Questera

Implementation worksheet · 2 min read

A holdout design worksheet for a lifecycle campaign

To estimate a lifecycle campaign's incremental effect, assign eligible users or accounts to a treatment group and a concurrent holdout before exposure. Keep assignment stable, define the primary outcome and observation window, and compare outcomes for the originally assigned groups. Report uncertainty and implementation failures alongside the result.

This is a design worksheet, not a claim that a campaign produced uplift. A before-and-after comparison can mix campaign effects with changes in traffic, seasonality or product behavior. The appropriate sample size and inference method depend on your baseline rate, effect of interest and randomization unit.

Put it into practice

1. Choose the unit

Randomize at the level where interference is limited. If teammates share a workspace and influence one another, account-level assignment may be more suitable than assigning individuals independently.

2. Define eligibility and assignment

Specify entry conditions, exclusions and a stable assignment mechanism. Record assignment before the send or action occurs. Preserve a holdout even if the treatment fails to deliver, so the analysis does not select only successful sends.

3. Choose outcomes in advance

Name one primary outcome, its observation window and the guardrails that could stop the test. Estimate the sample required for a useful result. If traffic is insufficient, state that limitation before launching.

4. Check implementation

Compare intended and observed allocation, duplicate exposures and cross-channel leakage. A holdout that receives the same campaign through another workflow is not a clean control for that intervention.

5. Report the full result

Show group sizes, conversion counts, rates, absolute difference and uncertainty from the chosen method. Distinguish percentage points from relative lift. Record whether the experiment was stopped early and why.

Holdout planning worksheet

Copy this structure into your review document and record your observed result for each row.

Holdout planning worksheet
DecisionRecord before launchReview
PopulationEligible cohort and exclusionsReproducible
AssignmentUnit and allocationStable
Primary outcomeEvent and windowUnambiguous
GuardrailsComplaints and adverse outcomesOwned
AnalysisMethod and stopping ruleChosen in advance

A failure worth checking

Comparing people who opened an email with people who did not open it does not isolate the email's effect. Those groups selected themselves after assignment. Analyze the original randomized groups for the intended causal comparison.

Common questions

Can I claim uplift from a few extra conversions?

Only with an analysis that accounts for uncertainty and the design. A small observed difference may be compatible with chance.

Must a holdout last forever?

No. Its duration should match the predeclared observation window and operational constraints. Do not repeatedly change it to obtain a preferred result.

Basis and scope

This is a proposed implementation method using illustrative examples, not a measured benchmark or a customer case study. Prepared with AI assistance. Validate product-specific behavior against current documentation and your own test environment.

Continue with Questera

Discuss your workflow