Meta Is Removing View Attribution Windows: Rebuild Reporting Now

Picture a Tuesday morning in mid-January. You open the weekly report. Spend looks normal. Results look a little thin. Nothing is red, no error banner, no failed refresh. So you do what most of us do. You assume the ads got worse and start editing creative.

That is the scenario I want you to avoid. According to Meta, on January 12, 2026 the Ads Insights API will stop returning the 7-day view and 28-day view attribution windows.¹ If any report you depend on asks for those windows, the thin numbers will not mean what you think they mean. The risk is a report that goes quiet, not one that breaks.

You have about twelve weeks. This piece explains exactly what is changing, who is affected, and how to rebuild in an afternoon.

What exactly is Meta changing?

Meta published the details on its developer blog on October 16, 2025. The announcement was reported by PPC Land on October 18.¹² Here is the substance, with the sources marked.

The attribution windows

Meta says the 7d_view and 28d_view values of action_attribution_windows will no longer be returned. The windows that remain are 1-day click, 7-day click, 28-day click, 1-day engaged view and 1-day view, plus the default values based on each advertiser's attribution setting.¹

The breakdown limits

Starting the same day, Meta says it will limit how far back certain breakdowns can reach:¹

• Unique-count fields such as unique_actions and cost_per_unique_action_type: all breakdowns limited to 13 months.
• Hourly breakdowns: limited to 13 months.
• Frequency breakdown (frequency_value): limited to 6 months.
• Marketing mix modeling breakdown: limited to asynchronous jobs.

Total values for API fields remain available for up to 37 months, consistent with Ads Manager.¹

Why Meta says it is doing this

Meta says it regularly reviews the metrics it offers and is limiting some windows and breakdowns to keep them meeting its standards.¹ PPC Land reports the stated aims as reducing discrepancies with Ads Manager and improving API performance.²

Will your Ads Manager numbers change?

Probably not, and this is the part that confuses people. Meta describes the change as consistent with Ads Manager behavior, which means it is bringing the API in line with what the Ads Manager interface already reports.¹

So the honest answer is: it depends on where you read your numbers.

Who is affected, and how would you know?

Anything that talks to the Insights API on your behalf. That includes:

• BI dashboards and spreadsheet connectors that refresh nightly.
• Agency or freelancer reports built on a template someone made years ago.
• Custom scripts that compute blended ROAS from several windows.
• Any marketing-mix or reporting tool that uses the Meta connection.

You may not know which of these ask for the removed windows. The person who built the report may be long gone. That is the real problem. The failure is not technical. It is a gap in who knows what the report does.

What a quiet failure looks like

PPC Land reports that requests for the deprecated windows return empty data sets rather than error messages, and that out-of-retention queries can return null or empty responses.² Meta's post does not describe the response, so this comes from PPC Land's reporting, and you should test your own setup rather than trust any summary.

If that behavior holds, a report that adds a 28-day view column would just show zeros or blanks. A year-over-year chart that reaches past the retention limit would have a hole in it. Nobody gets an alert.

How do you audit your reporting in an afternoon?

You can do this in about three to four hours. Do it on a day you are not launching anything.

What does a before-and-after look like in a real report?

Here is an invented example, only to show the logic. A report has these columns for one campaign: 7-day click purchases, 1-day view purchases, 28-day view purchases, and a "total" that adds all three. After January 12, the 28-day view column goes blank, and the total drops.

Nothing about the campaign changed. The total changed because the report was adding a window that no longer exists. Worse, the old total may have double counted the same customer in overlapping windows, so the "before" number was never as clean as it looked.

That is why I tell people to rebuild, not patch. A patch hides the gap. A rebuild forces you to decide what each column means.

Rules for the rebuilt report

• One window per column, and the window is in the column name.
• No summing across windows. Windows overlap. Pick one to headline and show others beside it for context.
• A reconciliation row. Meta-reported purchases next to purchases from your store or processor for the same dates.
• A notes cell. Put the date of the change and the new definition where anyone opening the file will see it.

Tell your team and your clients early

If other people read your numbers, warn them. A short message does the job: "Meta is changing what its reporting API returns on January 12. We are rebuilding our reports now. You may see a one-time shift in some columns, and it will not reflect ad performance." Explaining a change in advance builds trust that no dashboard can.

Which window should be your headline number?

Choose one, write it down, and do not change it for a year. A consistent window matters more than a perfect one.

I lean toward a click-based window as the headline, because clicks are the clearest record of a person choosing to arrive, and I use shorter view windows only as a secondary signal. That is a preference, not a law. The better answer for you comes from comparing each window to cash in your bank for a month or two.

Anchor to cash, not to the dashboard

Whatever window you choose, reconcile it against your payment processor, your store or your CRM. If Meta says 100 purchases and your store says 62, you do not have an attribution problem to solve with a different window. You have a measurement problem to understand.

Jon Loomer explained the mismatch in a video posted on April 2, 2025:

Consider an incrementality lens

Meta has been rolling out an incremental attribution setting, and Seer Interactive tested it on about $1.05 million of spend in April 2025. Meta's own reporting classified 87% of conversions as incremental, against 67% under GA4. Seer cautions that Meta's self-reported incrementality is directional and should be cross-checked against independent sources.³ I would test it beside your normal reporting, not instead of it (why incrementality matters).

Where does signal quality fit in?

Windows decide how long Meta credits an ad after someone sees or clicks it. They do not decide whether Meta learned from a real purchase in the first place. That is the layer under the layer.

If your Purchase events arrive twice, arrive with no value, or arrive only when the browser cooperates, changing windows will not fix the picture. A basic hygiene check is that browser and server events share an event name and an event ID, so Meta can discard the duplicate. Meta's documentation says it discards the later event when the same pair arrives within 48 hours.⁴ The Conversions API setup basics cover the rest.

Window changes are a good prompt to check this, because you are already looking at your reporting end to end.

What should the next twelve weeks look like?

Here is a simple schedule.

• This week: inventory every report and owner.
• By mid November: rebuild on supported windows and relabel columns.
• By early December: export any history you might need, and write your measurement definition.
• Early January: freeze report changes before the cutoff and note benchmark numbers for the week before January 12.
• January 13: compare, annotate and tell your team what you expect to see.

If you work with an agency or freelancer, send them this question today: "Which attribution windows do our reports request, and have you checked them against the January 12 change?" Their answer will tell you a lot.

Platforms change what they show you. They will do it again. The operators who never get ambushed have done two things: they own a definition of success that is written down, and they reconcile their ad numbers against money they can count.

You have twelve weeks. An afternoon of work now beats a month of confusion in January. This kind of reporting hygiene is built into the systems I build for clients.

Keep building.Ekai