Trackingplan

fig. tree

Schema · semantics & context

The context your AI needs, built automatically.

From the first hit, Trackingplan learns your schema, your naming, and what “revenue” or “signup” mean in your stack — and keeps that knowledge current as your data changes. No rules to build, no setup project, no time spent teaching AI your data.

How it works

  1. 01
    Understands what existsEvents, properties, values and providers discovered from real traffic.
  2. 02
    Understands what it meansConnect implementation-level data to the concepts your team actually uses.
  3. 03
    Understands what changedTrack how the model evolves across releases and changes in production.
  4. 04
    Keeps it currentNo spreadsheet or documentation project to maintain.

Auto-inferred tracking plan

Every event, property and value range, discovered from real-user traffic and kept up to date.

Semantic model

Raw events mapped to the concepts your team talks about, so anyone can ask about revenue, not about purchase.value.

Release awareness

What changed, when, and in which GTM version or app release. Every answer knows what shipped.

Living documentation

The plan you never had time to write, always current, with the evidence one click away.

“Trackingplan picks up everything automatically, allowing us to see the data and dig into it without us configuring things on our own.”
Persida Roata · Digital & Analytics Project Manager, MicroAnalytics

Ask it

Why is ‘country’ missing from our events? How do I fix it?Send Review the dataLayer and the main events — tell me if you see any implementation problem.Send
Start asking your data Book a demo

Ask it

Your schema, as it is in the wild

Trackingplan gives the agent the context it needs to understand your implementation — your stack, events, releases, history, campaigns, tolerances, and the patterns behind what usually breaks.

Event identity · providers · payload differences
Trackingplandemo workspace
You · Analyst

Which providers receive add_to_cart, and which don’t?

coverage · add_to_cart · by provider · 7dcompare · payload · by provider
Trackingplan

Six providers receive it: GA4, Meta, TikTok, Pinterest, Mixpanel and your server-side stream. Google Ads gets it as the AddToCart conversion on web only. Two gaps: Amplitude stopped receiving it when v90 removed its tag, and Bing has never had it, although Bing Ads is live.

Same event, different payloads: Mixpanel gets it without items[], and TikTok sends the variant SKU as content_id where everyone else sends the parent id. If you compare add-to-cart rates across tools, that is why they differ.

6 providers · google_ads web only · amplitude since v90, bing never
Continue this thread with the agentSend
Versions · historical changes
Trackingplandemo workspace
You · Analyst

Which events changed schema in the last release?

release_diff · v94 → v95 · all events · 24hsample_hits · changed properties
Trackingplan

Two events changed with v95, published yesterday at 17:02. checkout_step lost coupon on web — coverage went from 41% to 0% in the same minute, because the variable now points to ecommerce.coupon_code, which the checkout doesn’t set. And sign_up.method changed values: “google” became “Google”, so GA4 now reports two methods.

The other 29 events are identical to v94: same keys, same types, same coverage. Consent flags and provider coverage didn’t move either.

v95 · 2 events changed · 29 unchanged · web
Continue this thread with the agentSend

That’s why it can answer questions like these. It knows which event is which, where it goes, how payloads differ, what changed in the latest release, and what the data looked like before.

FAQ

About schema management.

?What does Trackingplan learn automatically?
tp

Events, properties, values, providers, naming patterns and the relationships between them, observed in real traffic. Nothing is declared up front: the model is built from what your implementation actually sends.

?Where does the semantic context come from?
tp

Trackingplan builds the model from your actual implementation and the traffic it generates, rather than relying only on a manually maintained specification. That is how it knows which providers receive an event, how the payloads differ between them, and which release changed what.

?Does the model stay up to date?
tp

Yes. As your implementation changes, Trackingplan continuously updates what it knows about your tracking: new events, new properties, new providers and changes across releases, with the history kept so you can see when something changed.

?How is this different from a tracking plan?
tp

A traditional tracking plan describes what your team intended to implement. Trackingplan builds a living model of what is actually happening in production, and keeps it current. When the two disagree, you find out from the data, not from a spreadsheet nobody updated.

Book a demo

Want to see your tracking model in action?

See what Trackingplan can understand from your real traffic.