Quick answer
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.
What is behavioral data?
Behavioral data is the record of what people did when they interacted with a product, site, store or service: which pages they viewed, what they clicked, what they bought, which features they used, and when. It answers three questions about a person that other data cannot: what they did, in what order, and how often.
It helps to separate three kinds of statements about a customer. Did: "opened the pricing page twice this week" (behavioral). Said: "I plan to buy next quarter" (attitudinal). Is: "works at a 50-person company in Spain" (attribute). The three often disagree, and when they do, the record of what was done is usually the most reliable guide to what happens next. The worked example below shows the size of the gap. The cost: behavior tells you what happened, not why.
What does one behavioral record contain?
Most behavioral data is stored as events: one row per action. A usable event answers who, what, when, where and with what details.
| Field | Question it answers | Example |
|---|---|---|
| Identity | Who did it? | visitor id v_8231, or a login id once known |
| Event name | What did they do? | click_start_trial |
| Timestamp | When? | 2026-03-02 10:04:11 |
| Context | Where and on what? | page /pricing, mobile Safari, referrer a search engine |
| Properties | With what details? | plan = pro, price shown = 29 |
Sample data: five events from one visitor, in order.
| Time | Event | Page |
|---|---|---|
| 10:02 | pageview | /pricing |
| 10:03 | scroll_75 | /pricing |
| 10:04 | click_start_trial | /pricing |
| 10:05 | signup_complete | /signup |
| Next day 09:10 | project_created | /app |
Read in order, the five rows already tell a story that a count of "5 events" does not. This is why behavioral analysis is about sequences, not just totals. A stream limited to page views and clicks is called clickstream data.
Behavioral, attribute, attitudinal and transactional data compared
| Type | What it records | Example | Collected by | Changes | Good for | Weak at |
|---|---|---|---|---|---|---|
| Behavioral | Actions | Searched, viewed, clicked, logged in | Event tracking | Daily | Predicting next actions, finding friction | Explaining why |
| Attribute | Who or what the person or company is | Country, plan, company size | Forms, accounts, enrichment | Slowly | Targeting, grouping | Predicting use |
| Attitudinal | Stated opinion or intent | "Very likely to recommend" | Surveys, interviews | When asked | Understanding motives | Matching later actions |
| Transactional | Money events | Order of 42.00, refund | Order ledger | Per purchase | Revenue, repeat rate | Pre-purchase behavior |
Transactional data is a special case of behavioral data with money attached; it is kept separate because it comes from a different system and has to reconcile with finance. For how analysts use all of these together, see behavioral analytics and event analytics.
Why trust what people did over what they said?
Sample data: a software company surveys 200 free users. "How likely are you to use the new export feature?" 124 say yes (62%), 76 say no. The team then checks usage over the next 30 days.
| Said they would use export | Respondents | Actually used it | Rate |
|---|---|---|---|
| Yes | 124 | 31 | 25.0% |
| No | 76 | 6 | 7.9% |
| All | 200 | 37 | 18.5% |
The recomputation: 31 ÷ 124 = 25.0%, 6 ÷ 76 = 7.9%, 37 ÷ 200 = 18.5%. Two findings. First, the stated intention is not worthless: yes-sayers used the feature at 3.2 times the rate of no-sayers, and a two-proportion z-test gives z ≈ 3.0, so the difference is real. Second, 62% said yes and only 25% of them followed through; 3 in 4 "yes" answers never turned into an action, and 6 of the 37 people who used export had said they would not. If the team had planned capacity from the survey (124 users), it would have been off by a factor of more than 3 compared with the 37 who acted.
The practical rule: use attitudinal data to learn reasons and wording, and behavioral data to size and time anything.
What questions can behavioral data answer?
- What happened? Visits, signups, purchases per day: totals and trends.
- In what order? Paths and funnels: where do people drop between viewing the price and signing up?
- Who comes back? Retention: of people who first did X in week 0, how many did anything in week 4?
- Which group acts differently? Compare users who used a feature with those who did not, by outcome.
- Is something broken? Errors, rage clicks and dead ends show up as behavior before anyone complains.
It cannot answer why (needs interviews, replay or a survey) or what people would do if you changed something (needs a controlled test). Be careful with the second one: two groups that behave differently are not proof that one behavior caused the other.
How behavioral data misleads
- Counting events instead of people. 10,000 clicks can be 10,000 people or 50 people clicking 200 times each. Report people (distinct ids) alongside events.
- Anonymous and identified records mixed. The same person appears as two visitors before and after login unless the two are linked. Counts of "new users" are then inflated.
- Bots counted as people. Automated traffic produces perfect-looking behavior. Remove it before computing rates; see bot filtering.
- Survivorship. Records only exist for people who reached the page. Someone who never arrived leaves no event.
- Over-reading a short window. Two weeks of data can show a pattern that disappears across a month.
- Reading cause into correlation. People who use more features retain better, but heavy users may simply be the ones who were going to stay.
Is behavioral data personal data?
Often yes, depending on how it is stored. GDPR Article 4 defines personal data as any information relating to an identified or identifiable person, and profiling as automated processing of personal data to evaluate aspects of a person, including behavior. A stream of events tied to a login id is personal data in that sense; a daily count of page views with no identifiers is not. This is not legal advice: decide what to collect, why, for how long and who can see it before you start, and treat free-text properties (search terms, form values) with particular care.
Behavioral data in TapCub
Take the sample visitor above and follow them through TapCub:
- Collect. The website script records pageviews and, by default, outbound link clicks and file downloads. Your own events (
click_start_trial,project_created) come from calling the tracker or the server interface, and are listed under Web analytics › Events. - Analyze sequences. In All-platform analytics › Behavior › Behavior analysis, funnels, retention and paths use the event stream; in Behavior › Event analysis you count events and distinct people for any event.
- Attach identity. Call
tapcub.loginwhen someone signs in. They then appear under All-platform analytics › Users › People, with a day-by-day timeline of what they did. Anonymous visitors are counted but have no profile. - Keep bots out. Recognized bots are reported separately and do not inflate human counts.
- Set limits on collection. In Configuration › Privacy center you choose the collection tier, retention and data-subject request handling.
TapCub records and analyzes what people did. It does not run surveys, so the attitudinal side has to come from a tool you already use.
Sources
- arXiv, The structure of behavioral data (2020).
- European Union, Regulation (EU) 2016/679 (GDPR), Article 4: Definitions.
- NIST/SEMATECH e-Handbook of Statistical Methods, 7.3.3 Comparing two proportions (test used in the example).
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.

Technical Engineer
Focuses on tracking code, SDKs, and API integrations, with practical guidance on implementation checks and troubleshooting.