Open the CSV in Excel, or import it into Google Sheets and choose comma-separated values. Save your working copy in a controlled location. The CSV contains the rows below and space for your notes; it has no macros, formulas or live account connection.
How to use the template
Write the mechanism before the asset
Explain why the proposed creative should change behavior. A different background color is a treatment description, not a customer hypothesis. A clearer demonstration of product size may reduce uncertainty; an earlier price may change which visitors continue. Those mechanisms create different outcome and guardrail choices.
Choose a design that matches the claim
Use a randomized setup for a causal comparison and document its assignment unit. If the platform distributes impressions adaptively across ads, label the result a delivery observation. The template supports both workflows but does not turn one into the other.
Check feasibility before production
Use the sample-size planner for a two-arm binary conversion experiment. Estimate the required recruitment time, then decide whether the business can wait. If the answer is no, reduce the number of concepts, choose a materially different treatment or make a limited operational decision without claiming statistical proof.
Keep the decision attached to the evidence
When the experiment closes, record the effect estimate and uncertainty as well as the winner label, if any. Include failed tracking checks and exclusions. A valid inconclusive result can prevent an unsupported rollout and still improve the next hypothesis.
Preview the worksheet
Every row shown here is included in the download. Scroll the table horizontally on a small screen.
| Planning field | What to record | Illustrative example |
|---|---|---|
| Business decision | The choice the result will inform | Whether to replace the control landing-page promise |
| Audience | Eligible population and exclusions | New paid visitors in one serviceable market |
| Hypothesis | Expected mechanism and falsifiable prediction | A clearer delivery timeline reduces uncertainty and improves accepted inquiries |
| Control | Exact existing version and immutable reference | Current hero copy; archived version A |
| Treatment | The specific difference and fixed elements | Change the timeline promise only; preserve offer and form |
| Design | Randomized experiment or observational comparison | Randomized two-arm landing-page experiment |
| Assignment | Experimental unit and allocation rule | Unique eligible participant; persistent 50/50 assignment |
| Primary outcome | One outcome and exact denominator | Accepted inquiries per assigned participant |
| Guardrail | A harmful outcome that should block adoption | Inquiry quality falls below the pre-agreed tolerance |
| Baseline and MDE | Comparable baseline and meaningful effect | 5% baseline; 20% relative increase to 6% |
| Sample plan | Method, power, confidence and exclusions | 95% confidence; 80% power; 8158 participants per arm |
| Timing | Recruitment horizon and conversion maturity | At least 16316 total participants plus outcome maturation |
| Stopping rule | Predefined completion and failure rules | Fixed horizon; pause only for tracking failure or agreed safety condition |
| Quality assurance | Assignment and event-delivery checks | Verify persistent assignment and one accepted event per inquiry |
| Result | Effect estimate, uncertainty and data completeness | Leave blank until the planned analysis |
| Decision | Adopt, retain control, or gather more evidence with reason | Record what the result supports and what it cannot establish |
Worked review example
Illustrative treatment: replace a vague 'fast service' promise with a supported description of the appointment process. Keep price, audience and form unchanged. The primary outcome is an accepted inquiry, with qualification rate as a guardrail.
If inquiries rise but qualification falls, the business decision depends on the mature number and value of qualified opportunities. A higher form conversion rate alone does not settle whether to adopt the treatment.
What this template does and does not establish
The sample-plan example is for a two-arm, equal-allocation, independent binary-outcome experiment. It is not appropriate for 20 simultaneous ad variants, clustered geographic tests or revenue significance analysis without a different design.