Analytics & Attribution
If the Ad Creates Demand, Name the Search It Should Create

Updated: October 2026
Google Ads API v26 adds Search Lift keyword bundles and new Search Lift measurements, giving analysts a clearer way to define which searches a campaign is expected to influence. The opportunity is not to stuff every branded query into one bucket. It is to register a defensible hypothesis before the result is visible.
What changed in API v26
Google added campaign associations for Search Lift keyword bundles and new metrics for generated searches, search lift and relative search lift. Those fields can support a repeatable export and QA workflow around a Search Lift study.
The release does not turn every search into a sale or prove long-term profit. Search behavior is an intermediate outcome. It can show whether exposure changed query activity inside the study design, then business teams must connect that movement to site behavior and revenue.
Write the keyword-bundle hypothesis first
| Bundle | Question | Guardrail |
|---|---|---|
| Brand | Did exposure increase searches for the company? | Exclude ambiguous names |
| Product | Did people search for the promoted product? | Separate generic category terms |
| Problem | Did the ad create category exploration? | Avoid overly broad queries |
| Campaign phrase | Did a distinctive message enter search behavior? | Register exact variants |
| Competitor or alternative | Did comparison behavior change? | Interpret carefully and legally |
Prevent bundle drift
Freeze the approved keywords, match logic, market, language and study dates. Store the bundle resource names and campaign associations. If the bundle changes mid-study, document the version and ask whether the analysis needs to split or restart.
Keep a control list for accidental overlap
Review misspellings, common words, product names shared by other brands and local meanings. A bundle that captures unrelated searches can produce a confident but useless lift number.
Interpret three layers separately
- Generated searches: the estimated additional search activity.
- Search lift: the absolute difference associated with exposure.
- Relative search lift: the proportional change from the baseline.
A large relative lift on a tiny baseline may represent little volume. A modest relative lift on an established category can represent more searches. Publish the absolute and relative context together.
Connect search lift to business outcomes without overclaiming
Compare the timing of lifted query groups with organic and paid search visits, qualified sessions, conversions and revenue. Treat the relationship as a funnel hypothesis unless the study design supports a stronger causal statement at each stage.
Use MTC’s lift-measurement configuration QA workflow and attributed branded-search guide as companion controls.
Publish an auditable readout
Include the campaign list, bundle definitions, study window, markets, absolute lift, relative lift, uncertainty, missing data and any changes. State what the test measured and what it did not. Avoid converting generated searches directly into revenue unless a separately validated model supports that step.
Pre-register the interpretation rules
Before results arrive, write what would count as a meaningful absolute lift, how uncertainty will be handled and which bundle outcomes would support or contradict the campaign hypothesis. Include a rule for mixed evidence—for example, brand searches rise while product and problem searches do not. This prevents the strongest-looking number from becoming the story after the fact.
Decide whether bundles are primary, secondary or exploratory outcomes. Primary bundles should be few and directly tied to the campaign strategy. Secondary bundles can explain how demand moved. Exploratory bundles can generate the next hypothesis but should not be presented as pre-planned proof.
Audit campaign-to-bundle associations
Export every participating campaign and its bundle resource before launch and at readout. A missing campaign can dilute the result; an unexpected campaign can contaminate it. Pair this mapping with the lift-configuration check so the final analysis documents both study enrollment and query definitions.
Frequently asked questions
Should a bundle include every branded query?
No. Include terms that answer the registered question and remove ambiguous or irrelevant variants.
Is Search Lift the same as incremental revenue?
No. It measures search behavior. Revenue impact needs downstream measurement.
Can keyword bundles be edited during the study?
Treat changes as measurement interventions. Version them and consult the study owner before editing.
