JuristAI: a GA4 audit and analytics instrumentation fix

This is a GA4 and web analytics audit engagement: we diagnosed the tracking, locked a baseline, and added event instrumentation to an existing React product so acquisition steps could be measured. JuristAI develops an AI-assisted legal platform with research, document and case-management features. Paid plans were available through Stripe, and its own team was actively developing the product. Autonomous’s role was to make the journey into that product easier to understand through measurement.

The initial audit found an analytics property without configured key events and a payment journey that was not represented as purchases in GA4. Acquisition attribution was also difficult to interpret. The question was practical: how could the team judge growth activity when the instrument did not reliably describe the steps toward becoming a customer?

This was an integration engagement within an existing product. JuristAI retained its application and engineering team; Autonomous contributed analytics diagnosis, implementation guidance and a focused instrumentation change.

Before touching the tracking we recorded what the numbers already looked like, so a later change could be judged as a real change in behaviour rather than better recording. The work began with the client’s GA4 and Search Console data and an inspection of the site. The diagnostic covered tag architecture, event completeness, conversion setup, attribution and data quality. It produced a prioritised set of issues rather than assuming every weak metric was a product problem.

A pre-engagement snapshot then fixed the comparison window for sessions, engagement, key events and organic-search measures. The purpose was to make the starting point explicit. Once tags or event definitions change, an apparent improvement can come from better recording rather than a different customer experience.

The implementation plan addressed conversion tracking, a form-start event firing on page load, canonical routing and campaign-tagging strategy. Those are different causes of misleading data and needed to be treated separately.

Workflow: Acquisition question → Locked baseline → Application events → Funnel analysis.

The infrastructure change went to the client’s own team with a runbook, and the tracking code went into their application through their normal review process. Autonomous supplied a CloudFront redirect function and a deployment runbook for the canonical-domain change. Strategy and implementation responsibilities were explicit, allowing the client to make the infrastructure change within its own environment.

For the application work, an Autonomous engineer added PostHog captures across the marketing navigation, signup form, dashboard entry points, Law Library and Cite Checker. The client merged that change into its main branch.

The events included context about the page and the component involved. That is useful in a product with several routes to signup: a visitor arriving through a tool limit is expressing a different intent from someone using the main navigation. The instrumentation creates a basis for separating those entry points rather than placing them all in one undifferentiated click count.

Installing tracking code is not the same as having data you can trust, and that difference matters before anyone spends money on what the report says. The delivered event calls remained in the application months later, through subsequent changes by the client’s team. Their continued presence shows that the implementation fitted into the existing codebase and was retained as surrounding interfaces evolved.

It does not establish that the receiving analytics project collected every event or that a funnel report was used to make a decision. We also identified taxonomy issues, including a path where a click and an impression share a click-event name. Those distinctions matter before using raw event totals as evidence of customer behaviour.

This is why a measurement engagement needs more than inserting capture calls. The definition of an event, the conditions that trigger it and its receipt in the destination all contribute to whether the resulting report can answer a business question.

The engagement was completed and delivered. The case demonstrates a locked starting baseline, targeted deployment guidance and signup instrumentation integrated into an existing React product.

The client received concrete work connecting acquisition questions to the application surfaces where intent appears. The next measurement decision would require checking receipt and taxonomy against the questions the team wants the data to answer.