One reader identity across eight publishing properties

The situation

Eight properties had accumulated analytics independently over a decade. Event names collided, the same concept was called four things, and a reader moving from a newsletter link to the app looked like four different people. Every question from the subscriptions team started with a two-week reconciliation.

The constraint

A news business cannot pause instrumentation for a rebuild, and traffic is spiky by nature — a breaking story produces an order-of-magnitude jump within minutes. Consent handling had to be correct per jurisdiction from day one, not added later, and the properties had separate engineering teams with their own release cadences.

What we built

A tracking plan that is a contract

Roughly 90 events with names, properties, types, required flags, an owner and a written definition, versioned in the repository. Payloads are validated in CI. An event that is not in the plan does not reach the warehouse, which sounds severe and is the only thing that keeps a plan alive past month three.

Identity resolved deliberately

Anonymous, registered and subscribed identities linked through a documented merge rule, with the timing of the merge specified — because the difference between merging at login and merging at first authenticated event changes every funnel number and someone will eventually ask why.

Consent as a first-class state

Consent state travels with the event, per jurisdiction. Downstream models know which events they can use for which purpose, so a marketing question and a regulatory answer can coexist.

A pipeline sized for the spike, not the average

Buffered ingest with backpressure and replay, so a breaking-news surge queues rather than drops. We load-tested at nine times average, which turned out to be roughly what the biggest story of the year produced.

Adoption over dashboards

Two workshops per property, a starter set of charts each editorial and subscriptions team helped define, and an office-hours slot for six weeks. The instrumentation was the easy half.

What changed

The subscriptions funnel is measurable end to end across properties and devices for the first time. The first insight to survive scrutiny was mundane and expensive: a registration wall placed one scroll too early on two properties, costing a measurable share of conversions.

What we would do differently

We wrote the tracking plan with the analytics team and reviewed it with engineering. It should have been written with both in the room. Three events had definitions that were unimplementable on iOS without a release cycle we did not have.

Outcomes

Events with an owner and definition

a minority → 100%

Registration-to-subscription funnel

unmeasurable → tracked end to end

Peak ingest sustained

several times average, no loss

Client identity withheld under a mutual NDA. Figures are illustrative — rounded and directional, meant to show the shape of the change rather than an audited result. We will walk through the real numbers, and how they were measured, under NDA on a call.