Measurement and iteration — treating ads as experiments

hard

Learn with your AI

Open this lesson in your favourite AI. It'll walk you through the why, explain the demo, and quiz you on the try-it list.

Open in Claude Open in ChatGPT

Why this matters

Commercials used to be made once and broadcast many times. Modern AI-driven commercial production runs the opposite loop: ship fast, measure, iterate, replace. The feedback signals — watch-through rate, CTR, conversion, cost-per-action — are the brief-pressure-test the old system never had. The filmmaker who closes the loop between creative and performance data is the one who builds an actual practice rather than a portfolio of one-off spec spots. This is the workflow the top 1% of DTC creative teams run.

Demo

A production loop that actually learns: (1) Ship 5 hook variants this week. (2) Watch-through at 3s, 15s, 30s and CTR at endframe — pull from Meta Ads Manager / TikTok Ads Manager / Google Ads. (3) Pick the two best hooks. Ship 5 variants of middle-act for each next week. (4) Pick the best middle. Ship 5 CTAs. (5) Now you have a winning :30 spot proven at every beat. Total production time: 3 weeks. Total AI cost: often under $300. Total learning: what your audience actually responds to — which feeds the next brief.

# Weekly ad-learning loop (template)
 
Week 1 — HOOK test
  5 variants × :06 pre-viz
  Matched $100/variant spend on Meta + TikTok
  Metric: 3-second watch-through rate
  Winner: pick top 1–2
 
Week 2 — MIDDLE test
  Use Week 1 winner as hook
  5 variants of middle (demonstration / testimonial / explanation)
  Metric: 15-second watch-through rate
  Winner: pick top 1
 
Week 3 — CTA test
  Lock hook + middle
  5 endframe/CTA variants (direct / soft / curiosity / FOMO / social-proof)
  Metric: CTR and cost-per-click
  Winner: your ad is born
 
Log at the end of every test:
  What the audience picked (the winner)
  What SURPRISED you (winner you didn't expect)
  What this teaches you about the AUDIENCE not just the ad
  → Feeds next month's brief.
 
Tools:
  Meta Ads Manager, TikTok Ads Manager, Google Ads
  Foreplay / Motion / Creative OS for creative analytics dashboards

Try it yourself

Run a 3-week hook/middle/CTA loop on one product at small budget ($300 total). Document what surprised you. Surprises are the insights; everything else is execution.

For a past campaign, reconstruct the loop: which variant was the winner, which the loser, and why? If you can't answer 'why,' you're measuring without learning.

Build a weekly creative retro: 30 minutes, review what shipped, what performed, what's next. The retro is where the learning compounds.

Audit: do your ads test hook, middle, CTA in isolation? Or are you shipping 'a whole ad' and hoping? Isolated testing scales knowledge; monolithic testing doesn't.

Share a test summary with a peer for blind review — they name the winner from metrics alone. If they can't pick it, your metric isn't sharp enough.

Prompt your AI

Use these three in order. Each builds on the one before.

1. Basics & terminology

In one paragraph, explain the hook/middle/CTA test loop — what gets tested in which week, what metric matters at each stage, and why this beats shipping a monolithic ad and guessing.

2. Why it works (the mechanism)

Walk me through how modern creative-analytics tools (Motion, Foreplay, Creative OS) attach performance data to creative variants — what they ingest, what they surface (thumbnail-level winners, hook-type trends), and why the workflow around the data matters more than the dashboard.

3. Advanced — application & what's next

I'm running creative for a DTC brand at $500k/month. Walk me through the team structure, the weekly production-to-measurement cadence, how many variants to keep in rotation, when to retire winners, and how to turn performance learnings into the next quarter's brief.