Hook rate vs hold rate: define what you measure

Define the event behind the name

A dashboard’s “hook rate” can conceal a different calculation from the same label in another report. Before comparing hook, hold and thumbstop rates, write down the counted event, denominator and reporting window. If those differ, the labels may describe different questions. Those definitions determine whether two results belong in the same comparison and which creative question they can help investigate.

This guide proposes a reporting contract; it does not assign undocumented definitions to Meta. In our production in October 2026, attention metrics inform creative hypotheses rather than acting as a complete verdict. creative fatigue meta ads provides the fatigue context. The immediate task is to know what a number means before using it to choose an opening or reject a full ad.

Two hypothetical reports may both describe attention while dividing by different populations. One uses impressions; the other uses qualifying starts. Their percentages should not be read as interchangeable measures, even if the dashboard headings sound similar. Preserve the named events and date range, then align the cohort where comparison is appropriate. Only afterward ask whether the opening or later explanation deserves inspection; the numbers alone do not establish which creative change will help.

An early-attention measure asks whether someone reached a defined opening event. A continued-viewing measure asks whether a viewer stayed to a later point. Choose the events explicitly. A source may count a particular duration, a fraction of the video or another documented action.

Do not assume that similarly named rates use the same denominator. One could divide by impressions while another divides by people who started viewing. Those results cannot be read as consecutive steps without checking the source definition. For a hypothetical report, write “opening event divided by eligible impressions” in the column note rather than using a familiar label whose meaning the team has never agreed.

Source: Turntake editorial checklist (proposed), October 2026.

1. Opening
2. Later

Compare the same cohort and reporting window

Keep market, placement, reporting period and video length visible. A short clip and a longer explainer expose different opportunities to reach later viewing events. A comparison across unrelated windows can also mix creative changes with changes in delivery context.

This does not mean every comparison is impossible. It means the conclusion should match the available controls. “Opening A reached the defined event more often in this supplied window” is narrower than “A is the better ad.” Keep source fields and exclusions beside the analysis. If the denominator is unavailable, report that limitation rather than reconstructing it from a rounded displayed rate.

1. 01Name counted eventDefine numerator
2. 02Name eligible populationDefine denominator
3. 03Match cohort and windowKeep comparisons interpretable
4. 04Inspect creativeWrite a bounded hypothesis

Use the shape of attention to ask a question

An opening can attract attention while the next beat fails to explain the product. Another ad can begin quietly but carry a clear demonstration for those who stay. Inspect both the opening and the transition into the body before deciding what to change.

1. Opening A
2. Opening B
3. Opening C

In our production in October 2026, a high stop-rate did not reliably settle the wider outcome question. The public lesson is qualitative: examine the full journey and keep the creative hypothesis specific. If the proposed problem concerns the handoff from opening to explanation, changing only the opening may not address it. Read the spoken thought and watch the next shot together.

Avoid benchmarks without a comparable source

Do not copy a percentage threshold into a production rule unless the source, metric and context justify it. A benchmark for another format or reporting definition can create false certainty. A team can still compare its own consistent records without inventing a universal target.

State whether an observation concerns one ad, one window or several comparable examples. Keep exact business results out of public production guidance. The practical output is a question worth testing: perhaps the product becomes recognizable too late, or the first explanation repeats the opening instead of advancing it. Those hypotheses can guide an edit without claiming that one rate predicts a business outcome.

Measurement questionRequired definition

• Measurement questionOpening attentionRequired definitionExact early event
• Measurement questionContinued viewingRequired definitionExact later event
• Measurement questionRate comparisonRequired definitionShared denominator context
• Measurement questionCreative conclusionRequired definitionObserved limits and hypothesis

Pair the metric review with creative inspection

Describe the opening without seeing its rank first, then compare the descriptions with the supplied report. This can reveal a mechanism worth investigating while reducing the temptation to praise whichever clip carries the larger number.

Pretest ads with synthetic personas: compare pairs covers synthetic persona pretests, which can explore clarity questions but do not supply observed audience behavior. How many ad creatives? Plan useful variety first covers creative variety across a test set, and Meta Andromeda: what creative variety changes discusses the platform context without replacing careful metric definitions. Keep these tools in their own roles. A synthetic preference, a visible creative difference and a measured viewing event are three different kinds of evidence.

Save the calculation contract with the hypothesis

Record event, denominator, window, exclusions, source and the creative question being proposed. Keep the definition version with the report so a later dashboard change does not silently break the comparison. If the definition changes, start a new comparable series or recalculate the old data where possible.

Turntake’s production approach describes Turntake’s production approach. This reporting method helps a creative team interpret supplied data; it is not an assertion that Turntake currently exposes a cross-account analytics dashboard.

For a suggested final check, ask another operator to calculate one example from its numerator and denominator. They should reach the reported value using the written definition. Then ask what creative conclusion that value alone supports. The first question checks arithmetic and the second checks interpretation. Both are necessary before a metric becomes a production decision.

Resolve a definition mismatch before ranking

Suppose one export labels a three-second event as the opening measure while another labels the first quarter of a video that way. Keep them as separate fields until their definitions are reconciled. Do not merge the columns because their names sound similar. The useful next action may be collecting a consistent export rather than editing a creative. This hypothetical mismatch shows why the calculation contract belongs beside the number, not only in a distant dashboard help page.

Inspect the source fields without inventing rate names

Meta’s official Business SDK Insights object lists distinct video-action fields. Its report-run object includes date and status fields. That supports keeping the selected field and reporting window in the evidence record. It does not establish that a custom label such as hook rate has one universal formula.

For a hypothetical export, keep the original column name beside the team’s display label. Add the calculation separately and preserve the raw numerator and denominator where available. When a dashboard changes its label, compare the underlying field and calculation before deciding that the metric changed. Conversely, the same label can conceal a changed calculation.

The distinction matters when comparing early attention with continued viewing. Two fields can describe different events even if both are summarized as a rate in a presentation. Keep the raw event identity, the local formula and the creative interpretation in separate columns. Then a reviewer can inspect the arithmetic without accepting the interpretation, or accept the observed difference while questioning the proposed creative cause.

Who wrote this We made the ads first, for SmartSolar in Denmark, Sweden and Ireland. Then we built Turntake. Where Turntake comes from