Anchor on a North Star

Pick one metric that captures the value you deliver — qualified pipeline, MRR, repeat purchase rate. Every team metric should ladder up to it. Without that anchor, optimization scatters and budget follows.

The North Star should be customer-value-aligned (not vanity), measurable weekly (not just quarterly), and influenceable by marketing in the next 90 days. "Revenue" fails the third test; "qualified pipeline created" passes all three.

Once chosen, every channel and campaign should be able to answer one question: "how does this move the North Star?" Anything that cannot is either a brand investment (defend it as such) or a distraction (kill it).

Instrument GA4 for events that matter

GA4 is event-based by design. Define a clean event taxonomy: lead_submit, demo_booked, signup_complete, purchase. Attach revenue values where possible so attribution and reporting stay honest.

Use consistent naming conventions (snake_case, verb_noun structure) and document them in a tracking plan that lives in Notion, Confluence, or wherever your team actually looks. Tribal knowledge dies the day the analyst leaves; a documented plan does not.

Server-side tagging via GTM Server-Side or stape.io improves data quality dramatically — fewer ad-blocker losses, longer cookie life, and cleaner first-party data pipelines. The ROI on a 2-week server-side migration is usually visible in 60 days.

"A dashboard nobody acts on is a vanity project. Build for decisions, not for decoration."

Move beyond last-click attribution

Last-click flatters bottom-of-funnel channels and starves brand and content. Use data-driven attribution in GA4 or Looker Studio, then compare it to last-click — the gap is exactly the budget you are misallocating.

For longer B2B sales cycles, layer on time-decay or position-based attribution and triangulate against self-reported attribution from a "How did you hear about us?" form field. No single model is perfect, but the consensus across 2–3 models is usually directionally correct.

The point of attribution is not to assign perfect credit. It is to inform reallocation. Use it to ask "should I move budget from channel A to channel B next quarter?" and you will get more value than chasing methodological purity.

Use cohorts to expose retention truth

Group customers by acquisition channel and month, then plot retention or repeat-purchase curves. Channels that look cheap on CAC often look expensive on LTV. Cohorts surface that mismatch within 60 days.

A channel with 30% lower CAC but 50% lower 6-month retention is a money-loser dressed up as a win. Cohort analysis is the only way to see this clearly — blended averages hide the truth and let bad channels grow.

Build cohort views in BigQuery or Looker for the long term, not in spreadsheets. Spreadsheets are great for one-off analyses; recurring decisions need a single source of truth that updates automatically and gets reviewed in a standing meeting.

Build dashboards by audience

An exec dashboard has 5 numbers and a trend line. A channel-owner dashboard has 15 numbers and a diagnosis path. Stop building one dashboard for everyone — it serves no one.

For execs: North Star, pipeline, blended CAC, payback period, and one leading indicator. One screen, one minute of read time, one decision per number ("invest more, hold, cut").

For channel owners: full funnel from impression to revenue, segmented by campaign, creative, audience, geo. The job here is diagnosis, not summary — so optimize for clickable filters and drill-downs, not visual polish.

Respect privacy by design

Server-side tracking, consent mode v2, and first-party data pipelines are no longer optional. Build them in early — retrofitting after a regulator letter is ten times more expensive.

For EU-traffic-heavy sites, a properly configured CMP (Cookiebot, OneTrust, Iubenda) plus consent mode v2 typically recovers 30–50% of conversion data lost to non-consent users via Google's modeled conversions. That is a free uplift that costs only configuration time.