Quick answer
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).
What analytics means
Analytics is the systematic, computer-assisted analysis of data to find patterns and support decisions. The IEEE describes it as a field that uses structured, computational examination of data to uncover patterns, reach conclusions and inform decisions, drawing on statistics, operations research and computer science (IEEE Technology Navigator). Wikipedia's wording is similar: the systematic computational analysis of data or statistics, used to discover, interpret and communicate meaningful patterns (Wikipedia).
Two words get confused with it. Analysis is the act of examining something; analytics is the discipline and the tooling around doing it with data, repeatedly. And data is the raw material: a table of events is data, and "signups rose 8% because of email" is the output of analytics.
The working test: analytics is only analytics if it ends in something a person can decide or do. A chart nobody looks at is a record.
The four types of analytics
The common way to organise analytics is by the question it answers. The IEEE page describes four types:
| Type | Question | What it does | Typical methods |
|---|---|---|---|
| Descriptive | What happened? | Summarises historical data | Aggregation, charts, summary statistics |
| Diagnostic | Why did it happen? | Identifies factors linked to an outcome | Breakdowns, correlation, root-cause decomposition |
| Predictive | What will happen? | Forecasts from current and past data | Regression, decision trees, machine learning |
| Prescriptive | What should we do? | Recommends actions to reach a goal | Optimisation, simulation |
The types build on each other. You cannot explain a number you have not measured, and you cannot optimise a decision you cannot predict. Most teams spend most of their time in the first two, and that is fine, because they pay off soonest.
A worked example: one dataset, four questions
Sample data: monthly signups for a product, January to June: 3,100, 3,300, 3,500, 3,600, 3,900 and 4,200.
Descriptive. June had 4,200 signups against 3,900 in May: 4,200 ÷ 3,900 − 1 = +7.7%.
One refinement: May has 31 days and June has 30. Per day, June is 4,200 ÷ 30 = 140 and May is 3,900 ÷ 31 = 125.8, so the daily rate rose by 11.3%, a bigger change than the monthly total shows. Compare periods of equal length, or per day, before drawing conclusions.
Diagnostic. The +300 split by channel is:
| Channel | May | June | Change |
|---|---|---|---|
| Organic search | 1,500 | 1,550 | +50 |
| Paid | 1,200 | 1,240 | +40 |
| 1,200 | 1,410 | +210 | |
| Total | 3,900 | 4,200 | +300 |
Check: 50 + 40 + 210 = 300, and email's share of the change is 210 ÷ 300 = 70%. The rise is mostly email. A breakdown explains the change only if you then ask what happened inside the biggest bucket: if the 210 came from one campaign sent on a single day, the diagnosis is a campaign, not a trend.
Predictive. A straight line fitted to the six months (least squares) has a slope of about 211 signups per month and gives about 4,340 for July. It is a rough projection: it assumes the recent trend continues, which a one-off email campaign may not.
Prescriptive. Suppose you have a $10,000 budget, email costs $4 per signup but can deliver at most 2,000 signups, and paid costs $12 per signup. Put the budget into the cheaper source first: 2,000 × $4 = $8,000 for 2,000 signups, then $2,000 ÷ $12 ≈ 166 more from paid, about 2,166 in total. A 50/50 split gives $5,000 ÷ $4 = 1,250 plus $5,000 ÷ $12 ≈ 417, about 1,667. The recommendation is not obvious from the descriptive chart, and it depends on the cost and capacity numbers being right.
The decision rule: stop at the type that is enough for the decision. A weekly review usually needs descriptive plus diagnostic; budget allocation needs more.
Which type for which decision
Matching the type to the decision avoids both under- and over-engineering.
| Decision | Question behind it | Type that is enough | What you need first |
|---|---|---|---|
| Weekly update to the team | What happened this week? | Descriptive | A trusted count and a comparison period |
| Explaining a spike in signups | Why did it change? | Diagnostic | The same metric broken down by channel, device or page |
| Planning support staffing | How many conversations next month? | Predictive | Months of stable history |
| Splitting a fixed budget | Where should the money go? | Prescriptive | Cost and capacity per option |
Three checks before any analysis starts: the metric has one written definition (what counts as a signup); the data is complete enough (bots and test traffic removed, consent coverage known); and you have named the comparison (last month, last year, a control group). Most disagreements about analytics are really disagreements about one of these three.
Analytics versus related terms
| Term | Focus | Difference from analytics |
|---|---|---|
| Web analytics | Visits and behaviour on a website | Analytics applied to one kind of data |
| Digital analytics | Web, app and other digital channels together | Wider set of sources |
| Business intelligence | Reporting on the state of the business | Mostly descriptive, from curated data |
| Data science | Building models and algorithms | Heavier on predictive methods |
| Statistics | The mathematics of inference | The theory analytics borrows |
For a longer comparison of methods, see the blog on types of data analytics.
Common misconceptions
- "Analytics is a tool." Tools collect and display. Analytics is the process of getting from a question to a conclusion.
- "More dashboards means more analytics." Reports are only useful if they change a decision.
- "Predictive is better than descriptive." A forecast built on an unmeasured baseline is a guess with decimals.
- "Correlation is the answer." Diagnostic work finds linked factors; proving cause usually needs a controlled comparison.
- "One number is the answer." A rise of 7.7% means little without the comparison period, the definition and the size of the usual month-to-month swing.
- "The data speaks for itself." Someone chose what to collect, how to define a visitor and what to compare.
What analytics looks like in TapCub
Take the signup example and see where each type lives in TapCub.
- Descriptive. Web analytics › Overview and Web analytics › Sources show visitors, sessions and conversions over time and by channel. Goals count signups; the channel split above comes from the source report.
- Diagnostic. In All-platform analytics › Operations › AI Assistant you can ask why a metric changed. Its change attribution breaks the change down by channel, referring domain, country, device and landing page and shows how much each part contributed. The numbers come from analysis queries run on your project, not from a lookup.
- Checking the baseline. All-platform analytics › Operations › Monitors compare a metric with a baseline band built from the last 7 days and the same weekday, and notify you when it is outside. That is alerting against a recent norm, not a forecast model.
- Predictive and prescriptive. TapCub does not provide forecasting or optimisation. It supplies the measured history those methods need, which you can export.
TapCub reports what happened and helps explain why; the decision about what to do stays with your team. Bots and AI crawlers are counted apart from people, so the baselines are not inflated.
Sources
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Data Analyst
Explores traffic sources, conversion funnels, and retention, using clear metric definitions and concrete examples to explain what the numbers mean.