How to Measure Whether Creative Is Working

Measure the Intended Effect

Every campaign has one primary communication job: earn attention, change a belief, improve comprehension, create memory, or drive action. Pick the metrics that match that job before launch, and write them into the brief. A brand film judged on click-through rate will always look like a failure, and a direct-response ad judged on "sentiment" will always look like a success. Both readings are useless.

The single most common measurement mistake we see is not bad math. It's grading creative against a job it was never given.

Read the Signal in Layers

Creative works in a sequence: it has to hold attention, then register as your brand, then move someone to act. Measure it in the same three layers, because each layer diagnoses a different fix.

Layer 1: Attention

Metrics: 3-second holds as a share of impressions (thumb-stop rate), average watch time, and completion rate on short cuts. Most social ads lose around half their viewers in the first three seconds, so a thumb-stop rate north of roughly 25–30% is doing its job, and anything under 15–20% usually means the hook, not the offer, is broken. When attention fails, fix the first three seconds. Nothing downstream matters yet.

Layer 2: Branding

Metrics: aided and unaided recall if you can afford a lift study, and if you can't, branded search volume and direct traffic in the campaign window. A campaign that holds attention but produces no movement in branded search after several weeks is entertaining people on behalf of nobody in particular. The fix is earlier and more distinctive brand presence: your assets should be recognizable before the logo appears.

Layer 3: Action

Metrics: click-through rate, conversion rate, and the trend in customer acquisition cost across the flight. Judge trends, not single days. A healthy pattern for strong creative is CAC drifting down over the flight as the audience warms; CAC climbing while attention metrics hold usually means fatigue or a landing-page problem, not a creative problem.

When to judge

Not before the data can carry a conclusion. Our working floor: don't judge an ad before it has a few thousand impressions per variant, don't judge a campaign before two weeks, and never kill your control asset on 48 hours of data. Early numbers are noise wearing a trend costume.

A simple weekly scorecard

One table, filled in every Monday, per active asset. It takes ten minutes and ends most measurement arguments.

Two habits make the scorecard work. Keep the verdict column to a decision, not a description: keep, kill, iterate, or scale. And log qualitative signal in the same pass, because saves, shares, and the comments people actually write are early evidence of memory that the platform dashboards undercount.

Turn Results Into the Next Brief

The scorecard's real product isn't a grade, it's the next brief. Document which hooks, claims, characters, visual worlds, and calls to action moved which layer. "Founder-to-camera holds attention but doesn't convert; the sizzle close-up converts but can't stop a thumb" is a finding you can build a better ad from next month.

Do this for two or three cycles and creative stops being a debate about taste. You'll know your brand's hooks the way a chef knows their pantry. If you'd like help setting up the layers for your next campaign, book a 15-minute call and bring last quarter's numbers.