Keep a marketing change log that makes before-and-after reviews useful

Record a hypothesis before the edit

Write the problem being addressed, the change, and the outcome it should affect. A form-label repair may reduce submission errors; a service clarification may reduce wrong-service inquiries. Those are different hypotheses from increasing total demand. Assign a person to verify the effect. Without that note, the next review can turn whichever metric moved favorably into the retroactive purpose of the change.

Source checks: About change history · About the Experiments page

Include operational changes in the same timeline

Record closures, staff coverage, booking capacity, promotions, supply restrictions, and unusual local events that affected the business. Keep the log focused on changes relevant to interpretation rather than every daily task. A campaign adjustment made while the shop was closed cannot be reviewed as if customer handling stayed constant. These conditions help the owner decide whether the next fix belongs to marketing, staffing, or fulfillment.

Preserve measurement definitions

Document conversion action edits, new tags, consent changes, inquiry deduplication, and qualification-rule changes. A cleaner definition can reduce an event count while leaving real customers unchanged. Link the measurement note to the report period and preserve the old definition for historical context. Do not silently combine pre-fix and post-fix totals under a single unchanged label. This is especially important when the report is shared with someone who did not attend the implementation discussion.

Use platform history to verify implementation

Google Ads change history helps inspect account edits, while the website's deployment record and form-system notes establish other changes. Confirm that the planned edit was actually applied to the intended path. Record differences between the plan and implementation. Timing alone still does not prove impact, but verified implementation prevents a reviewer from evaluating a change that never reached the live campaign or advertised page.

Choose a comparison appropriate to the evidence

Use comparable operating periods and allow outcomes to mature. Where feasible, a supported experiment can compare a base and trial while holding other variables stable. A small business may need an operational before-and-after review instead; state its limits and check confounders. Do not manufacture statistical significance from a few inquiries. If several changes were made together, assess the bundle honestly rather than assigning the whole effect to one favorite edit.

Turn the review into the next specific task

Record whether the observed problem improved, worsened, stayed unclear, or revealed another bottleneck. Include the evidence and the next action. A useful log supports decisions such as keeping a fixed form behavior, clarifying the offer further, or collecting more comparable outcomes. It should not become a gallery of unverified wins. Keep the report concise enough that the owner can understand both the result and why the evidence supports only that conclusion.

Worked example: illustrative figures

For example, delivered inquiries rise from 10 to 15 after a contact-page edit. During the same period, the business also resumes full staffing and begins a promotion. The 50% increase is arithmetically correct, but attributing all five extra inquiries to the page is unsupported. The log records all three changes and checks form errors specifically, because that was the page edit's stated target. If errors fall while inquiry demand rises, the report describes both observations without claiming a measured causal share for the edit.

Questions about this task