Meta will begin measuring some website purchases on shoppers' own phones and computers from November 12, 2026, through a new Meta Pixel setting that is already switched on for advertisers who take no action. The setting, labelled "Measure conversions on customer devices," is designed to recover sales the company says never reach its campaign reporting today. Yiqi Wu, chief executive of Aimerce, drew wider attention to the change in a post on X on October 26, writing that Meta "will start counting more conversions starting November 12, 2026."

In Short

Meta is changing how its website tracking code checks whether an ad led to a purchase, so that part of the check happens on the shopper's own phone or computer instead of only on Meta's servers. It affects businesses that use the Meta Pixel, because the setting switches on by itself from November 12 and may add sales to Meta's reports that were missing before. The numbers in Ads Manager could go up without anything changing in the shop, which is why some practitioners suggest comparing Meta's figures with real orders.

The setting in Events Manager

The new control appears in the Meta Pixel settings panel of Events Manager, where it sits directly beneath an older toggle called "Automatically include more detailed page and product info." Both are shown as on. Under the new toggle, Meta adds a line in grey: "Measurement starting November 12, 2026."

Meta's own description of the feature in that panel is two sentences long: "Meta measures conversions from your ads so you can see which ads are working. With this feature on, that measurement happens on your customer's device instead of exclusively on Meta servers."

The final phrase carries more weight than its length suggests. "Instead of exclusively" does not say that measurement moves off Meta's servers; it says the servers will no longer be the only place it happens. Read literally, the setting adds a second, device-side path alongside the existing server-side one rather than replacing it.

The toggle above it belongs to an earlier change. In April, Meta introduced an AI feature for the Pixel that adds page and product information to the events sent to Meta. The panel describes it as a feature that "uses AI to identify and send more website details like basic page info, reviews and pricing in addition to your existing Meta Pixel data." That rollout gave existing Pixel users a notification and a 30-day window to review the feature before it turned on, and it did not reactivate once disabled.

What Meta told advertisers

Meta's notice about the November change reached advertisers in mid-October. German e-commerce blog Wortfilter reported it on October 18, describing an email from Meta's business team whose subject line, in its German version, promised advertisers a view of more of the conversions they already generate. Philippine news site ASTIG covered the same notice that day.

According to the notice as those outlets described it, the feature recovers conversions that are missing from campaign reporting because they never make it back to Meta to be measured. Those conversions are measured on the customer's device instead, using information the customer has already shared with the business, such as an email address, together with temporary, non-identifying device data. Both stay on the device. Meta learns only that a purchase took place and which campaign should get credit for it.

No code changes or integration updates are needed, according to the notice. The feature applies to datasets that already send events - Meta's term in Events Manager for Pixel and Conversions API connections - and advertisers can turn it off at any time in the settings of each dataset. Wortfilter wrote that Meta plans to roll it out gradually from November 12 and that there is no opt-in step.

MetricMaven, which publishes guides on ad measurement, went further in a piece by Jamie Isabel on October 19. Citing two Meta Business Help Center articles, it listed limits it described as 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; and the website must require people to log in with an email address or mobile phone number. MetricMaven also wrote that, according to Meta, the attribution window and attribution model do not change, and that the toggle sits under Events Manager, Data sources, the relevant pixel and then Settings.

None of those conditions appears in the settings panel text or in the notice as Wortfilter and ASTIG reproduced it. The login condition in particular, taken literally, would leave out shops where a purchase can be completed without an account.

How the Pixel matches purchases now

The Meta Pixel is JavaScript that runs in the shopper's browser. When someone loads a checkout confirmation page, the Pixel sends a purchase event to Meta's servers, where Meta tries to connect it to a Facebook or Instagram account that saw or clicked an ad. It does this with first-party cookies, click identifiers carried in ad links and, where advertisers enable advanced matching, contact details such as email addresses and phone numbers hashed with SHA-256 before they leave the browser. The Conversions API sends the same kind of event from the advertiser's server, which avoids some browser losses, and Meta deduplicates the two streams.

Events go missing at several points in that chain. Browsers restrict cookies, strip click identifiers or block requests to tracking domains; ad blockers stop the Pixel from loading; consent banners pause it. Meta's notice describes the recovered conversions as ones that never make it back to Meta to be measured, without saying which of those causes the feature addresses.

The change described in the notice inverts part of the process. Instead of sending identifiers to Meta and matching them on Meta's servers, the matching inputs stay on the device and only the outcome - a purchase occurred, a campaign gets credit - is passed on.

That raises technical questions the published material does not answer. For a match to happen on the device, the device needs some record of which ads its user saw or clicked; Meta has not said where that record comes from. Meta's wording also leaves room for two readings of what is being recovered: purchases whose events never reach Meta at all, or events that reach Meta but cannot be tied to an ad. The notice does not say which, and the answer determines what kind of purchases the feature can add. And the notice says Meta learns only that a purchase happened. It does not say whether purchase value and currency travel with it, which decides whether recovered conversions show up in return on ad spend reports and value-based bidding or only in conversion counts.

Wortfilter listed three further gaps: how attribution works without a click identifier or cookie, which browsers and operating systems are supported, and whether the measurement runs in Meta's apps, in the browser, or in both.

What Wu wrote

Wu's post opened bluntly: "Meta's update in the worst possible moment! Meta will start counting more conversions starting November 12, 2026." Wu's X profile describes the company as helping "Shopify brands build Meta Pixels on steroids in the cookieless world" and lists ten years of engineering work at Meta, Instagram and Reddit. Because Aimerce's stated business is Pixel tracking for Shopify merchants, its chief executive has a commercial interest in how merchants respond to changes in Meta's measurement.

The post's description matches Meta's notice closely. "Meta is rolling out a Pixel feature that recovers conversions it currently can't see," Wu wrote. "The purchase is matched on the customer's device, using the email they gave you plus temporary device data. Meta is only told that a purchase happened and which campaign gets credit." It added: "It turns on automatically for any dataset already sending events however you can opt out in Events Manager."

The post does not link to a Meta announcement or help page.

The post then asked "Should you opt out?" and answered: "I wouldn't opt out by default." It then set out three steps: "Export Meta-attributed purchases next to Shopify paid orders, by day, for the last month," "Track the ratio of Meta-attributed sales to actual orders" and "Keep your server-side tracking/CAPI active."

The reasoning was conditional. "If the recovered conversions are accurate, purchases your ads drove are finally getting counted. If Meta also uses them for optimization, delivery gets better signal," Wu wrote. The case for switching the feature off, in the post's terms, is a reconciliation failure: "You would want to opt out there if your numbers stop matching the orders and you can't figure out why."

Wu did not explain the phrase "worst possible moment." The calendar offers one reading. November 12 falls 15 days before Black Friday on November 27, so the change in counting lands as many retailers are setting holiday budgets and benchmarks. Replies under the post split along familiar lines. "Sounds like reconcile nightmare. I'm with you on this," wrote Dan Trapp; another user replied: "Why is this a bad thing? Don't understand."

Questions left open

Privacy compliance firm Captain Compliance, in an October 18 article by Maria Sanzio that rests on the same notice, raised four questions: whether visitors who declined consent or sent opt-out signals are excluded, what the on-device code reads, which identifier links a conversion to a campaign, and whether businesses in sensitive sectors are treated differently.

The optimisation question is the one Wu left conditional. MetricMaven also lists it as unresolved: whether recovered conversions feed delivery, bidding, audiences or value optimisation, or only reporting. That piece recommends recording a baseline before November 12, tracking the ratio of platform conversions to backend orders, and running a holdout test before crediting any lift to Meta.

The distinction matters because more counted conversions and more sales are different things. A recovered conversion is a purchase that already happened and that Meta's reporting now attributes to an ad. Whether the ad caused the purchase is a separate matter, and the feature, as described, does not address it.

The timing

The change arrives in a crowded month for Meta defaults. On October 28, Meta will switch on in-app product browsing in ads that are already running, after a notice that circulated around October 14 and gave advertisers about two weeks. Meta justified that change with a 3% average increase in conversions from an experiment across about 117,700 advertisers, without defining the conversion metric. The on-device measurement notice was first reported on October 18, about 25 days before November 12. On November 20, Meta's revised Commercial Terms and Self-Serve Ad Terms take effect, after a delay from October 30 that followed complaints from publishers and advertisers to EU and UK competition authorities.

Peak-season guidance gives the counting change a practical edge. In a post on October 21, Meta said ad sets need roughly fifty conversions a week to learn, and that ad sets below that level ran about 28% more expensive per purchase in its own analysis, which it published without a sample size or method. If recovered conversions count as optimisation events, some ad sets could cross that threshold with no change in actual sales. Meta has not said whether they do.

The money at stake is considerable. Meta reported advertising revenue of $63.06 billion for the third quarter of 2026, up 26% from a year earlier, with ad impressions up 10% and the average price per ad up 14%.

Meta's long argument about undercounting

Meta has argued for years that standard reporting understates what its ads do. In its "Suite of Truth" white paper, based on 307 studies across 54 advertisers, the company said rules-based attribution undervalues Meta by 31% at the median. Independent work points in both directions: Karan Dhir, a marketing measurement product lead at Genentech, drew on two peer-reviewed papers built on Facebook experiments to argue that observational methods frequently miss experimental results by a factor of three or more, sometimes overestimating and sometimes underestimating.

Meta's own reporting rules have moved both ways in the past year. On January 12, 2026, the Ads Insights API stopped returning 7-day and 28-day view-through windows. In March, Meta limited click-through attribution to link clicks for website and in-store conversions, moving likes, shares and saves into a renamed engage-through category, with the stated aim of narrowing the gap with Google Analytics. The on-device setting, by MetricMaven's account of Meta's help pages, leaves windows and models alone and changes which conversions Meta can see.

The Conversions API has been Meta's main answer to signal loss. In April, Meta launched a free one-click version that mirrors every Pixel event from a Meta-managed server, and the company says advertisers using it alongside the Pixel see an average 17.8% lower cost per result. Wu's advice to keep server-side tracking active fits that pattern. If the MetricMaven description of the eligibility rules is accurate, however, the on-device feature depends on the browser Pixel being installed directly on the page, which suggests server-only setups would not benefit.

Browsers, Google and earlier on-device work

Browser makers have been closing the routes that email-based matching relies on. On October 19, WebKit said Safari 27 will block identity resolution providers that turn email addresses entered at login or newsletter sign-up into persistent tracking IDs, at the domain level, on iOS, iPadOS, macOS and visionOS 27. That block follows nearly a decade of Intelligent Tracking Prevention restrictions on cookies and cross-site storage. Meta's notice says the matching inputs on the device include the customer's email address. It does not say whether the feature works in Safari.

Google has tried similar ideas. In August 2025, Google added on-device conversion measurement to iOS app campaigns, processing de-identified app event data locally, and excluded the European Economic Area, the United Kingdom and Switzerland. The EU and UK carve-out MetricMaven describes for Meta's feature covers similar ground. In June 2026, Google began testing a beta that lets website tags draw on CRM data to recover conversions the browser tag missed, matched by transaction ID and with a 14-day trial in which the new data appears in reports but does not affect bidding. Meta's notice describes no comparable trial period.

Meta's own interest in on-device processing goes back further. In August 2021, Facebook described on-device learning and Private Lift Measurement, the latter built on secure multi-party computation, as responses to new privacy measures from Apple and Google.

The Pixel has been at the centre of privacy litigation on both sides of the Atlantic. On February 3, 2026, the Dresden Higher Regional Court awarded four users €1,500 each over Meta's Business Tools, a category that includes the Pixel and the Conversions API, and barred Meta from collecting data about them on third-party sites and apps. In August 2025, a federal jury in San Francisco found Meta liable under the California Invasion of Privacy Act over health data collected from the Flo app through Meta's software development kit. On October 8, 2026, the Washington Supreme Court affirmed the dismissal of a Pixel claim against Seattle Children's Hospital, holding that clicks and searches answered by an automated server are not communications between individuals under the state's Privacy Act.

The most recent episode involving the Pixel and a device's internals came in June 2025, when researchers found that the Pixel script on Android browsers sent its first-party cookie to local ports where the Facebook and Instagram apps could receive it. Meta stopped that communication as of June 3, 2025. Nothing in the November notice describes any link between the browser and Meta's apps, and Meta has not said, according to Wortfilter, whether the new measurement runs in the app, the browser or both.

On-device processing does not by itself settle questions of consent. If the EU and UK exclusion that MetricMaven reports is accurate, the feature would not run in Germany, where the Dresden ruling applies, but would run in the United States, where the Flo and Seattle Children's cases were decided. The consent questions Captain Compliance raised apply there.

Timeline

Summary

Who: Meta, which operates the Meta Pixel and Events Manager; advertisers whose datasets already send events to Meta; Yiqi Wu, chief executive of Aimerce, who highlighted the change on X.

What: A new Meta Pixel setting, "Measure conversions on customer devices," switched on by default, that matches some purchases on the shopper's device using information such as an email address and temporary device data, and tells Meta only that a purchase happened and which campaign gets credit. Advertisers can opt out in Events Manager. Meta has not said whether recovered conversions feed bidding, whether purchase values are included, or how consent is handled.

When: Meta's notice was reported on October 18, 2026, Wu posted on October 26, 2026, and measurement starts on November 12, 2026, 15 days before Black Friday.

Where: In the Meta Pixel settings of Events Manager, for websites using the Pixel. MetricMaven reports, citing Meta's help pages, that the feature is not active for people in the EU or the UK.

Why: Meta says the feature recovers conversions that never make it back to its servers and so do not appear in campaign reporting. For advertisers, reported conversion counts may rise without any change in sales, which makes reconciliation against backend orders the main way to tell recovered purchases from changes in performance.