What this checkpoint covers
The first module of the intermediate analytics level: the event model in GA4Google's event-based analytics platform., the data layer that Tag Manager (GTMTag management system.) reads, and the routine that catches failures. If you have done the booking flow tracking plan exercise, several scenarios will feel familiar; the point is to make the same decisions without the worked answer in front of you.
Task
- Write a one or two sentence answer to every question before opening the explanations. A choice without a reason does not count.
- Compare your reasoning, not just your conclusion, with each explanation.
- For every question where you missed part of the reasoning, add a line to your own tracking plan template: a column, a negative test or a weekly check that would have caught it.
Questions
- A marketing manager asks for three new events:
contact_form_submit,quote_form_submitandcallback_form_submit. The quote form is the only one that produces a sales lead. What do you build? - A developer shows you a parameter,
plan_interest, arriving correctly in DebugView. A week later the parameter is missing from every report. What happened, and what can you recover? - The developer proposes pushing
purchasewhen the visitor clicks "Pay now", because the confirmation page is slow. What do you say? - Customer tier (free, trial, paid) is being sent as a parameter on every
page_view. Someone asks how many paid customers visited the pricing page last month and nobody can answer. What is wrong? - Your weekly check shows the ratio of lead events to leads in the CRM, normally steady, has fallen sharply from a Wednesday. A release that day redesigned the form. The container version has not changed. The paid search team wants to cut bids because lead volume is down. What do you do?
- After a consent banner update, users fall in one market and conversions fall further. A colleague wants to rebuild the conversion trigger. What do you check first?
- A single-page application adds a router push for each view. Page views per session double overnight. Why?
- Items from a customer's previous
purchasestart appearing on their nextadd_to_cart. What is missing?
Answers explained
- One
form_submitevent with aform_nameparameter for the contact and callback forms, andgenerate_leadfor the quote form. One event with a parameter keeps the list short and the reports split it; the exception is an action that means something different to the business, and a sales lead does. The events and parameters guide sets out the rule and its exception. - The parameter was collected but never registered as a custom definition. Collection and reporting are separate: the data sits with the event and in the export, but the reporting interface builds dimensions only for registered parameters, and registration is not retroactive there. Register it now and record the gap; the same events guide explains why registration belongs in the schema before release.
- Push when the booking or order is confirmed, not when it is attempted. A declined card or a failed server call would otherwise become a purchase. Speed is solved with the push timing and, where navigation follows, the callback pattern in the data layer guide, not by counting attempts.
- Customer tier is an attribute of the person, so it belongs in a user property. As an event parameter on page views it cannot answer user-level questions cleanly. Set it as a user property, register it at user scope, and stop repeating it on every event.
- Treat it as missing events until proven otherwise, and stop the bid change. Spend and traffic continuing while tracked leads fall on a release date is the signature of a trigger that no longer matches the redesigned form. Fix the trigger in preview, document the gap as a known hole, and confirm the ratio recovers. The weekly routine in the data quality guide is built to catch this before money moves.
- The consent configuration, before the trigger. A fall in users after a banner change may be the tags now correctly waiting for consent. The data quality guide lists consent gaps mistaken for errors as a failure of their own; check the consent settings and the consent rate by market first.
- The automatic page views from history changes are still switched on beside the router push. Every view after the first is now counted twice. Keep the router push, which carries fresh context, and switch the automatic version off, as the data layer guide describes for single-page applications.
- The reset before each e-commerce push. The data layer merges nested objects across pushes, so without setting
ecommerceto null first, old items leak into the next event. Add the reset to the specification and to the negative tests.
Where to go next
If you missed the reasoning on questions about timing or merging, reread the data layer guide. If you missed the diagnosis questions, run the weekly routine from the data quality guide on a property you have access to. Then move to the second module, starting with the attribution models exercise.
