The 3-Second War: the creative scoreboard

Yesterday you learned to name every part of an ad. Today you learn the four numbers that grade those parts — and the chain that tells you exactly which one is bleeding.

1Account metrics judge the buy. Creative metrics diagnose the ad.

In the media-buying course you lived inside account-level numbers: CPM, CPA, ROAS. Those are the right metrics for grading the buy — are you paying a fair price in the auction, are conversions coming in under target. But ask CPA a different question — "why did this particular video flop?" — and it goes silent. CPA is a single lagging score at the end of a long pipeline. It can't tell you whether the ad failed in the first frame or on the landing page.

So creative needs its own scoreboard: metrics that measure how the ad performs as a piece of content, stage by stage, before money ever changes hands. There are four that matter, and they sit in a strict order — each one is the audience that survived the stage before it.

• Thumbstop / Hook rate — 3-second video plays ÷ impressions. Did the first frame stop the thumb? This is the front line of the 3-second war.
• Hold rate — ThruPlays (or 15-second views) ÷ 3-second plays. Of the people you stopped, how many stayed for the middle?
• CTR (outbound) — outbound link clicks ÷ impressions. Of everyone who saw it, how many wanted what you're selling enough to leave the feed?
• CVR — conversions ÷ landing-page visits. Of those who clicked, how many actually did the thing once they arrived?

Each grades a specific part of the anatomy you decomposed in Day 2. Hook rate grades the first frame and the thumbnail. Hold rate grades the footage and the middle. CTR grades the desire built by the body plus the CTA. CVR grades the match between the promise in the ad and the page it lands on. You couldn't even ask these questions until you named the parts — that's why Day 2 came first.

2The diagnostic chain: where it leaks is what to fix

Here's the move that makes these four numbers worth more than the sum of their parts. Read them as a sequence, find the first stage that under-delivers, and you've localised the fault. The pattern:

• Low hook rate → weak first frame. People aren't even stopping. Fix the opener — thumbnail, first 3 seconds, pattern interrupt.
• Strong hook, low hold → boring middle. You earned attention and squandered it. The footage drags or the promise of frame one isn't paid off.
• Strong hold, low CTR → weak desire, CTA, or offer. They watched the whole thing and didn't want it badly enough to click.
• Good CTR, low CVR → creative-to-landing mismatch. The ad sold one thing; the page delivered another. This isn't a creative-content problem so much as a continuity problem.

Notice this chain is exactly what the learning loop will read in Week 4. When you dissect winners and losers in Day 17, you won't just ask "did it win?" — you'll ask "where in the chain did the winners separate from the losers?" The scoreboard you learn today is the instrument the whole engine reads later.

Let's put numbers on it. Two videos in the same test, both served to 100,000 impressions:

Now a second creative with the same 27 conversions but a completely different leak:

Stare at those two panels. On CPA alone they look like twins — both produced 27 conversions on 100k impressions. Judge them only by that lagging number and you'd treat them identically, probably kill one at random, and learn nothing. Read the chain and they're opposites: A needs a new middle, B needs a new opener. Same outcome, completely different fix — and only the creative scoreboard can see the difference. One cheap re-edit on each, instead of two dead ads.

3Leading indicators, not the verdict

One guardrail before you fall in love with these numbers. Hook, hold and CTR are leading indicators — fast, cheap, available within hours, and perfect for diagnosing where an ad is leaking. They are not the verdict on whether the ad is a winner. The verdict is still the deepest real-money event: CPA, ROAS, cost-per-lead. We'll formalise this as the fitness function in Day 16, and it echoes the media-buying rule — optimise for the deepest event that matters (media-buying Day 4).

Why this matters: if you crown winners by hook rate alone, you will breed a stable of scroll-stopping ads that sell nothing. A shock open, a celebrity look-alike, a "wait for it" — these spike the thumbstop and tank the till. The vitals tell you how to treat a creative; the money event tells you whether it deserved to live. Today's metrics are the diagnosis. The winner-crowning logic comes in Week 4, and it's strict about not being fooled by leading indicators — which, incidentally, is why you'll set a minimum-data bar before trusting any of these reads (a noise trap we'll meet head-on in Day 16, echoing the ~50-events / learning-phase patience from media-buying Day 5).