What to analyse besides CTR
CTR tells you an ad was interesting. It does not tell you it was true, profitable, or worth repeating. Optimising toward it reliably produces cheap attention and expensive customers.
The problem with CTR as a target
Any metric that can be improved by exaggeration will eventually be improved by exaggeration. A hook that overpromises lifts CTR immediately and damages conversion rate, refund rate and retention later — in that order, and often in different reports owned by different people. By the time the damage is visible, the creative has been declared a winner and scaled.
CTR is still useful, but as a diagnostic, not a target. It answers one narrow question: did the creative earn attention? Pair it with the metric that answers the next question, and it becomes informative again.
Six metrics that decide things
- Hook rate (3-second views / impressions). Isolates the opening from the rest of the video. When hook rate is strong and hold rate is weak, the promise is not being paid off.
- Click-to-conversion rate. The bridge between media and product. A falling ratio with stable CTR is almost always message match, not audience quality.
- Cost per qualified outcome. Not cost per lead — cost per the event your commercial team recognises as real. Everything upstream of it is a proxy.
- Payback window by cohort. The only metric that tells you whether scaling is safe. A campaign at target CPA with a 14-month payback is a cash flow problem wearing a success costume.
- Incremental lift. Platform-reported conversions include people who would have converted anyway. Geo holdouts or conversion lift tests are the only honest answer.
- Creative concentration. What share of spend sits on a single concept? Above roughly sixty percent, you are one fatigue curve away from a bad quarter.
If a metric cannot change a decision this week, it belongs in an appendix, not a dashboard.
Read the funnel as ratios, not absolutes
Absolute numbers move with budget and hide diagnosis. Ratios isolate causes. Impression-to-click identifies the creative. Click-to-landing-view identifies page speed and tracking. Landing-view-to-initiate identifies message match and offer clarity. Initiate-to-complete identifies friction in the form or checkout. When CPA rises, one of those four ratios moved — find it before touching bids.
Segment before concluding
A blended metric is an average of things that behave differently. The same campaign can be excellent on Android in Poland and disastrous on iOS in Sweden, and the blended number will read as mildly disappointing. Segment by market, device, placement and new-versus-returning before you conclude anything. Most 'the algorithm changed' stories dissolve on the second segment.
Decide what a good week looks like in advance
The most useful discipline in analytics is not a tool. It is writing down, before launch, which metric will decide whether the test succeeded and what threshold counts as success. Do this and the analysis takes twenty minutes. Skip it and every review becomes a negotiation about which chart to believe.
Takeaways
- Treat CTR as a diagnostic for attention, never as an optimisation target.
- Read the funnel as four ratios so you know which stage actually moved.
- Define the success metric and threshold before launch, not during the review.