Campaign postmortem template: decisions from evidence

1. Define the result that missed expectations

A disappointing return on ad spend can reflect a weak campaign, immature sales data, a measurement error or an unrealistic expectation. Before the review, write one sentence identifying the agreed objective, observed result and reporting period. If the objective was only defined after launch, record that as a process finding. Do not rewrite the original brief to make a later expectation appear to have been agreed from the start.

This guide is for a completed campaign or a finished phase. The blog's troubleshooting checklist addresses a different need: diagnosing ads that are running now. A postmortem preserves learning for the next launch. It can also be useful after a successful campaign when the team needs to distinguish conditions worth repeating from outcomes that may have depended on a favorable coincidence.

2. Assemble evidence before the meeting

Prepare the original brief, approved constraints, spend export, event definitions, change history and lead-handling data. Record when each export was produced and the expected conversion delay. If sales are still developing, mark the commercial result as interim and schedule a later calculation. The meeting should not become a group search for which dashboard column everyone intended to use.

Divide statements into observations, explanations and unknowns. A form was unavailable for two hours is an observation if supported by logs. Therefore half our sales were lost is an explanation requiring additional evidence. We cannot identify affected visitors is an unknown and possibly a measurement improvement task. This distinction allows the team to investigate without prematurely agreeing on a single root cause.

3. Reconstruct what was known at each decision

For important events, record the time, information available, decision and owner. Distinguish what the team knew then from what became obvious after completion. Otherwise the review rewards hindsight rather than better judgment. Useful entries include launch, an offer change, a landing-page failure, a budget adjustment, a warning and the decision to pause.

Google SRE's blameless postmortem approach comes from system operations. The principle adapted here is to examine contributing conditions and prevent recurrence. In marketing, that means asking why the process allowed spend to continue while the form was broken. Possible findings include a missing check, unclear duty ownership or an alert nobody received. A blameless review still assigns specific responsibility for completing the corrective action.

Completed example: the cheap lead campaign

Illustrative USD data: Campaign A spent $600 on 30 inquiries, six qualified; Campaign B spent $400 on ten inquiries, six qualified. A’s CPL is $20 against B’s $40, but its cost per qualified lead is $100 against B’s $66.67. This does not prove that B will win a larger test; it shows why the next review needs qualification evidence.

Completed postmortem decision: inspect A’s offer-to-landing-page consistency and rejected inquiry reasons; retain the same qualification rule; define a capped follow-up experiment and a review date after equal maturation time. The downloadable worksheet preserves the observation, uncertainty and accountable owner. It contains synthetic data, not client results.

Review your report draft · Check cost per lead · Audit the next campaign

4. Work through a failure without overstating causality

Illustrative USD example: a campaign spends $6,000, generates 100 inquiries and produces 10 qualified leads. CPL is $60 and cost per qualified lead is $600. The plan expected 20 qualified leads at the same spend. These are teaching figures, not client results or a conversion of historical Russian spending. An attractive inquiry cost does not establish that the quality objective was met.

Halfway through the campaign, the team changed both targeting and the form copy. The overall result cannot isolate which change contributed more. First check the qualification definition, distribution of inquiries over time and lead handling. If the evidence is insufficient, the finding is a measurement limitation and a better design for the next test. One campaign with several simultaneous changes is not enough to dismiss an entire advertising platform.

5. Convert findings into verifiable actions

Monitor the agency more closely is not a usable action. A stronger entry specifies that the launch owner submits a test inquiry and attaches proof that the assigned employee received it; launch cannot proceed without that check. Each action needs an owner, deadline, acceptance criterion and the recurrence risk it addresses. Separate an immediate repair, a process change and an experiment.

Avoid generating twenty tasks from one review. Choose a small set that fixes a confirmed problem or reduces a consequential uncertainty. For a new ad test, define the comparison, primary metric and stopping rule in advance. A platform experiment may help, subject to its supported campaign types and conditions. Running a new version after an old one is not automatically a controlled experiment simply because the brief calls it a hypothesis.

6. Close the review with evidence of completion

Share the decisions and evidence links through the team's normal internal process, then set a date to check the actions. A closed task should contain proof such as a received test inquiry, a working alert, an agreed event dictionary or an approved experiment plan. If the action is incomplete, retain the reason and revised date. Quietly moving it into the next report hides the risk rather than resolving it.

The report-review tool accepts pasted text and suggests missing information and follow-up questions. It can help prepare the review, but it cannot establish why a campaign failed or replace source exports. Use a redacted report, assess its suggestions and move verified questions into the worksheet. At the next launch, reopen the previous postmortem. Its practical value is visible when the team actually uses the completed change, not merely when the document has been filed.