First-party, consent-gated tracking
Browser and server-side events pass through the same consent check before they reach your warehouse. If a visitor declines, the event isn't sent. What's left is data you can defend.
Digital Marketing
One set of numbers. Tests that settle arguments.
Consent-first tracking, a clean event model, attribution finance accepts, and an experimentation program that turns conversion opinions into measured wins.
What we set up
Measure with consent. Decide with tests.
Clean data first, then a testing program that uses it. Each part works alone, and they work better together.
A tracking plan in your repo
Event names, properties, and owners, documented and versioned. Changes reviewed like code.
SchemadbtVersioned
Attribution finance accepts
Multi-touch where it helps. Incrementality tests where it matters. A media mix model when the data supports one.
Lift testsGeoMMM
Dashboards on one model
Looker, Metabase, or your BI tool, all reading one semantic layer. Self-serve for marketing, reconciled with finance.
LookerMetabaseSemantic layer
A testing backlog, ranked
Pre-registered hypotheses, ranked by expected impact and the traffic they need. Tests that can't reach power don't run.
A methodology log
Every change to attribution windows, filters, or definitions is written down and signed by marketing and finance.
Analytics
One source of truth across teams.
Three teams, three revenue numbers for the same week. That's usually when we get the call. The fix is plain: one warehouse, first-party consent-gated server-side tagging, an agreed method, and retiring the side reports.
- One warehouse — Snowflake, BigQuery, or the one you already run — as the agreed source.
- Server-side tagging on your own domain, applying the visitor's consent choice before anything is sent.
- Attribution calibrated with incrementality tests and re-baselined each quarter.
- Every dashboard reads one semantic model. No hand-rolled queries feeding the board pack.
Experimentation
Tests that are pre-registered settle arguments.
Most A/B tests end in a shrug: too little traffic, too many metrics, a result read too early. We write down the hypothesis, the primary metric, and the sample size before launch. Then we wait for it.
- Hypothesis, primary metric, and guardrail metrics written down before every test.
- Power check before launch. If your traffic can't detect the effect, we don't run it.
- Fixed-horizon or sequential analysis chosen up front. No stopping on a good day.
- Every result archived — including the tests that lost.
0
Analytics events sent before a visitor consents
80%
Minimum statistical power we plan every test to
1
Event schema, versioned in your repo
Quarterly
Methodology review with marketing and finance
Finance and marketing report the same revenue for the same week now. Test results stopped being a debate.
Questions
What buyers ask us first.
Is server-side tagging a way around consent or ad-blockers?
No. We use it for data quality and control, not to get around anyone's choice. Server-side events follow the same consent decision as the browser, and a visitor who declines isn't tracked.
How much traffic do we need for A/B testing?
It depends on your baseline conversion rate and the smallest change worth detecting. We run a power calculation first. If the numbers don't work, we suggest bigger changes or qualitative research instead.
Will you replace our analytics tools?
Rarely. We usually clean up what you have — GA4, your warehouse, your BI tool — and add server-side tagging and a tracking plan on top.
Can you make us GDPR-compliant?
We implement consent, retention, and data-access settings to match the policy your legal team sets. We're not lawyers, so we don't give legal advice.
Who owns the data and the setup?
You do. Everything runs in your accounts, your cloud, and your repo. We document it so your team can run it without us.
Ready when you are
Send us your three reporting tools.
A 30-minute call. We'll find where their numbers diverge and tell you why.