Meta’s on-device conversion setting starts November 12. Check it before then

Meta’s “Measure conversions on customer devices” setting starts November 12, 2026, and it is on by default for eligible pixels. If you do nothing, your pixels start using it. The setting changes which purchases Meta can connect to an ad. The attribution rules stay the same, so the counts can move even when performance has not changed.

Here’s what’s happening with Meta’s on-device conversion setting

Meta’s help article describes the problem the feature targets. A shopper sees an ad, returns later, and buys on the website. Meta may hold the ad view and receive the purchase event, but it can’t always connect the two, so the campaign doesn’t get credit. The new setting, “Measure conversions on customer devices,” moves that match onto the shopper’s device, where Meta says both the ad view and the purchase are present.

According to Meta, the match uses information the advertiser already shares, such as an email address, together with temporary, non-identifying device information. Meta says it receives only confirmation that a purchase happened and which campaign should get credit.

Two statements in Meta’s materials matter most for measurement:

• Meta says the feature is “turned on by default.” The opt-out sits under Events Manager, then Data sources, then your pixel, then Settings, in the toggle labeled “Measure conversions on customer devices.”
• Meta says your “attribution window and attribution model don’t change.”

Those two statements together describe the risk. The default decides whether the feature runs, and the unchanged attribution rules mean the reporting change will look like ordinary performance movement unless someone labels it. Meta’s advertiser email gives November 12, 2026 as the start date. The help article itself does not repeat the date.

Some limits are also confirmed by Meta. The feature is not active for people located in the EU or the UK. The Pixel must be installed directly on the page, not inside an iframe or as an image-only pixel. The website must require people to log in with an email address or mobile phone number.

Here’s what you need to get ready for

Counts will move, and a rise in reported conversions does not by itself show that performance improved. Our inference is that CPA and ROAS will shift mechanically with the conversion count, even when spend and sales are unchanged. Meta does not say this, but it follows from the attribution rules staying the same.

Without a baseline of Meta conversions from before November 12, you cannot tell how much of any later change came from the setting. Capture these before the switch:

• Daily conversions by action type and attribution window for each client, exported from Ads Manager. Keep the window and date basis in the export.
• Events Manager event volume and match quality for each conversion event, saved as a dated snapshot.
• Platform-to-backend ratio: Meta conversions divided by orders or qualified leads from your CRM, Shopify or backend, computed daily. This is the check that shows whether a rise is real.
• Annotated dashboards: mark November 12 on every client chart that uses Meta conversions, so later movement is not read as the result of anything else you changed that month.
• Pixel placement audit: list every page where the Pixel runs today, including any login or account pages, so you know where the setting could apply. This is an inventory, not a reason to add the Pixel anywhere new. For healthcare and other regulated clients, a conservative design keeps third-party ad pixels out of portals and authenticated areas; see our guide to healthcare tracking without PHI.

Save the baseline outside the dashboards you will later compare against. A dashboard that has been edited after the change cannot prove what it showed before.

Capture email early, and change one thing at a time

Meta says this setting uses information shoppers have already shared with you, such as an email address, and that it only applies on sites where people log in with an email or phone number. Meta has not said whether the email you send through Advanced Matching or the Conversions API feeds the on-device match, so treat those as separate questions.

Sending a hashed email with your events is Meta’s documented way to improve matching in general, whatever this setting does. Whether it also helps the on-device match is our inference, not a Meta statement. Decide on it as a matching and privacy choice, not as a requirement of the November change.

• Capture email before the purchase: newsletter signups, account creation, gated downloads and quote or lead forms are all earlier points where a shopper can share it.
• Send only what you would send anyway: the Pixel’s Advanced Matching hashes the email in the browser before it is sent, and the Conversions API accepts a hashed email or phone number with each server event. Don’t add identifiers on the assumption that the on-device setting needs them.
• Normalize it the same way everywhere: trim and lowercase the email before hashing, so the Pixel and the Conversions API send the same value for the same person.
• Don’t change matching during the rollout: if you change what you send close to November 12, a before-and-after comparison can’t separate your change from Meta’s. Make identifier changes well before your baseline window or after you have read the rollout’s effect, or keep a comparable set of events unchanged as a control.
• Regulated clients decide first: for banks and healthcare, counsel should decide which events, if any, carry customer data to Meta. The healthcare tracking guide covers the conservative design.

Here’s what you can do

Start by deciding the setting on purpose. To turn off Measure conversions on customer devices, open Events Manager, go to Data sources, select the pixel, and open Settings. Check the toggle for each eligible pixel now, record the state, and tell the client what it is. A default you inherited is still a decision.

For most advertisers, the next step is a controlled read after the switch. Monitor the ratio of platform conversions to backend orders for four to six weeks and watch for duplicate spikes. Then run a geo or campaign holdout before crediting any lift to Meta. A holdout measures incremental orders in your own backend, so it doesn’t depend on how Meta counts conversions. That makes it the most direct check in this list for telling a real change from a counting change.

For banks, credit unions, healthcare providers and other regulated clients, the setting is a legal decision as well as a reporting one. Meta’s materials do not say whether the feature honors consent signals such as fbq('consent', 'revoke'), a consent management platform, or Global Privacy Control. Those are questions for counsel and the client’s privacy team. Until counsel has reviewed the feature, our recommendation is to opt out and document the reason. The requirement that the site use email or phone login is also worth reviewing, because it describes the authenticated pages where the sensitive content usually sits.

Finally, ask Meta in writing. Its public documents do not answer four questions:

1. How a recovered purchase is deduplicated against a Conversions API event for the same order.
2. Whether recovered conversions feed delivery, bidding, audiences or value optimization.
3. How consent signals affect matching.
4. Whether recovered conversions carry a separate label or column in Ads Manager and the API.

Until those are answered, treat any reported lift as provisional. For the broader reasons Meta and GA4 disagree, read why Meta Ads and GA4 numbers don’t match. If your conversions come from backend events, the offline conversions guide covers how those events reach ad platforms, and the platform mismatch guide sets out the shared rules to write down.

Open questions

These are not answered in Meta’s public materials as we read them:

• The exact device-side mechanism, including which component holds the ad-interaction records and how long the device token lasts.
• Whether the setting interacts with Advanced Matching in a way that changes how the email is normalized.
• Whether reported figures in Events Manager and Ads Manager diverge after the change.

We have not verified these. Revisit this article when Meta publishes answers.