Most performance dashboards drown you in fifty metrics, and most teams spend most of their time looking at the wrong three. When you start generating dozens of creative variants a week, the noise gets worse. Here are the only four numbers you actually need to make decisions, and the seven you should stop checking.

This post is the measurement framework we recommend to every team that adopts the platform. It is not novel. It is a deliberate distillation of what works after watching hundreds of teams overcomplicate their reporting and slow down their iteration cycles. The goal is not a beautiful dashboard. The goal is to know what to ship next by Friday afternoon.

Why traditional ad metrics break with generative volume

The metrics you grew up with — CTR, CPC, CPM, conversion rate, ROAS — were designed for a world where you ran four creatives per campaign and watched them for two weeks. That world is gone. When you run forty creatives in a week, those metrics still exist but they tell you different things, and the rituals that grew up around them (the Monday performance meeting, the end-of-quarter creative review) actively slow you down.

The new failure mode is analysis paralysis. With forty variants, you can always find an interesting pattern in CPC by hour-of-day segmented by audience. Most of those patterns are noise. The teams that win at high-volume creative testing are aggressive about ignoring almost all of the data so they can actually move.

Metric #1 — Thumbstop rate (the only top-of-funnel signal worth watching)

Thumbstop rate is the percentage of impressions where the user stopped scrolling long enough to watch at least three seconds of your creative. Different platforms call it different things — Meta calls it "hold rate at 3s," TikTok calls it "engaged view rate" — but it's the same concept.

This is the only top-of-funnel metric that maps to creative quality without being heavily distorted by audience targeting or bidding strategy. CTR conflates creative quality with offer strength. CPM conflates creative quality with auction dynamics. Thumbstop isolates the question we actually care about for creative testing: did this visual earn three seconds of attention?

The benchmark we use: a creative that hits 30% thumbstop is a strong winner, 20-30% is a fine performer that's worth iterating on, under 20% gets killed without a second test. The bar will be different for your industry, but the spread will be similar.

One important rule: compare thumbstop only within the same placement. A 9:16 Reel and a 1:1 feed ad will have wildly different thumbstop benchmarks. Stacking them on the same chart and ranking by absolute thumbstop will mislead you.

Metric #2 — Cost per outcome (the only mid-funnel signal worth watching)

Whatever your actual business outcome is — purchase, signup, demo booked, free trial started — your second metric is the cost to acquire one. CPA, CAC, cost per signup, whatever the local name is at your company.

We deliberately do not say "ROAS." ROAS is contaminated by your pricing decisions, your discount strategy, and your AOV — none of which are creative levers. Cost per outcome is clean: holding offer and targeting constant, if creative A produces signups at $4 and creative B produces them at $7, A is the better creative. The math is too obvious to argue with.

The discipline here is patience. Cost per outcome is noisy at low sample sizes. We require at least 50 outcomes per variant before we declare a winner on this metric. If you do not have 50 outcomes per variant in 72 hours, your spend is too thin and you cannot use this metric for creative selection — you have to fall back to thumbstop.

Metric #3 — Hook freshness (the metric you've probably never tracked)

This is the one we evangelize hardest because almost nobody tracks it explicitly. Hook freshness is the rate at which your top-performing creative's thumbstop rate decays over time after launch.

Why it matters: a creative that hits 35% thumbstop on day one but drops to 18% by day fourteen is fundamentally different from one that hits 28% on day one and stays at 26% on day fourteen. Conventional ranking would pick the first creative. Creative-fatigue economics pick the second.

To track it, store the day-one thumbstop and day-fourteen thumbstop for every creative, and graph the decay curve. After a quarter of data you will have a feel for which prompt patterns produce fatigue-resistant creative and which produce flash-in-the-pan winners. The fatigue-resistant patterns are gold — they let you run a single creative for six weeks while you generate variants for the next quarter.

This is also where the prompt-tagging discipline we mentioned in the workflow post pays off. When you know which prompt produced each creative, you can correlate prompt structure with fatigue resistance. Over time you build a private knowledge base of "prompts that stay fresh" that no competitor has access to.

Metric #4 — Variant differentiation (the qualitative metric)

This one is qualitative, not quantitative, but it is the metric that prevents the failure mode where all your creative testing converges on a single winning visual and your account collapses when it fatigues. Variant differentiation is a one-sentence weekly answer to: are the creatives we are testing this week meaningfully different from each other?

The test is simple: lay this week's top four creatives on a wall and ask a person who has not been involved in their production to describe each one in one sentence. If they describe them in similar terms ("a person using the product, against a soft background"), your variants are not different enough. You are testing local optima, not exploring the creative space.

If they describe them in genuinely different terms ("a close-up of the product texture," "a wide shot of a real-world environment," "a flat-design illustration," "a UGC-style talking head"), your testing is healthy.

The reason this matters: A/B testing only learns when the variants are different. If you test seven nearly-identical blue backgrounds, you learn which shade of blue is best — which is a true answer to a useless question. You want to know whether blue beats red, lifestyle beats pack-shot, or illustration beats photography. Those answers compound. The shade-of-blue answer does not.

The seven metrics to stop checking weekly

This list will feel uncomfortable. That is the point.

• CTR. Useful at the campaign level, useless at the creative level because it's dominated by offer strength.
• CPM. A reflection of your audience and bidding, not your creative.
• Frequency. Track at the audience level monthly, not the creative level weekly.
• Engagement rate (likes/shares/comments). Tracks creative virality, which is almost orthogonal to creative performance. Many high-engagement ads convert poorly.
• Quality score / relevance score. Useful for diagnosing why an ad has stopped serving. Not useful for ranking winning creatives.
• Reach. A function of your budget and audience size, not your creative.
• Video completion rate. Looks important, almost never affects business outcomes once you control for thumbstop. The viewer who stopped at three seconds was always going to stop. The one who watched to ten seconds was always going to watch.

You will get pushback on this list. Senior stakeholders are emotionally attached to CTR and frequency. The argument that works in those meetings: "we can revisit these any time we have a hypothesis that needs them. We have stopped tracking them weekly because tracking them weekly slows down decisions without changing the decisions."

The weekly ritual that holds it together

The four metrics need a forcing function or they will get drowned in the existing reporting cadence. The ritual we recommend is a 30-minute Monday morning meeting with a single document open. The document has four columns (one per metric) and a row per creative shipped in the last seven days. The four columns get filled in live during the meeting. At the end of the meeting, three decisions are made: which creatives to scale, which to kill, and which prompt patterns to generate more of next week.

That's it. No quarterly reviews, no creative deep-dives, no end-of-month performance recaps. The meeting that matters is the one where you decide what to ship next week. Everything else is theater.

How to set this up without buying new tools

You do not need a new analytics platform. Every metric in this post is available in Meta Ads Manager, TikTok Ads Manager, and Google Ads. The work is not pulling the data — the work is deciding to ignore the other forty columns.

Build one spreadsheet (or one looker dashboard, or one slice of whatever BI tool you already use) with exactly four columns plus a creative-identifier column. Populate it weekly. Hide every other dashboard from the team. If someone asks for additional metrics, ask them what decision they would make differently with that data. If they cannot answer, you have your justification for keeping the dashboard simple.

What happens after six months of this

The teams we work with that adopt this framework typically see three changes within a quarter. First, decision speed roughly doubles — the Monday meeting goes from contentious to mechanical. Second, the creative win rate (defined as variants that beat the previous winner) climbs from maybe 15% to over 25%, because the team stops testing micro-variations and starts testing meaningful differences. Third, and most importantly, the conversation in the room shifts from "what does this data mean" to "what should we make next." That second conversation is where creative careers get built.

Simple metrics force creative thinking. Complex metrics absorb creative thinking. Pick the simple ones, and watch the team get faster, more confident, and more interesting in their output. That is the actual value of measurement — not the dashboard. The decisions.