Single publisher attribution is the practice of measuring conversions inside one publisher or ad platform, using only the ad exposures that publisher itself recorded. Meta counts the purchases that followed Meta ads, Google Ads counts the purchases that followed Google ads, and an app network counts the installs that followed its own clicks. It exists because the seller of the media usually holds the only exposure log, login data and click record needed to connect an ad to an outcome, and because advertisers want a number quickly, where they set their bids.

The phrase is descriptive rather than a defined industry standard. No IAB, Media Rating Council (MRC) or W3C document defines it, and practitioners use several near-synonyms: platform-reported attribution, in-platform attribution, self-attribution and, in mobile apps, self-attributing networks (SANs). The common thread is that the party doing the counting is also the party being credited.

How it works

The mechanics follow the same five steps on most large platforms.

First, the advertiser sends conversion data to the publisher. That happens through a browser tag or pixel, a server-to-server feed such as Meta's Conversions API, or an offline upload of sales records. Second, the publisher matches each conversion to people it has seen, using cookies, device identifiers, logged-in accounts or hashed email addresses and phone numbers. Third, it applies a lookback window: only exposures within a set number of days before the conversion count. Meta's default since 2021 has been 7-day click and 1-day view. Google Ads uses a conversion window that advertisers can set per conversion action. Fourth, it applies a model to its own touchpoints, whether last click, a rule-based split or a data-driven model. Fifth, the result is reported in the dashboard and, in most systems, fed back into automated bidding.

The defining feature is in step four. The model only ever sees one publisher's touchpoints. A conversion preceded by a Google search, a TikTok video and a Meta Reel can be credited in full by each of the three, because none of them sees the other two.

A simple worked example shows the arithmetic. Suppose a retailer records 1,000 online orders in a month. Meta reports 620 attributed purchases, Google Ads reports 540 and TikTok 180. The platform total is 1,340, or 134% of real orders. None is necessarily wrong by its own rules. The excess is the overlap, and no single platform can see it.

Google's Demand Gen product makes the logic explicit. Its "conversions (Platform Comparable)" columns, which Google rolled out in March 2025, limit credit to Demand Gen touchpoints and give "full credit" to the last interaction, following a waterfall of last click, then last engaged view, then last view, according to Google's help documentation. Google says the columns exist to make Demand Gen comparable with "other advertising platforms" that use similar defaults, and that they do not affect bidding.

Mobile apps and the claim system

App marketing formalised single publisher attribution earlier than the web. When an app is installed, a mobile measurement partner (MMP) such as AppsFlyer or Adjust queries each self-attributing network with the device's advertising identifier. Meta, Google Ads, TikTok, Snap, X and Apple Search Ads each check their own records and reply with any engagement they logged, a process AppsFlyer documents as SRNs that "self-attribute using the device ID when notified of an install." Every network can claim the same install. The MMP then deduplicates the claims and credits the install once, usually by last touch. That is why network dashboards rarely match MMP dashboards: networks report what they claimed, and MMPs report what they granted.

Apple's own frameworks narrowed this further. SKAdNetwork, introduced in 2018, and AdAttributionKit, introduced at Apple's WWDC in June 2024, send delayed, aggregated postbacks to the ad network that won the install, rather than a user-level log to the advertiser.

Origin and evolution

Search engines reported conversions against their own paid clicks from the early 2000s, and last click within one account was the default for years.

The academic framing came later. Ron Berman's paper "Beyond the Last Touch: Attribution in Online Advertising", circulated as a working paper in September 2013 and published in Marketing Science in 2018, modelled "single publisher campaigns" against campaigns where advertisers buy from several publishers that reach the same "multihoming" consumers. The paper found that last-touch rules "overincentivize ad exposures" once more than one publisher is involved, and that a Shapley value method raised advertiser profits. PPC Land summarised the work in The attribution illusion, quoting Berman: "Attribution creates a virtual competition between publishers, resulting in a team compensation problem." In the same data set, 54% of advertisers used last-touch attribution.

Trust problems followed. In September 2016 Facebook disclosed that it had overstated average video viewing time by 60% to 80% for two years, according to Tubefilter's account of the Wall Street Journal report. It was not an attribution figure, but it hardened agency demands for independent checks. Facebook opened its ads to comScore and Integral Ad Science in January 2017, while YouTube cut third-party measurement partners in May 2018, around the time the General Data Protection Regulation (GDPR) took effect.

Privacy rules then pushed the industry back toward the publisher's own data. Apple's App Tracking Transparency prompt, enforced from iOS 14.5 in April 2021, cut access to device identifiers, and cross-site cookies disappeared from Safari and Firefox. With fewer shared identifiers, independent multi-touch tools lost visibility, and platform-modelled conversions filled the gap.

The newest twist runs the other way. The W3C's Private Advertising Technology Working Group is drafting "Attribution Level 1", edited by engineers from Google, Meta and Mozilla. Its draft dated October 14, 2026 lets a browser match conversions to impressions in a differentially private, aggregated report. A conversion query can name the impressionSites it will match, so one API can serve one-publisher or cross-publisher queries. Mozilla tested a version of the idea in Firefox 128 in July 2024 and drew complaints from privacy groups.

Why it matters for marketers

Single publisher attribution feeds the bidding systems that spend most digital budgets. A platform's automated bidding optimises toward the conversions it can see and credit. If its window includes 1-day view-through, its algorithm learns to value people who were likely to buy anyway.

Platforms keep changing those rules unilaterally. Meta removed the 7-day and 28-day view-through windows from its Ads Insights API on January 12, 2026, according to an October 2025 developer notice. In March 2026 it restricted click-through attribution to link clicks, moved likes, shares and saves into a new "engage-through" category and cut the engaged-view threshold from 10 seconds to 5. Each change shifts reported return on ad spend (ROAS) without any change in actual sales.

Publishers argue their own numbers can also be too low. Meta's "Suite of Truth" white paper said that across 307 Conversion Lift studies with 54 advertisers, rules-based models undervalued Meta by 31% at the median. The figure comes from Meta's own experiments, which is exactly the critics' point.

Limitations and disputes

The central criticism is that the publisher marks its own homework. Exposure logs, matching logic and models are proprietary, and outside auditors see little of them.

Over-counting is the second problem. Because every publisher works from its own touchpoints, summed platform conversions routinely exceed real conversions, as in the worked example above.

Third, attribution is correlational. Brett Gordon and colleagues compared observational methods with 15 randomised Facebook experiments in a 2019 Marketing Science paper, and a 2023 follow-up covered 663. Karan Dhir, a measurement product lead formerly at Amazon and Walmart, argued in September 2026 that such methods miss experimental results "frequently by a factor of 3 or more". The 3x figure is his characterisation; the papers state no single multiplier.

Not everyone sees accuracy as the main problem. A March 2026 study by the Coalition for Innovative Media Measurement (CIMM) and the 4As, surveying 197 senior brand marketers, found 43% rated cross-platform measurement a major barrier, and framed the issue as reconciling systems rather than distrusting any single one. "They're not looking for a single source of truth, but clarity about how different truths relate," said Jon Watts, managing director of CIMM.

Not the same as

Multi-touch attribution (MTA) splits credit across touchpoints from several publishers, usually run by an advertiser or an independent vendor. Single publisher attribution can be multi-touch inside one platform and still blind to everyone else.

Last-click attribution is a rule about which touch gets credit. A last-click model can be run across all channels in an analytics tool; single publisher attribution describes whose data is used.

Closed-loop measurement connects ad exposure to a sale inside one company's systems, typically a retailer that sells both ads and goods. It is a form of single publisher attribution where the publisher also owns the transaction, as with Amazon Attribution, which reports how off-Amazon media drives Amazon sales.

Incrementality testing uses holdout groups to estimate causal lift. Platforms run such tests too, so the method and the referee are separate questions.

In academic work, "single publisher" can also mean a campaign bought from one publisher only, as in Berman's model: a campaign structure, not a measurement method.

Recent developments

The past year has brought partial openings. On May 20, 2026, Google and TransUnion said TransUnion would become the only multi-touch attribution provider for YouTube, adding YouTube exposures to cross-publisher models after a pilot with more than 15 customers. Meta, alongside its own changes, said in March 2026 that Northbeam and Triple Whale would incorporate both clicks and views from its ads. As of October 2026, though, each of these is still controlled by the publisher whose ads are being measured.

Timeline

  • 2000s - Search engines report conversions against their own paid clicks, with last-click inside one account the common default.
  • September 2013 - Ron Berman circulates "Beyond the Last Touch", contrasting single publisher and multi-publisher campaigns.
  • September 2016 - Facebook discloses that average video viewing time was overstated by 60% to 80% for two years.
  • January 2017 - Facebook adds comScore and Integral Ad Science as third-party measurement partners.
  • 2018 - Apple introduces SKAdNetwork; Berman's paper is published in Marketing Science.
  • May 2018 - YouTube limits third-party measurement partners as GDPR takes effect.
  • 2019 - Gordon, Zettelmeyer, Bhargava and Chapsky publish their comparison of attribution methods with 15 Facebook experiments.
  • 2021 - Apple enforces App Tracking Transparency from iOS 14.5 in April; Meta moves its default attribution setting to 7-day click and 1-day view.
  • February 2022 - Mozilla and Meta propose Interoperable Private Attribution in the W3C Private Advertising Technology Community Group.
  • June 2024 - Apple introduces AdAttributionKit at WWDC24.
  • July 2024 - Firefox 128 ships an experimental Privacy-Preserving Attribution feature.
  • August 2024 - Meta announces its incremental attribution setting.
  • March 2025 - Google adds Platform Comparable conversion columns to Demand Gen.
  • May 28, 2025 - Meta publishes its "Suite of Truth" white paper.
  • January 12, 2026 - Meta's 7-day and 28-day view-through windows stop returning data in the Ads Insights API.
  • March 2026 - Meta limits click-through attribution to link clicks and introduces engage-through attribution.
  • May 20, 2026 - TransUnion becomes the sole multi-touch attribution provider for YouTube.
  • October 14, 2026 - Date of the latest Attribution Level 1 draft from the W3C working group.

Summary

Who: Ad platforms and publishers such as Meta, Google, TikTok, Snap, Amazon and Apple Search Ads, which measure conversions from their own ads; advertisers and agencies who rely on those reports; and independent vendors, MMPs and standards bodies that try to reconcile them.

What: A measurement approach in which one publisher credits conversions using only the ad exposures it recorded, applying its own lookback windows and models, so several publishers can claim the same conversion.

When: Present since the earliest search advertising, studied academically from 2013, and reshaped since 2021 by privacy changes and repeated platform rule changes through 2026.

Where: Inside platform dashboards and bidding systems, in mobile app install reporting through self-attributing networks, and increasingly in browser-level privacy APIs being drafted at the W3C.

Why: The publisher holds the exposure data, the reports are immediate and they drive automated bidding. The cost is over-counting across platforms and a referee who is also a player.