How do you measure MFA exposure after a campaign?
Part of MFA avoidance
The short answer
An MFA exposure analysis joins delivered impressions to a dated list of domains classified as made for advertising, then divides media cost on those domains by total media cost in scope. Report masked and unclassified delivery apart, and compare the result with the prelaunch screen to see whether a gap came from the list, the setup or the sites.
Key takeaways
- Exposure is measured on delivery, the impressions that actually served, not on the list that was planned.
- The share of spend needs one cost definition and one currency in both the numerator and the denominator, with the scope written down.
- Masked URLs and unclassified domains are a separate bucket, never counted as clean and never counted as MFA.
- Compare each served domain with the classification that existed at screening time and with today's, because the two answer different questions.
- The output is a short list of changes: list entries, platform settings, deal terms and the screening rules themselves.
Which data shows MFA exposure after a campaign?
Exposure analysis starts from what served. A prelaunch screen says which sites were allowed; delivery data says where the money went. The MFA avoidance hub puts a monthly MFA share of spend on its review calendar, and this page is the method behind that number. The data source sets the ceiling on what the analysis can say, so choose it before choosing the classification list.
| Source | What it gives | Limit to plan for |
|---|---|---|
| Display & Video 360 Data Transfer files | Impression level rows with event time, line item, exchange, the bid request URL and media cost | Files stay in Cloud Storage for 60 days and are then deleted; some exchanges mask the URL |
| Display & Video 360 reports | Delivery grouped by App/URL, exchange, environment, country and inventory source, with media cost | Grouped by site or app, so page level questions need the log files |
| Google Ads Where ads showed report | Websites, pages or apps where Display Network ads showed | Google says not to use it as an exhaustive list or as an accurate representation of billing |
| Exchange or deal reports | Delivery inside a private marketplace or curated deal | Fields and naming vary by seller; map them to the same domain format first |
Google's Data Transfer overview (opens in a new tab) describes the files as "raw reporting data in log level format" and states that "After 60 days files are deleted." For a flight longer than two months, copy the files out as they arrive. The Google Ads Where ads showed report (opens in a new tab) adds two cautions: placements shown "only a couple of times may not be included," and invalid traffic "may not be removed" from it. Matching rules for turning raw URLs into domains are covered in our guide to served domain reconciliation.
How do I compare an MFA screen with impression level exposure later?
Keep the classification version that existed on the screening date, and run each served domain against it and against the current version. The first comparison tests the screen and the setup: did delivery follow the list the planner approved? The second tests the screen's foresight: did sites that passed later turn into MFA sites? Mixing the two blames the planner for changes that happened after approval.
Steps to compare the screen with delivery
- Freeze the inputs. Save the approved list version, the MFA classification list with its date, and the platform change history for the flight.
- Normalize served URLs to domains. Use the same rules used to build the prelaunch list, so a subdomain does not appear as a new site.
- Join on the event date. Match each impression row to the list and settings in force when it served, using the event time field.
- Sort into the matrix. Place each served domain in one quadrant, with its media cost.
- Trace each leak. For domains that were screened out but still served, find the line item, inventory source and setting that let them through.
Why can exposure differ from the prelaunch screen?
Some differences come from the buying setup, some from the sites and some from the measurement. Display & Video 360's brand suitability help (opens in a new tab) lets advertisers "bypass all advertiser-level brand suitability settings for specific deals," so a deal line can serve where the open auction line could not. Its targeting help (opens in a new tab) notes that "a few hours may pass" before a targeting update reaches bidding on an active line item.
| Cause | Evidence in the data | Fix |
|---|---|---|
| Deal exempt from suitability settings | MFA rows concentrate in one inventory source | Review the deal's domain list and its bypass setting |
| Exclusion list applied at the wrong level | MFA rows come from one campaign that lacks the shared list | Apply the list at the account or advertiser level |
| Late change | MFA rows stop a few hours after an exclusion was added | Note the lag in the report; add exclusions before launch |
| Site changed after screening | Domain sits in the turned MFA later quadrant | Recheck sites on the plan during long flights |
| Broad targeting beside the list | MFA rows come from lines not limited to the inclusion list | Limit those lines or add the MFA exclusion list |
In Google Ads, account level placement exclusions (opens in a new tab) override campaign and ad group settings, and new ones "typically take effect within 12 hours." Check those dates against the first and last MFA impression before calling a row a leak.
What should change after an MFA exposure analysis?
Actions an exposure analysis should produce
- Domains in the leak, still MFA quadrant are traced to a line item and inventory source, and the path is closed.
- Domains that turned MFA after screening are added to the exclusion list with the evidence and the date.
- Deals carrying MFA delivery have their domain lists reviewed before the next flight.
- The recheck interval for long flights is set from the share of sites that changed during this one.
- The unclassified bucket is reviewed by hand for its largest domains by media cost.
- The method, scope, cost definition and list versions are saved with the result for the next comparison.
Report the result to the client as a scope, a share and a list of changes, not as a verdict on the campaign. The ANA benchmark page explains how the industry's participant figures are built; a single campaign's share is a different population and should be read on its own terms.
What does an exposure analysis look like on a real flight?
Fictional example
A modular home builder in New Jersey
A fictional modular home builder serving Morristown, Toms River and Cherry Hill runs a ten week English display flight in Display & Video 360 beside home building, real estate and local news content. The agency copies the Data Transfer files out each week so none are lost to the 60 day deletion.
After the flight, the analyst joins the impressions to the screening date MFA list and to the current one. Two home improvement listicle domains sit in the leak, still MFA quadrant. Both trace to one private marketplace line where the deal bypassed advertiser level suitability settings and carried the seller's own package of sites.
A real estate tips site that passed screening now appears on the current MFA list after a template change in week six, so it lands in the turned MFA later quadrant. Masked URL rows from one exchange go in the unclassified bucket. The agency asks the seller for the deal's domain list, adds the real estate site to the exclusion list with the date of the change, and shortens the recheck interval for flights longer than eight weeks.
How does PlacementGraph give an exposure analysis a dated list to compare against?
Frequently asked questions
How soon after a campaign ends should the exposure analysis run?
Before the delivery data expires. Display & Video 360 deletes Data Transfer files after 60 days, so either run the analysis within that window or copy the files to storage the agency controls while the flight is live. The classification list should also be saved on the day of the analysis, since it changes.
Can an exposure analysis be run without log level data?
Yes, at the site level. Grouped reports by App/URL with media cost support a domain level MFA share. They cannot show which pages served, and Google says its Where ads showed report is not an exhaustive placement list or a billing record, so state the source and its limits beside the result.
Does MFA exposure analysis apply to YouTube delivery?
Only in part. MFA classifications describe websites, and Google reports ads outside the YouTube Watch page, such as the Home feed, as youtube.com in the Where ads showed report. Measure YouTube channel quality with a channel review rather than folding it into a website MFA share.
Who should own the exposure analysis, the agency or the brand?
The party with access to the delivery data runs it, and the agency scope should say which party that is. The brand approves the method and the scope so that the figure means the same thing each quarter. Whoever runs it saves the inputs, so the other side can reproduce the result.
Sources
- Data Transfer v2.0 overview (opens in a new tab)Google, Display & Video 360 developer documentation, updated October 14, 2024
- Data Transfer file format (opens in a new tab)Google, Display & Video 360 developer documentation, updated September 14, 2026
- Metrics in reports (opens in a new tab)Display & Video 360 Help
- Dimensions in reports (opens in a new tab)Display & Video 360 Help
- About the “Where ads showed” report (opens in a new tab)Google Ads Help
- Brand suitability (opens in a new tab)Display & Video 360 Help
- App, channel, and website targeting (opens in a new tab)Display & Video 360 Help
- Exclude placements at the account level (opens in a new tab)Google Ads Help
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.