Blog RSS feed

Everyone announced the death of tracking when ATT landed and cookies started crumbling. The funeral was premature. Measurement didn’t vanish, it migrated: off the client and onto the server, out of the browser and into CAPI and server-side tagging, away from per-user rows and toward privacy-safe aggregates. The signal changed shape, lost some resolution, and quietly changed owner. Whoever controls the server-side endpoint now controls the data, which is usually the platform, not you. Across 200M-plus devices I see less granularity and more modeling, but the spend still flows and the optimization loops still close. So stop mourning the device identifier. Start asking who owns your conversion endpoint, because that is where the real tracking moved.

Ask an ad network how many conversions it drove and it will tell you a beautiful number. Of course it will. It is grading its own homework, it gets paid on the grade, and its attribution window conveniently swallows every conversion that would have happened anyway. This is not malice, it’s incentive, and incentive beats integrity every single time you leave it unchecked. The entire MMP and neutral-measurement industry exists for exactly one reason: somebody who doesn’t get paid on the answer has to count. When two networks each claim the same install, the math is obvious and the conflict is structural. So I never accept a partner’s self-reported numbers as truth. I accept them as a bid for credit, and credit gets adjudicated by someone neutral.

Post-IDFA, a lot of what gets called “attribution” is modeled inference wearing a measurement costume. We lost the deterministic join, so we infer the link from timing, geography, and device signals, then report the result as if a human clicked and converted. That is fine, until you forget the confidence interval and start moving real budget against the point estimate. I’ve watched teams optimize aggressively toward a channel whose lift was statistically indistinguishable from noise. If your model can’t tell me how sure it is, it isn’t measuring, it’s guessing politely. Demand the error bars. Spend like you believe them. Modeled truth is still useful; modeled truth mistaken for ground truth is how you set money on fire.

Treat your UTM taxonomy like a free-text field and you will pay for it forever. Facebook, facebook, and FaceBook are three campaigns to a dashboard, and no analyst catches it until the quarterly numbers refuse to reconcile. Casing, delimiters, abbreviations: every inconsistency silently splits one truth into many, and aggregation quietly lies. I run a pipeline that ingests billions of ad events a day, and the dirty secret is that most “data quality” problems are really upstream naming problems nobody governed. So I treat the taxonomy as a schema. Enumerated values, a validator at link generation, an owner. You do not let people invent column names in production. Stop letting them invent UTM values.

If a product manager cannot articulate why a feature is being built, in one sentence, tied to a user problem or a business outcome, then the feature should not be built yet. The inability to explain the “why” is not a communication problem. It is a thinking problem. Clarity of purpose precedes quality of execution.

Talking to users once a quarter is research. Talking to them weekly is a habit. The difference matters because product judgment decays. What you learned about your users three months ago may no longer be true. Regular conversations do not just generate insights. They keep your instincts calibrated and your assumptions honest.

Shipping is not the goal. Changing something is. A team that ships ten features and moves no metric is less effective than a team that ships one feature and changes a behavior. The obsession with output volume is a defense mechanism, it is easier to count what you shipped than to measure whether it mattered.

If you can only ship one thing this quarter, what would it be? That question strips away the politics, the pet projects, and the backlog inertia. Whatever you answer is your actual priority. Everything else is negotiable. If the team cannot agree on the answer, you do not have an alignment problem. You have a strategy problem.