Problem
A growth team cannot form hypotheses or run experiments until it knows how customers enter the funnel and what they do on the site and in the app. Without a deliberate setup, event data turns "junky": events have different shapes, nobody remembers what they mean, and bots or monitoring data pollute the counts. Vendor pricing also grows with event volume. The author, a growth engineer, gives the order in which to build instrumentation and the rules that keep it healthy.
Example from the article: a visitor lands on the homepage from a Google search, clicks Sign up, submits the account form and becomes a registered user with a UUID. Each step is an event tied to one identity.
Solution in brief
Send an event for every touch point, with a fixed schema and an anonymous id plus a user id. Grow the setup in stages: web analytics tag, vendor SDK, event router, in-house stream pipeline. Validate events against a schema and describe their purpose in a tracking plan.
Questions this case answers
- How to structure event instrumentation and its delivery pipeline so that behavioral data stays consistent, privacy-compliant and affordable as the product and the number of data consumers grow
- How to design analytics events and a schema so that growth teams can analyze the funnel from anonymous visit to registered user?
- How to decide when to move from a single analytics tag to a vendor SDK, an event router or an in-house event pipeline?
- How to keep analytics data clean and privacy-safe as the number of events and teams grows?
Tools and methods
- Web analytics tag — traffic sources, sessions, UTM parameters on the marketing site.
- Vendor SDKs (Amplitude, Mixpanel) — custom events from frontend and backend.
- Event router (Segment, Tealium) — routes events to warehouse, martech tools and ad networks.
- Apache Kafka, Amazon Kinesis, Amazon S3 — in-house event stream and storage.
- Schema validation and tracking plan — keep events consistent and documented.
Google or a one-time email code. No password.