Keep an ad test hypothesis log you can act on

A test log should explain the next production decision. A folder of successful exports cannot tell you what changed, why it changed or whether the result answered the original question. Keep a short written record that travels with the actual ads, including comparisons that remain unresolved.

What decision should the test resolve?

Begin with a decision you can act on. For a hypothetical product demonstration, the question might be whether the opening should show the object working before the narrator explains it. That is narrower than asking which ad people like. The proposed production change is visible, and the editor can preserve the explanation and closing action while changing the opening.

Write the decision in ordinary language before selecting metrics. If the result would not change a production choice, clarify the purpose of the comparison. Creative fatigue and variation gives the wider portfolio context; this page focuses on the record for one test rather than the number of concepts to produce.

Write the hypothesis before the variants

Record the audience situation, the proposed creative change and the reason it might matter. Describe the expected observation without promising that it will occur. A useful hypothesis can be contradicted by the evidence or remain unresolved. If every possible result would count as confirmation, the statement is too vague to guide a test.

In our production in October 2026, the hypothesis is written before the ad is made, and a narrow iteration changes one intended creative element. That convention helps us identify what the editor changed. It does not establish that delivery conditions are controlled. NIST’s guide to selecting an experimental design distinguishes design choices by objective and factors; choose the actual design separately from this production log.

Source: Turntake’s proposed editorial test record, October 2026; the example is hypothetical.

For a hypothetical organizer ad, the question concerns the opening demonstration. Showing loose tools and showing a removable divider are distinct visual treatments of the same product. The log keeps the rest of the approved ad fixed and records the actual exports used. If delivery conditions change, the editor has a place to describe that deviation instead of treating the report as a clean answer. This example contains no campaign result.

Save the reviewed versions with stable identifiers and a reference to the exact exports. A title such as demonstration variant does not identify which voice, caption track or opening was uploaded. Keep the reference export as well as the editable recipe, so another person can inspect the difference without trusting a summary.

Record any production deviation. If the second version also received new music, that belongs in the log even when the brief requested only a picture change. Correct the variant or acknowledge the broader comparison. Do not silently treat the original intention as a description of the files that were actually delivered.

1. Variant A
2. Variant B

Define the signal and comparison

Choose the main observation before reading the result. Define how it is measured, the reporting interval and which assets belong in the comparison. If a derived metric is used, keep its numerator, denominator and available source fields with the record. Hook rate and hold rate explains why similarly named video metrics need explicit definitions.

In our own analysis in October 2026, a stronger stopping signal did not reliably establish a stronger downstream outcome. That is a reason to inspect the wider path, rather than a new universal rule about which metric matters. Preserve the distinction between an opening that attracts attention and an ad that supports the intended action.

1. 01State the decisionName the intended creative change
2. 02Bind reviewed versionsCheck what was actually delivered
3. 03Record the observationKeep the metric definition
4. 04Choose the next actionPreserve limits and uncertainty

Record conditions as they happen

Keep a dated note when the destination, audience, offer, placement or reporting setup changes. A remembered change after the result arrives is harder to assess. The log should show which conditions were intended to remain comparable and which were actually different.

A short deviation field is more useful than an elaborate form that nobody updates. Record the fact, when it happened and which comparison it affects. Creative test confounders provides the diagnostic checklist when a result has several plausible explanations. Do not repair an inconvenient history by overwriting the original setup record.

Separate observation from interpretation

Write what the report showed first, then explain what you think it means. Keep the reporting reference attached. An observation might identify a difference between specific versions during a particular period. The interpretation should state whether the design supports attributing that difference to the intended creative change.

Use an uncertainty field when evidence is missing or conditions differ. It can say that the files were comparable but delivery was not, or that the opening changed alongside the offer. Such a note protects the next editor from treating an ambiguous result as a permanent instruction about the audience.

ObservationInterpretation

• ObservationSpecific assets and reporting intervalInterpretationProposed explanation of the difference
• ObservationDefined source fieldsInterpretationAssumptions and competing explanations
• ObservationWhat the evidence containsInterpretationWhat decision the evidence supports

Choose the next action

End with a concrete decision: retain the opening, revise the explanation, investigate measurement or run a narrower follow-up. Name who will make that decision and which evidence they still need. A request for more testing is incomplete unless it says what uncertainty the next comparison should resolve.

Planning clip reuse helps turn the decision into a new edit without repeating the same visual sequence everywhere. Link the next brief to the original log entry rather than copying a confident conclusion into several documents. The original evidence and its limitations should remain available as the idea develops.

Carry the learning into production

Keep confirmed working instructions separate from tentative interpretations. A practical improvement such as identifying an approved source clip can enter the library record immediately. A claim that a visual approach works for every audience needs evidence beyond one comparison. Preserve the scope of what was observed.

Meta creative testing covers the broader testing workflow, while Turntake’s production approach explains how editable assets and review fit together. In our production in October 2026, a person approves activation. The log supports that decision and the next creative brief; it does not activate campaigns or certify a result automatically.

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