The challenge
Three stakeholder requests, all landing on Day 1, plus four data‑layer issues to navigate.
Product
Medium-sized campaign
Hard deadline: Thursday. AI Accounting feature launch. 4 days.
Growth
A/B test SMS in onboarding
Clear hypothesis. No deadline. ~1 hour of work.
Sales
Re-engagement campaign
No clear goal yet. Broad. Lowest urgency. Highest ambiguity.
And the broken data layer:
- ·VAT obligation status — no updates since August. Critical for the AI Accounting launch.
- ·Onboarding step 3 — no event tracked. Only the first and last steps are.
- ·Lead source = sales — no attribute. Can’t distinguish sales-touched leads in the CEP.
- ·user_type values — inconsistent ("freelancer", "Freelancer", "FL", "SME", "sme", "small_business").
My read
The data problems are the real test. Triage is theater. Anyone can prioritize three tickets — fewer can ship through a broken layer.
- —Don’t wait for clean data. Patch, prioritize, ship.
- —Triage by deadline × bandwidth × clarity, not by who shouts loudest.
- —Every “I need data” request is also “I need to talk to the data team.” First Slack message of the morning.
Framework 01 · Triage
Monday triage: deadline × bandwidth × clarity.
Monday triage ranks competing stakeholder requests on three axes — deadline, bandwidth, and clarity of the ask — instead of seniority or volume. Anything under an hour ships the same day; hard deadlines get the mid‑week build slots; ambiguous requests get scoped before they get built. The order is decided by the work, not by who shouts loudest.
Framework 02 · Patches
Patch, don’t block.
Patch, don’t block is a data‑layer pattern: every broken field or missing event gets two answers at once — a workaround that ships today (a CSV match, a regex rule, an escalation) and a structural fix queued for next week. Campaigns never wait for clean data; the data layer never stays broken.
Each data issue gets a workaround that ships today, plus a fix that lands next week. The pattern: unblock now, fix the layer later.
-
VAT obligation stale → escalate, then carve.
Slack the data team immediately. Ship the AI Accounting campaign without VAT‑segmentation. Add VAT‑aware audience as v2 the moment data is fresh.
-
Missing onboarding step 3 event → don’t block.
The Growth A/B test measures completion of the final step. Step 3 visibility is nice‑to‑have, not blocking.
-
No "lead_source = sales" attribute → CSV‑match patch.
Sales exports leads they touched in the last 60–90 days. I match against SME leads in the CEP and flag with a custom attribute.
-
Messy user_type values → regex now, dropdown next.
OR rule (or regex) catches all variants for today’s campaigns. Backlog: replace free‑text field with strict dropdown + retroactive cleanup.
If this is your Monday
Triage, patches, and a launch that still ships Thursday — this is the retainer work. 30 minutes tells us if it fits.
Framework 03 · Campaign
I chose the AI Accounting launch for end‑to‑end design. Reasoning: clear hypothesis, sharp deadline, useful as a launch playbook.
- —Audience. Freelancers + SMBs with VAT‑tracking behavior in last 90d, scoped first to DE. Excludes the universal holdout.
- —Channels. In‑app primary, email secondary, SMS for re‑activation only. No push for launch — too noisy.
- —Timing. Wave 1 Thursday morning Berlin time. Wave 2 (non‑openers) the following Tuesday. 7‑day cliff before trial trigger.
- —Personalization. Three layers — country/locale, VAT regime, business stage. Liquid templating.
- —Success metric. Trial activations measured against holdout, not opens. Secondary: % completing onboarding step 3 (after fix lands).
Framework 04 · Measurement
Universal holdout. Incrementality, not opens.
A universal holdout is a randomized, persistent slice of the audience excluded from all CRM intervention — every campaign, every channel, all year. Comparing it against the contacted audience is the only defensible way to measure true incremental revenue, because open rates and clicks never reveal whether a customer would have converted anyway.
That’s why the AI Accounting launch above counts trial activations measured against the holdout as its success metric — not opens, not clicks. The methodology I pushed for in large‑scale merchant CRM and would carry into any lifecycle engine.
Full artifacts
Working doc with the live thinking process + slide deck. Available on request.