In this article
Define the incident before looking for a culprit
Name the business outcome, affected scope and observation period. “Ads stopped working” can mean lower delivery, fewer submitted forms, fewer qualified leads or a broken reporting event. Those incidents require different investigations.
Preserve the report used to identify the issue. Record the extraction time, conversion definition and date basis. Delayed conversions can change a later report and make the original observation difficult to reproduce.
Set a baseline that reflects the question. Compare equivalent weekdays and promotion conditions where possible. If the business changed its offer or service area, a simple previous-period comparison may mix several effects.
Build a timeline from several systems
Google Ads documents a ChangeEvent resource for supported account changes. That is useful evidence within its documented coverage and query limits. It is not an unlimited history of everything that affected the business.
Bring in website releases, checkout changes, promotions, stock availability and tracking modifications too. Ad-platform history cannot explain an unpublished landing-page edit or a sales team that stopped contacting leads.
| Timeline entry | Evidence to retain |
|---|---|
| Campaign edit | Object, old state, new state and timestamp |
| Website release | Changed page, deployment time and intended behavior |
| Tracking change | Event definition and verification result |
| Offer change | Price, conditions and active dates |
| Operational issue | Stock, sales follow-up or service interruption |
| Performance observation | Source, metric and extraction timestamp |
Normalize time zones before ordering events. An edit recorded in the account time zone and a deployment recorded in UTC can otherwise appear in the wrong sequence.
Separate plausible mechanisms from coincidental changes
Suppose a hypothetical campaign receives a new headline on Monday and orders decline on Tuesday. The same week, the store removes a discount and changes its checkout. The headline is one hypothesis, not the established explanation.
Describe the mechanism each hypothesis predicts. A checkout failure should affect purchase completion after checkout starts. A misleading ad could change visitor intent and qualification. Reduced delivery should appear earlier in the acquisition chain.
Look for the stage where behavior changed. This narrows the investigation without inventing a causal story. If the available data does not include that stage, ask for the relevant event or native report.
Use comparison groups carefully
An unaffected campaign or page can help distinguish a local issue from a wider change. But check whether it has comparable audiences, offers and tracking. A different product category may respond differently to seasonality.
A comparison is evidence, not automatically a controlled experiment. If several systems changed together, an observational report may only establish likely contributors. State the limits clearly.
Avoid recommending a full rollback solely because it is easy to name the most recent edit. A rollback may be useful when a verified technical defect exists. Otherwise define which change is being reversed and what result would test the hypothesis.
Write the postmortem as a decision artifact
Use four sections: established facts, plausible explanations, unresolved evidence and the next action. Assign an owner to each investigation. Include the expected observation if the proposed explanation is correct.
Record what the team would do differently next time. That might mean separating releases, saving configuration snapshots or validating an event immediately after a change. The purpose is a better decision process, not a dramatic explanation.
Revisit the memo when mature outcomes arrive. Correct the conclusion if the later evidence changes it, while preserving the original observation and decision context.
Ask 1ClickReport to organize the evidence
1ClickReport can analyze connected marketing performance. Supply the verified change timeline as context and obtain native history or deployment records where connector coverage is insufficient. Do not ask the analyst to invent missing edits.
Investigate this campaign incident using the connected performance data and supplied change timeline. State account, dates, time zones and conversion maturity. Identify where the acquisition-to-outcome chain changed. Separate verified facts, plausible mechanisms and missing evidence. Do not treat sequence as proof of causation. Propose the smallest investigation or reversible test and name the observation that would support it.Frequently asked questions
Does a decline after an edit prove that the edit caused it?
No. Check other changes, reporting maturity and a plausible mechanism.
Does Google Ads change history cover website deployments?
No. Collect those records separately and align their timestamps.
Can AI complete the postmortem without every data source?
It can organize available evidence, but it should leave unsupported explanations unresolved.
Sources and research notes
Platform statements are linked below. The checklists, worked examples and analyst prompts are original editorial frameworks. Examples labeled hypothetical are not customer results. Research checked October 13, 2026; account availability may differ.