and What Blocking It Changes

CCBot builds Common Crawl's free public dataset — a documented training source for GPT-3, LLaMA and BLOOM. What it fetches, how to verify it, robots.txt rules, and why blocking only shapes future snapshots.

Read more
Sample data: Linear credit for the Acme demo request — each of 9 sessions gets 1/9 = 11.11%, so the three two-session channels (organic search, organic social, email) get 22.22% each. Under last click, Direct takes 100% and the other five channels get 0; four of them (organic social, paid search, email, referral) never appear as first or last touch.
Growth & MarketingNov 16, 2026 · 17 min read

B2B Marketing Attribution: Count Credit at the Account Level

B2B marketing attribution has to be counted at the account level, not the person level. A reproducible example — one account, four people, nine sessions across 88 days — scored under five models, plus lookback-window truncation, model choice and the analytics–CRM split.

Read more
Sample data: server records. Applebot verified by reverse DNS to applebot.apple.com; Extended never crawls; the wrong rule leaves only robots.txt fetches.
AI SearchNov 16, 2026 · 10 min read

Applebot and Applebot-Extended: Two Agents and How to Control Each One

Applebot crawls for Spotlight, Siri and Safari search; Applebot-Extended is only a robots.txt switch for AI training. What each does, three rule choices, DNS and IP verification, and a worked example of what the wrong block costs.

Read more
Sample data: demo project with server records, last 30 days. After the rule, Amazonbot only fetched robots.txt. TapCub runs no DNS check for Amazon, so trust shows Unverified; check source IPs against the Amazon IP file.
AI SearchNov 16, 2026 · 10 min read

Amazonbot: Amazon's Three Crawlers and How to Control Each One

Amazonbot is Amazon's crawler whose reads may be used to train Amazon AI models — one of three Amazon agents with independent robots.txt settings. What each does, rules per goal, IP verification, and a worked example of reads versus traffic sent back.

Read more
Sample data: TapCub counts both Ahrefs user agents in one AhrefsBot row (38,400 + 3,400 requests). Trust shows Unverified because TapCub runs no DNS check for Ahrefs; verify with reverse DNS or the Ahrefs IP list.
AI SearchNov 16, 2026 · 9 min read

AhrefsBot: What It Does and Whether to Allow or Block It

AhrefsBot reads your site for Ahrefs users — including competitors — and sends back no visitors. What its two crawlers do, robots.txt allow/block/rate-limit rules, verification, and how to decide.

Read more
Sample data: 200 recent questions at a clothing store. The dark bars (140 questions, 70%) can be answered from policy or product text; the light bars (60, 30%) need an order record or a decision and go to a person.
Live ChatNov 15, 2026 · 15 min read

AI Chatbot for Ecommerce: What to Automate and What to Hand Off

How to launch an AI chatbot for ecommerce that answers sizing, shipping and returns accurately and hands order and refund questions to people: a 200-question sizing example, five launch steps, legal risks and metrics.

Read more
Sample data: the only call to action sat in a section 38% of visits reached, so 2.3% of visits clicked it. Adding the button to the hero lifted clicks from 182 to 384 (+111%).
Growth & MarketingNov 15, 2026 · 14 min read

Website layout: order the page by visitor task, then check what people see

Website layout types and when to use each, what eye-tracking research actually shows, what belongs above the fold, and how to use scroll reach and click maps to check a layout before and after you change it.

Read more
Sample data: orders and human visitors never change, yet counting 2,500 bot visits as visitors cuts the reported rate from 3.00% to 2.40%, a 20% understatement.
AnalyticsNov 15, 2026 · 16 min read

Types of bias in data analysis: the four stages and what to check first

Bias in data analysis enters at one of four steps: sampling, measurement, analysis or interpretation. A table of each type with examples, a worked bot-traffic calculation and a checklist for your own reports.

Read more
Sample data: revenue fell 11.8% because conversion dropped to 2.10%; traffic and order value were normal, so the diagnosis takes one line.
AnalyticsNov 15, 2026 · 13 min read

Small business dashboard: 5–7 numbers that tell you when to act

A small business dashboard needs five to seven numbers, each answering whether to act this week. How to pick them, examples for three business types, thresholds and actions, and a worked week.

Read more
Sample data: India and Brazil mobile are 18.7% of loads but 43.9% of slow ones, and their TTFB alone is 1.7–2.1 s. The fix is edge caching near them, not more image compression.
AnalyticsNov 15, 2026 · 13 min read

Slow loading websites: find who is slow and where before you fix anything

Why websites load slowly: the server, resource downloads or the browser's main thread. A diagnosis tree, real-user data split by country and device, and the order to fix things in.

Read more
Sample data: lifting verification from 65% to 80% adds 234 active users; deleting the company size field adds 42.
Growth & MarketingNov 15, 2026 · 14 min read

Sign up flow: measure each step before you cut a field

A sign up flow runs from the sign up page to a working account. The trade-off behind each element, a six-event tracking plan, and a worked funnel showing why the step that leaks most is often not the form.

Read more
Sample data: price rises and higher unit volume masked a shift to cheaper products; at last year's prices, this year's sales would be 6.1% lower.
Growth & MarketingNov 15, 2026 · 14 min read

Sales analysis: split revenue into volume, price and mix

Sales analysis explains why revenue changed. Split it into volume, price and mix, or a price increase looks like growth. Types of sales analysis, a worked three-category example, steps and a report structure.

Read more
Sample data: accounts that connected a data source in the first 7 days (A + B, 40 accounts) paid at 37.5%; the other 60 at 3.3%. A form-based MQL flag on the same trials only separated 20% from 15%.
Growth & MarketingNov 15, 2026 · 15 min read

SaaS Lead Generation: Qualify Leads by What They Do in the Product

SaaS lead generation channels compared by lead quality, how to write a behavior-based PQL rule, and a worked example of 100 trials where product behavior predicts payment far better than form fields.

Read more
Sample data: closing the activation gap is worth $4,320 of new MRR a month, more than the other four gaps combined ($1,980). Each gap is closed one at a time, others held constant.
Growth & MarketingNov 15, 2026 · 13 min read

SaaS Growth Strategy: Start From the Stage That Is Most Stuck

Choose a SaaS growth strategy by diagnosis, not by trend: measure each AARRR stage as a rate, convert each gap into MRR, and fix the biggest one first. Worked example, strategies per stage and a quarterly cadence.

Read more
Sample data: 90% of repeat orders arrive within 60 days, so 60 days is this store's dormant window. At 14 days, 40% of returns were still to come.
Growth & MarketingNov 15, 2026 · 13 min read

Re-engagement campaign: define dormant before you write the email

A re-engagement campaign brings back customers who went quiet. How to set the dormant window from your own return data, why the wrong window inflates results 30 times, and how to tier, message and measure with a holdout.

Read more
Sample data: adoption ran ahead of plan while the announcement was fresh, then flattened at 23%. Repeat use beat its target, so the decision rule points to discoverability, not value.
Product NewsNov 15, 2026 · 14 min read

Product launch plan: decide how you will judge success before launch day

A product launch plan covers goals, audience, timeline, roles, rollout and rollback. The part most plans leave vague is how to judge the result. A timeline from six weeks before to four weeks after, a success-criteria card, and a worked review.

Read more
Sample data: on site purchases, only brand clears the 2.5 break-even line; generic reaches 2.52 only when the platform's 168 conversions replace the site's 126. Bars are scaled to brand's 7.60.
Growth & MarketingNov 15, 2026 · 16 min read

PPC Analysis: Reconcile the Numbers, Then Decide Where Budget Goes

A step-by-step PPC analysis: reconcile platform clicks and conversions with your site's sessions and goals, compare campaigns on CPC, CPA and ROAS against break-even, check landing pages and attribution, then write decision rules.

Read more
Sample data: compression alone saves 40%; serving the displayed size saves 75% and WebP takes it to 82.5%. Together with removing lazy loading, LCP falls from 3,000 ms to 1,045 ms.
AnalyticsNov 15, 2026 · 14 min read

How to optimize images for web: fix size and priority before format

The order that matters when you optimize images for the web: right dimensions with srcset, hero image loaded first, then format and compression. A worked example shows why, plus how to check LCP afterwards.

Read more
Sample data: third-party scripts own 720 of 990 ms of blocking time (72.7%). Minifying first-party code would touch only the 25% slice; removing and deferring third parties starts with the 73% one.
AnalyticsNov 15, 2026 · 14 min read

Minimize main thread work: find the long tasks before you fix anything

What Lighthouse's 'minimize main-thread work' audit measures, why it shows up as slow interactions (INP), how to attribute long tasks to the scripts that own them, and which fixes pay off first.

Read more
Sample data: of 12 metrics on a growth team's dashboard, four pass all four questions and become KPIs; organic visitors stays as a supporting metric.
AnalyticsNov 15, 2026 · 12 min read

Difference between metrics and KPIs: four questions that decide

Every KPI is a metric, but only a metric with a target, one owner and a threshold that triggers action is a KPI. Four questions, a worked 12-to-4 shortlist and how to calculate an alert threshold.

Read more
Sample data: paid social's leads cost half as much, but its customers cost 2.5 times more (8,000 ÷ 6 vs 4,800 ÷ 9). Bars are scaled to $1,333.
Growth & MarketingNov 15, 2026 · 14 min read

Lead Generation Metrics: Read Them in Pairs, Not Alone

The lead generation metrics that matter in four groups — volume, quality, cost and speed — why each should be read next to a partner metric, and a two-channel example where optimising cost per lead cuts customers by more than a third.

Read more

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics