Use attribution to operate campaigns and well-designed experiments to investigate causality. An inconclusive study does not prove that advertising has no effect.
Someone already knows your brand, searches its name and purchases after clicking an ad. The platform may record a valid conversion. That record does not establish whether the person would have purchased without advertising. Confusing these questions can shift money toward capturing existing demand while reducing investment in creating new opportunities.
Causal measurement is part of today’s discussion about automated advertising. Every decision does not need an experiment; the useful skill is recognizing when the available evidence is sufficient. This is the working framework Patagonia recommends when a business needs to reassess a meaningful advertising investment.
Define the decision before choosing a method
Start with a specific decision: retain a brand campaign, expand into a market or add a channel. Write down the budget at stake and the result that would change your recommendation. “Does our advertising work?” is too broad to guide a useful test.
Attribution allocates credit across recorded interactions. Incrementality estimates what changed because of an intervention. Google offers Conversion Lift for supported scenarios; eligibility, volume and setup need checking in each account. Access to a feature does not establish that a study can answer your business question.
Separate a comparison from causal evidence
Comparing October with September can identify changes, but it also mixes seasonality, promotions, pricing and service capacity. Investigating causality requires a defensible comparison group or condition. A controlled test needs planning; randomly stopping ads and checking tomorrow’s sales is not a reliable substitute.
Agree on the primary metric, comparison unit, duration and effect size that would justify a decision before the test begins. Consider customers crossing between locations or devices and other campaigns contaminating the comparison. If these conditions are not reasonable, describe the analysis as exploratory rather than causal.
Sources: Think with Google: Use incrementality testing for effective marketing measurement
An example of how the decision changes
Hypothetical scenario: a multi-location business sees many purchases attributed to branded search and considers doubling its budget. It first reviews fulfillment capacity, new customers and total sales by location. A measurement specialist then assesses whether comparable locations and sufficient volume make a controlled study feasible.
If attributed conversions rise while total results barely move, that creates a question; it does not independently prove the campaign is useless. If a sound experiment identifies a gain, the incremental contribution still needs comparison with its cost. The point of the study is to avoid conclusions that exceed the evidence.
Read uncertainty alongside the estimate
A single number can look precise even when evidence is weak. Ask for the uncertainty range, exclusions and changes that occurred during the study. A result compatible with both a small and a large improvement cannot support an exact return promise.
Keep revenue distinct from contribution. Additional orders carry product, service, discount and operating costs. Budget decisions should consider what remains after those costs and incremental advertising spend. An ROI calculator can explore assumptions, but it does not measure causality or replace a properly designed experiment.
When the business has limited volume
A small business may not generate enough transactions to distinguish a modest effect from normal variation. In that situation, avoid pretending the estimate is precise or splitting an already small sample by campaign, city and day. Improve record consistency first and observe periods that reflect the complete sales cycle.
Limited decisions remain possible with imperfect information: exclude irrelevant searches, repair a broken destination page or constrain spend when fulfillment capacity is exhausted. Those actions do not require a broad causal claim. Reserve an incrementality study for a question whose value justifies the effort and whose design is feasible. Agree in advance what an inconclusive result will mean: continue observation, redesign the study or retain a cautious budget while collecting stronger evidence.
What to prepare this week
Create a one-page decision brief and a consistent export of spend and results. The first useful finding may be “we do not yet have enough data to test this,” paired with a plan to collect it. That assessment prevents spending on a study whose result cannot be interpreted.
At Patagonia, this review connects measurement with the commercial objective and inquiry follow-up. The priority is to decide whether to retain, correct or investigate an investment further, rather than produce an impressive chart.
- Define one decision and one primary metric.
- Record promotions, price changes and capacity constraints.
- Review account eligibility and feasibility with the measurement owner.
- Agree which conclusions the test can and cannot support.
- Keep a record of findings and decisions so each test builds on earlier learning.
REFERENCES
Sources and review
Source documentation consulted on October 24, 2026. Platforms and features may change; check the official source before implementation. The English adaptation retains the source review date of the original article.
- About Conversion Lift Google Ads Help
- Use incrementality testing for effective marketing measurement Think with Google
This guide provides general guidance. It does not guarantee search rankings or replace a review of your business, locations and data.