Rules
Part of Web analytics for marketers: the parts worth your attention
Web analytics for marketers: user, session, event, conversion
How marketers should read four web analytics terms: what a user, session, event and conversion each count, and what a number can never establish.
What to take away
- A user, a session, an event and a conversion are product constructs with definitions, not plain facts about people.
- Each definition has a limit, and the limit is part of the answer.
- Totals arrive late and tools disagree for named reasonsscope, identity, time, filtering and models.
- Test every definition with a short drill before anyone gets dashboard access.
Web analytics terms sound more certain than the data behind them. A marketer who knows what each term counts, and what it leaves out, can read two conflicting reports without assuming that either is wrong. Begin with the business question, framed in a marketing analytics strategy, then inspect the product definition and the implementation.
Four terms and what each one cannot tell you
Each core term is a construct built by the tool. The table gives the working definition, the limit that goes with it and a counterexample worth keeping in the glossary, so reviewers hold the limit in mind before they read the number.
| Term | What the tool counts | Limit | Counterexample |
|---|---|---|---|
| User | An identifier-based product construct | Not a person | One person on two devices may appear under two identifiers |
| Session | A tool-defined grouping of activity | Not a customer visit | One visit can contain several page views, and session rules differ by product |
| Event | A recorded interaction under a trigger | Not every eligible interaction | An interaction with no trigger leaves no record |
| Conversion | A selected event | Not automatic business value | Someone chose which event counts |
Read the limit column before the definition column. A conversion count is a count of the event someone selected, so it answers how often that event fired and leaves what it was worth to a separate calculation.
Why totals arrive late and tools disagree
Totals are late because collection and processing have latency, so yesterday's figure can still move. Tools disagree because scope, identity, time, filtering and models differ, not because one of them is broken. Settle which difference you are looking at before reporting a gap.
Google's explanation of how Analytics works describes collection and processing for that product. It supports product interpretation, not a claim that a local implementation captures every eligible interaction. The federal Digital Analytics Program adds a warning from practice: that combining the wrong dimensions and metrics or using the wrong report can give misleading or even completely wrong data about content engagement.
Missing data is a related trap. A gap can be a true zero or a failed tag, and only evidence tells you which. A consent gap is one cause, as the piece on reporting Canadian data shows for click identifiers.
Sessions need a product definition
Google separately documents its GA4 session dimension and the dimensions associated with session acquisition. Those definitions should not be assumed to match another tool or a company's customer visit. When two tools disagree, compare the dimension first, and read GA4 attribution models for the wider GA4 picture.
The federal Digital Analytics Program's glossary of terms defines a session, or visit, as a period of time during which a user interacts with a website or app. That is one program's definition, not a universal rule, so write down whose definition your report uses.
Session Definition Verification Checklist
- Read the metric help text
- Check property, date, time zone, filters, identity
- Inspect thresholding, sampling, modeling, freshness
- Compare event and business-source totals
- Record the permitted conclusion
If the answer changes a budget or customer experience, reproduce it outside the presentation. Keep the query, configuration, quality note and review owner with the decision record.
The glossary and the interpretation drill
A useful glossary includes examples and counterexamples. State what a term excludes, how it can fail and which source resolves a dispute. Then run a short interpretation drill before granting dashboard access.
- Give reviewers one user row, one session row and one event row.
- Ask what each row can and cannot establish.
- Record every disagreement in the glossary.
- Grant editing rights on dashboards only after the drill.
The drill catches hidden assumptions about people, visits and outcomes before an executive meeting.
Keep the rows used in the drill, because the next reviewer can reuse them and compare answers. The drill turns an argument about a number into a short edit of a definition.
Answers that need a record
A useful answer names the decision owner, the population, the period, the method and the action, and states what the result cannot establish. Save the definition and the date with the answer, and turn the record into a routine as in measuring marketing attribution.
If the same question returns, compare the records before changing the metric. That step exposes a changed population or source instead of treating every difference as new behavior.
Attribution is the standing example of a number that is easy to over-read: it assigns credit and does not prove cause. For more on credit that quietly inflates results, see common attribution mistakes.
