Skip to main content
Get this template as a Google Doc: make a copy.
Run this after launch, not instead of the campaign launch checklist. A campaign can pass every pre-launch check and still be broken in ways only live traffic reveals.
Three passes with different jobs. The first hour catches things that are outright broken. The first 48 hours catches things that are working but wrong. The weekly pass catches drift.
Every pass can be run by a person at the dashboard or as a scheduled pass by an agent against the platform APIs, analytics, and the flat file: launch arms the three schedules (+1 hour, +48 hours, weekly). Either way, a failed check goes to the owner with the observed value against the expected one; rows tagged (human eyes) need a person’s read. Deciding what to do about a failure is the owner’s call.

First hour

  • Spend and impressions are registering on every platform. Zero after an hour usually means a rejected ad or an unmet audience condition, not a slow start.
  • No ads are in a rejected or limited state.
  • A real click on a live ad reaches the destination page. A scripted request to the final tagged URL, verified in analytics, is an acceptable equivalent; the one thing it misses is the ad rendering itself.
  • The conversion event fires on that click, verified in the analytics tool rather than in the platform’s own reporting.
  • The utm_campaign value arriving in analytics matches the campaign name exactly. Check the value that actually lands, not the one you intended.
If the UTM value arriving in analytics differs in case or separator from the campaign name, stop and fix it now. Every hour it runs adds rows that will not join, and those rows cannot be repaired retrospectively.

First 48 hours

  • The campaign appears as one row in the flat file, not two. A split row means the naming and UTM join is broken.
  • Conversion counts in the platform and in analytics are in the same order of magnitude. They will not match exactly; they should not differ by a factor of ten.
  • Spend is pacing against the flight, not front-loading the whole budget into two days.
  • Every ad in the campaign has delivered something. An ad on zero impressions after two days is usually a placement or aspect ratio problem.
  • The destination page is holding up: no console errors, no broken form, no slow load on mobile. The checks are scriptable; whether the page still reads well is (human eyes).
  • If a test is running, the experiment ID is present on the rows arriving in the flat file.

Weekly, while it runs

  • Names have not been edited in any platform. A renamed campaign splits its own history.
  • No new ad has been added outside the naming convention. Ads added mid-flight are where conventions usually break.
  • Budget changes have been recorded in the change log, so next week’s report can explain the movement. A change an agent executed through a platform API is logged there at execution time by construction; a manual change still needs the entry.
  • Creative is not being swapped without a brief. A new asset needs a creative brief even mid-campaign.
  • Any claim added to new copy is still on the approved list in messaging.
  • Guardrail metrics are checked against their normal ranges, not just the primary one. The check is mechanical; what a drift means is (human eyes).

Before you call anything a result

  • The campaign has run long enough for the conversion window to close.
  • The numbers come from the flat file, not from a platform dashboard.
  • The metric being reported matches its definition in the data dictionary.
  • Anything surprising has been checked against data quality and rollup concepts before it gets reported as performance.
Type: checklist
Domains: campaign, tracking, reporting
Provides: checklist-qa
Uses: approved-claims, budget-plan, creative-brief, data-dictionary, flat-file, library:change-log, naming-convention, utm-rules
Hands off to: process:budget-pacing, process:campaign-retrospective
Asks, in order:
  1. Which campaign, and which pass? Options: qa-pass. Skip when: scheduled from launch.
Output: Pass results with observed vs expected per failed row, human-eyes rows flagged for the owner; schedules armed at launch (+1h, +48h, weekly).

Get the complete Marketing OS

Subscribe to Snow Doughnut, our free weekly newsletter, and get the Notion workspace where you do the work: tasks, a marketing calendar, name generators, briefs, and more.

Have us run it for you

We implement this system, use it to help you grow and scale, then hand it all off to you when you’re ready.
Last modified on November 2, 2026