Costs
Part of Web analytics for marketers: the parts worth your attention
Eight web analytics mistakes for marketers and how to repair each one
Eight web analytics mistakes for marketers, from tags read as truth to attribution read as effect, with the check and the repair step for each one.
What to take away
- A metric can be calculated correctly from a broken implementation.
- Users, sessions and people are different units.
- A rate without an eligible denominator and a mature outcome can mislead.
Web analytics for marketers mistakes often begin with a familiar label that hides a product rule. The repair is to reconnect each number to its collection, definition, population and decision.
| Failure | Why it misleads | Repair |
|---|---|---|
| Tag equals truth | Firing may not equal accepted outcome | Reconcile to source system |
| Users equal people | Identifiers split or merge activity | State identity rule |
| Sessions equal visits | Session logic is product-specific | Publish definition |
| Page-view success | Traffic may not complete a task | Measure outcome and guardrail |
| Bad campaign taxonomy | Rows split or misclassify | Validate allowed values |
| Recent conversion rate | Outcomes are immature | Use closed windows |
| Attribution as effect | Credit lacks a counterfactual | Run a suitable test |
| Silent implementation change | Series is no longer comparable | Version and mark break |
How to check and repair each failure
- Tag equals truth.Compare the tag count with the source system for one week, such as orders in the store platform. Use GA4 DebugView and Google Tag Manager preview mode to confirm that each action fires the event once.
- Users equal people.Check whether one person appears under several identifiers, for example before and after a login or across two devices. Write the identity rule into the report label.
- Sessions equal visits.Read the GA4 session definition before comparing session counts with another tool's visit counts.
- Page views equal success.Pick one completion event, such as a confirmed order, and report the conversion against it.
- Bad campaign taxonomy.Open the campaign names report and look for case splits such as "Email" and "email", and for typos. Fix the naming template first, then map the old values to the new ones.
- Recent conversion rate.Report conversions only after the outcome window has closed, so late outcomes do not make the rate look worse than it is.
- Attribution as effect.Run a holdout or a controlled test before claiming that a channel caused an outcome.
- Silent implementation change.Log every tag, consent and site change with its date, and mark the break in the series.
Cross-domain tracking and consent settings are two common sources of silent gaps. Check the GA4 cross-domain configuration and the consent status for each domain before trusting a drop in traffic.
User label verification steps
- Inspect raw network and server evidence
- Compare event, order, lead, billing counts
- Check consent, browser, device, market coverage
- Review duplicates, bots, internal traffic, late data
- Retest critical journeys after every release
- Correct affected dashboards and decisions
Do not treat a user label as a person
Google's Analytics glossary defines a user as a person who interacts with a measured app or site. That product definition does not verify that a reported identifier maps one-to-one to a human or to a customer.
Beware of ratios whose numerator and denominator use different scopes. Landing-page orders divided by all users, for example, combines a session-level entry with a user-level population. Name the unit in the metric.
Worked cases in web analytics for marketers examples show which defects recur. Keep a defect register with the start time, affected pages and events, estimated impact, decisions exposed, repair, restatement and verification. Do not smooth away a broken period.
Prove the correction
A mistake is not closed when a dashboard changes. Trace the correction from the source record through processing, reporting, exports and the decision owner. Preserve the prior value, the reason, the date, the reviewer and the affected actions. Then test whether the same defect can recur. This turns an error list into an operating control.
The GAO guidance on assessing the reliability of data describes testing data against its intended use and documenting the work. It does not certify a local dataset, but it supports a written check of completeness, accuracy and known limits before release.
The FTC expects objective advertising claims to have a reasonable basis before they are made. Use that principle to test any public statement about a site's performance. It does not define the right method for every claim, and it does not replace legal advice. The common web analytics for marketers questions article collects the doubts that usually follow a repair.
