When to run this checklist
Run it before changing which conversions an ad platform bids or optimises towards, their values or counting, or the attribution model and windows, in Google Ads, Meta or any other platform. That includes promoting a new Conversion ActionA Google Ads definition of one action to count, such as a purchase or lead, with its source, category, counting method, value and conversion window. to primary, switching a Meta ad set's Optimisation EventThe conversion event an ad set or campaign is told to optimise delivery towards, such as a purchase, lead or install., moving from form submissions to qualified leads imported from a CRM, changing a Lookback WindowThe time after an ad interaction in which a conversion can still be credited to it., and the clean-up of an inherited account. Run it once per platform, because the list deliberately stages platforms one after another.
The paid media lead who will make the change runs the list, with the analyst who owns the tracking and the reports. Several neighbouring pages take over around it. Before the new event carries any bidding, the tracking verification checklist proves that its tag fires and counts once. After the change, the monthly paid media review checklist is where the result is read in the normal rhythm. A full clean-up of an inherited Google Ads account follows the conversion action clean-up tutorial, an inherited Campaign Manager 360 advertiser follows the tutorial on cleaning up an inherited Floodlight configuration, and the choice of strategy itself belongs to the Smart Bidding strategies guide. Tracking releases and re-baselines across the business are governed by the measurement governance playbook; this list is the per-change gate inside it.
The platform statements on this page come from those guides and the help pages they cite. Google lists changes to conversion actions among the changes that can put a bid strategy into a learning period; Meta lists changing the optimisation goal or conversion event among the significant edits that restart an ad set's Learning PhaseThe period after launch or a significant edit when an ad set's delivery is still being optimised.; and a change to a window or model breaks the trend line at the date it is made. How long learning lasts, and the screens where each setting lives, are matters for each platform's help centre, and the list does not state them.
Before you start
- The written definition of the result the new goal should count, in the business's own words.
- The conversion action or event register, with each action's source, primary or secondary status, counting, value, windows and attribution model.
- The list of campaigns, ad sets and bid strategies that use the action or event, including portfolio strategies and shared goals.
- The trading calendar, with sales, holidays, launches and seasonal peaks marked.
- The business's own records of orders, leads or qualified enquiries for a recent closed period.
- The measurement change log, open at a new entry.
- Current targets and budgets, with the period they were set against; the CPA calculator helps recompute a target once the new basis is known.
- A judgement of how reversible the change is. The one-way and two-way doors framework separates a setting that can be put back cheaply from one whose learning and history cannot.
The checklist
The groups follow the order of the change: define it and see where it lands, choose when it happens, prove the new event before it carries bidding, then record the change and judge it on the right basis. A line in the first three groups that cannot be ticked stops the change at that point.
The change
Timing
Before switching
Record and judge
Where each check is explained
The widget keeps the lines short so they can be ticked; the reasoning lives in the guides. Each check links to the page that explains it.
| Check | Explained in |
|---|---|
| Definition of the result written | Conversion events across ad platforms |
| Every campaign and bid strategy using the action listed | Cleaning up Google Ads conversion actions |
| One platform and one setting at a time | Conversion events across ad platforms |
| Targets and budgets left alone that week | Conversion windows and attribution in Google Ads |
| Scheduled outside sales, holidays and peaks | Smart Bidding strategies |
| Learning period expected and readers told | Smart Bidding strategies |
| Primary action changes applied together | Cleaning up Google Ads conversion actions |
| No bid strategy switch at the same time | Results dropped after a bid strategy change |
| New event observed or secondary first | Conversion events across ad platforms |
| New event confirmed against business records | Reconciling Google Ads conversions with sales |
| Event fires and deduplicates | The Conversions API and deduplication |
| Meta significant edits noted | The Meta learning phase |
| Old and new settings, date and reason logged | Keeping a measurement change log |
| Change date marked as a break in trend | Conversion windows and attribution in Google Ads |
| Judged on cost per real result after late conversions | Conversion events across ad platforms |
| Targets reset once the window has closed | Conversion windows and attribution in Google Ads |
| Attribution windows compared in reports first | Meta attribution settings |
If an item fails
The definition, the timing and the proof of the new event are the hard stops. A change with no written definition of the result has nothing to be checked against, so it waits until the definition exists and the analyst agrees it. A change scheduled into a sale, a holiday or a seasonal peak moves to a quiet period: the guide to sales and tracking outages explains why changes to conversion actions are held back while the model is already absorbing an unusual period, with the exception of a fix that an outage itself requires. A new event that has not been observed, or that the business's records show counting the wrong thing, never becomes the goal; correct the tag or the import and observe it again. The academy checkpoint on changing bidding goals and the measures behind them tests observing a new event before it steers bids alongside what a metric redefinition or a replaced north star adds: naming the new measure, finding everything that reads it, testing a replacement on cohort evidence and leaving live tests on their own rules.
The scope lines decide how far the change reaches. If the list of campaigns and strategies that use the action is incomplete, a portfolio strategy or a shared goal will pick up the change unannounced, so finish the list first. If a bid strategy switch or a target change is planned for the same week, separate them; the diagnosis of results dropping after a bid strategy change shows how two changes at once turn a readable result into an argument.
The record and judge group can be finished on the day of the change, but not skipped. If the change log entry or the break in the reports is missing, whoever reads the next month's figures will treat a measurement change as a change in the business, which is exactly the silent re-baseline the measurement governance playbook warns against. Targets can stay where they are until the new window has closed, with a note that they were set on the old basis; resetting them early, from figures that still lack late conversions, sets them against an incomplete count.
A solar installer switches its bidding goal (illustrative)
From form fills to qualified surveys, one platform at a time
Plethane Solar, an invented installer of solar panels and home batteries, bids on Google and Meta towards website form submissions. Its sales team finds that about four in ten forms come from renters or homes with unsuitable roofs, so the head of marketing decides to bid towards qualified site surveys recorded in the CRM instead. The paid media lead runs this list with the analyst in January, well before the spring rush that usually starts in March.
The change group passes once the analyst writes the definition: a survey booked after a call confirms the caller owns the home and the roof is suitable. The list of campaigns turns up a portfolio strategy shared by three search campaigns, which would otherwise have been missed. In the before switching group, the qualified survey import is added to Google as a secondary action and observed for five weeks. Comparing it with the CRM, the analyst finds that surveys later cancelled are still counted, and the import is corrected before it carries any bidding.
Google is changed first, in mid-February, with targets and budgets held for the week. The old target CPA of £60 per form becomes meaningless, and after the window closes the team sets a new target of £180 per qualified survey from the closed figures. Meta follows three weeks later, with the optimisation event switched on the main ad sets in a single edit, as the significant edits list warns. Both dates go into the measurement change log and the dashboards carry a break line. By April the monthly report shows fewer conversions, a higher cost per conversion and more installations booked per pound spent, and nobody mistakes the first two for a decline.
Common mistakes
- Changing every platform at once. The blended result moves and nobody can say which change moved it.
- Promoting the new event straight to primary. A tag that has not been observed beside the business's records can teach the bidding to chase the wrong thing for weeks.
- Changing during a peak. A change made in a sale teaches the system that the sale is normal and makes the period impossible to read afterwards.
- Changing in dribs. Small changes to primary actions spread over several weeks put the bidding through learning again and again.
- Bundling a target or strategy change. Two causes behind one shift leave the review guessing.
- Missing a shared strategy or goal. A portfolio or account-level goal picks up the change in campaigns nobody was watching.
- No break in the reports. A lower conversion count after a stricter definition is read as a collapse in performance rather than more honest measurement.
- Judging too soon. Results read before late conversions arrive, and targets reset from them, measure the window rather than the change.
The checklists library holds the lists that sit around this one, from tracking verification before the change to the monthly paid media review after it.
