In-House Analytics Tag Testing: What a Crawler Must Check

Export or describe the tests you currently run. For each one, record the journey, starting state, expected event, required fields, failure condition and evidence needed to investigate a failure. Separate page coverage from behavioural coverage. Visiting every URL will not exercise a form submission, a client-side route change or a multi-step checkout. Likewise, seeing an analytics request does not establish that its payload represents the action correctly. Prioritise the journeys that affect important reporting. A small suite with explicit expectations can be more useful than a large crawl that reports tag presence without explaining what passed.

Chrome's Network panel reference explains how to inspect requests and preserve a log across page loads. Use a manual session to understand the expected behaviour before encoding it in an automated test. For each journey, record the action, relevant request, payload fields and destination. Keep evidence that lets a reviewer distinguish a missing event from a duplicate, an incorrect value or a request blocked before it left the browser. Define what the test can observe. A request leaving the browser is different from confirmation that the destination processed it. If your requirement includes downstream receipt or reconciliation, scope a separate check with the appropriate system access.

A useful initial matrix includes the following cases, adjusted to your implementation: • A fresh visit with no prior browser state. • A journey before any consent selection, where applicable. • The same journey under each supported consent choice. • A client-side navigation and a full page reload. • A repeated click or submission that could create duplicate events. • A successful action and a validation failure. Start each case from a defined state. Reusing a browser profile without resetting it can make a test pass because an earlier run left a cookie or session behind. Record the build, test configuration and time so failures can be compared across releases. Use synthetic accounts and transactions where possible. Limit stored payloads to what is needed for the check, and review logs for personal information before sharing a failure report.

A failed check should name the journey, expected result, observed result and useful evidence. Include enough context for someone to reproduce it without guessing which environment or browser state was used. Keep unstable infrastructure separate from a confirmed behavioural failure. If the test environment was unavailable, report that the check could not complete. Do not quietly turn a timeout into evidence that tracking disappeared. Agree what blocks a release and what triggers investigation. An essential conversion event and a non-critical content event may deserve different escalation. The tool needs an owner who can distinguish a site regression from a test that no longer represents the product.

Compare the subset you need with the work required to maintain it: browser execution, scheduling, authenticated journeys, expected payloads, evidence storage, alerts and changes to the site. Include the time needed to investigate false alarms. Ask for a pilot covering a small number of important journeys and deliberately introduce a missing event or wrong field in a safe environment. The pilot should identify the fault and provide enough evidence for another person to find it. Keep the claim precise. These tests can provide evidence about specified technical behaviours. They do not, by themselves, establish legal compliance. Our build versus buy guide can help assess ownership, while a project brief can define the first testing scope.

Is finding a tag enough to pass a test? No. Tag presence is one check. Behavioural tests should specify the trigger, event count, required values and observable destination. Can this run on our own infrastructure? That can be a project requirement. Include browser execution, network access, credentials, storage, updates and an ongoing owner in the scope. Will an automated scan prove compliance? It can check agreed technical behaviours and preserve evidence. Compliance conclusions require the relevant policy and legal assessment, beyond the crawler's result.