Testing framework proposal · Underdays
Fewer bets, deeper reads.
The account has grown fragmented — too many campaigns, personas and ideas for the spend behind them. This is a proposal to simplify the testing structure so the data reads clearly and every ad gets a fair test.
Where we are
2 inches deep
across 10 ideas
across 10 ideas
→
Where we're going
10 inches deep
on 2 ideas
on 2 ideas
What changes & why
7 changes01
Collapse the testing campaign to 1–2 ad sets (max 3), by persona/archetype — concepts underneath.
WhySeven personas spread across many ad sets makes the data hard to read and starves ads of spend (~£100 each isn't enough to validate). Fewer, cleaner buckets = objective winner-vs-loser calls.
02
Test UK-first — stop splitting tests across UK and US. Roll winners into the US.
WhyUS CPMs run ~3× the UK. Testing both is a double creative pipeline for little extra signal. One pipeline, cleaner reads, less work for everyone.
03
Cap each sprint at 2 focuses + 1 product launch — not 5.
WhySprint 15 currently carries ~5 focuses. That's too many ads for the spend, so nothing gets a fair test. Fewer focuses = each idea actually validates. Needs alignment with Daisy.
04
Set defined kill/scale gates — spend + conversions + CPA, with harder targets for higher-performing formats.
WhyPulled from the trailing 60 days. Partnership/whitelist judged hardest, then video, then static. Removes subjectivity, stops us pausing too early or graduating too late, and frees spend for untested ads. 14-day window for winners.
05
Run on a two-week sprint rhythm — Week 1 launch net-new, Week 2 iterate the winners.
WhyPredictable cadence, and it mirrors the model working at Granola through its scaling arc. Run the kill/scale report weekly (daily isn't needed at this spend).
06
Brief a deliberate awareness-level mix — problem / solution / product / most-aware.
WhyIf we only brief bottom-funnel product wear, we run out of things to scale within 2–3 sprints and performance plummets. We need top-of-funnel in the pipeline before we need it.
07
Tie the Explore vs BAU balance to account health.
WhyStruggling → more iterations of our bread & butter (econ statics). Healthy → more experimentation. Underdays' USP is strong enough that iteration reliably works — no need to over-create.
Open question — volume. Proposing ~60 ads per two-week sprint (~20 iterations in the off week). Treat this as a starting dial we tune once the new structure runs.
Format opportunities
diversifyCrack partnership ads
45% of spend at Granola, but they haven't taken off here yet. Worth a real push.
Podcast & green screen
More podcast in our own branding, plus green screen — the podcast + green-screen combo already tested well.
Social proof
Customer-review ads via Zoom interviews with real customers. Social proof is the clear gap vs other accounts.
Next steps
4 actionsAlign with Daisy on cutting Sprint 15 to 2 focuses + 1 launch — keep the everyday-athlete work (tank + workout songs are launching anyway). — team + Daisy
Pull the trailing-60-day benchmarks to ground the kill/scale gates before switching them on. — Pascu / Naomi
Get sign-off on UK-first testing. — team
Pascu reworks Sprint 15 focus over the weekend; Naomi sends the first two personas + the thread. — Pascu + Naomi
Underdays · Testing Framework v2Recap of Fri session w/ Naomi