Incremental reach is the number of people a channel, publisher or placement reached that no other part of the same campaign reached. It is a subtraction rather than an addition: the audience of that channel, minus everyone in it already exposed somewhere else on the plan. The figure exists because reach reported channel by channel cannot be summed. Somebody who saw the same advertisement on a broadcast channel, in a streaming app and in a social feed appears three times across three reports and once in the population.

The question it answers is narrower and more commercially awkward than total reach. Not how many people a buy touched, but how many of them nothing else had touched. A streaming service that mostly reaches viewers the broadcast schedule already covered is selling duplication. One that reaches new households is selling growth. On a media plan the two look identical.

How the figure is produced

Three things are needed. An exposure record for each channel at person or household level. A method of recognising the same person across those records. An agreed universe against which percentages are expressed.

The arithmetic is then simple and order dependent. Deduplicated reach across two channels, minus the reach of the first channel alone, gives the incremental contribution of the second. Reverse the order and both numbers change, because whichever channel is treated as the base absorbs all the overlap. Google's Cross-Media Reach reporting fixes the base explicitly, comparing digital video against a corresponding linear television campaign using licensed third-party television data, available only in selected countries and in some cases only to allowlisted accounts.

Everything therefore rests on the deduplicated combined figure. Where a single data source covers every channel, whether a panel with cross-media meters or a device graph linking identifiers, overlap is observed directly. Where it does not, the overlap is modelled. The oldest model in the trade assumes exposures are independent: combined reach equals one minus the product of the non-exposure rates for each channel, a calculation known in media research as the Sainsbury formula or random duplication. It requires nothing but the individual reach figures, which is why it survived. It also overstates combined reach and understates frequency wherever audiences correlate, and audiences correlate constantly, since heavy viewers of one medium tend to be heavy viewers of another. Academic comparisons of the standard method against a demographically weighted variant reported the weighted version more accurate in 82 per cent of 9,680 within-medium cases and 77 per cent of 968 across-media cases.

The commercial answer has been hybrid measurement. Census-level delivery data from the platform is combined with an independent consented panel, and the panel supplies the overlap the census data cannot see. AudienceProject built its business on that architecture, describing its methodology in a white paper published in April 2026 as audited by outside parties including PwC and Mediasense. Its reports produce total and incremental reach for each medium, an output extended to Polish advertisers across the open web, social media, online video and connected television on 2 March 2026.

Buying platforms compute a restricted version inside their own walls. Amazon released an Exclusive Reach Rate for streaming television inventory on 23 December 2024, quantifying unduplicated audience across Prime Video, Twitch and its other first-party channels. Display and Video 360 added campaign-level reach overlap dimensions in April 2026, having revised the pre-filtering logic of its overlap reports the previous year. Forecasting tools express the same idea as a curve: reach against budget flattens as spend rises, because additional money buys frequency rather than new people.

Origin and evolution

The concept predates digital media. Random duplication has been used to combine vehicle audiences since the middle of the last century, and the reach curve with its diminishing returns is a fixture of the planner role from the era when a plan meant press, television and posters.

Standardisation arrived late. The Media Rating Council issued its Cross-Media Audience Measurement Standards for video in final form in September 2019, defining reach as unique audience with a viewable impression divided by the measured population, and setting out how duplication across media should be handled. Nielsen ONE Ads launched market-wide in January 2023. Google Ads introduced Cross-Media Reach measurement in June 2024, reporting deduplicated reach and frequency across video campaigns and, in the combined digital and television view, incremental and overlap reach against linear. Comscore extended its Social Incremental Audiences product to nine further markets on 15 July 2025.

Fragmentation has kept the problem open. The Coalition for Innovative Media Measurement and the World Federation of Advertisers launched a strategic review of competing European approaches on 15 January 2026, with a steering committee drawn from Unilever, Mediametrie, Auditel, Omnicom, the MRC, Publicis Media, RSMB, CESP, Utiq and WPP Media.

Why it matters for marketers

Incremental reach is the number that decides whether a channel keeps its budget. Combined with cost data it produces a cost per incremental person, which is the only basis on which a connected television line and a social line can be compared without double counting. Omnicom Media Italy integrated total and incremental reach reporting into its client offering on 27 May 2026, the first such agreement for a communications group in that market.

It is also the argument sellers use. Amazon reported advertisers applying its frequency group controls saving up to 26 per cent of budget and gaining up to 21 per cent incremental reach, and an earlier case study found 34 per cent reach overlap across orders from three brands inside one business line. TikTok said internal 2025 data put TopReach at 59 per cent incremental reach against buying its TopView placement alone, while a telecom case study measured by iSpot returned 30 per cent, a gap the company did not explain. Analysis of 519 United States campaigns run between July and October 2025 produced a claim that each 1 per cent of budget allocated to TikTok Pulse delivers 2.4 per cent exclusive reach.

Limitations and disputes

Almost every published incremental reach figure comes from the party selling the inventory. DAZN Media+ said its own cross-platform testing found streaming added 36 per cent reach to Australian Football League campaigns and 27 per cent to National Rugby League campaigns, without disclosing sample size, measurement window or method, and the point of the AudienceProject contract announced on 3 September 2026 is that such figures have not previously been produced independently.

The identity layer is the deeper constraint. Incremental reach is only as good as the recognition of one person across environments, and research commissioned by CIMM and Go Addressable, conducted by Truthset and published on 5 November 2025, examined close to one billion IP records from six providers and found the probabilistic linkage weaker than its pricing implies. Where recognition fails, overlap is invisible and incremental reach is overstated.

Definitions also diverge. Universes differ between measurement firms, household-level and person-level counting produce different answers, and shared screens complicate both. Co-viewing adds people to an impression that device-based counting cannot see, and YouTube's advanced analytics now carry separate co-viewed and unique-reach counters that answer different questions.

Finally, the metric says nothing about effect. A channel can add large numbers of new people who do nothing.

Not the same as

Incrementality measures causal outcomes, estimating the sales or conversions a campaign produced against a counterfactual. Incrementality is normally established through experiments, holdout studies or geographic tests. Incremental reach counts people, and a channel can add reach without adding outcomes.

Addressable reach describes how much of a population a targeting method can theoretically identify. High addressable reach with heavy overlap generates almost no incremental reach.

Deduplicated or unique reach is the combined total across a plan. Incremental reach is the marginal contribution of one component to that total.

Incremental reach in Campaign Manager 360 carries a different meaning entirely. There it refers to additional users reached in each successive time period who were not served an ad in any earlier period, a within-channel accumulation measure that requires a time dimension in the report.

Recent developments

Deduplication has widened beyond single publishers. Nielsen removed the single-publisher restriction on its four-screen deduplication in Japan on 3 August 2026, extending it from YouTube to all measurable connected television inventoryavailable through Nielsen ONE Ads, four months after closing the equivalent gap in Italy.

Amazon has pushed the same logic across accounts rather than across sellers. Cross-account reach and frequency reporting reached manager and cross-manager levels in April 2026, and a case study for Arla Foods in the United Kingdom reported 0.61 per cent audience overlap between Prime Video, Spotify and other streaming services bought through Amazon DSP.

None of this reconciles one seller's portfolio against another's. That is the gap the European review and the Australian collective arrangements are meant to close, on timetables nobody has yet fixed.

Timeline

  • Mid twentieth century: random duplication, or the Sainsbury formula, enters routine use for combining vehicle audiences
  • September 2019: the Media Rating Council issues final Cross-Media Audience Measurement Standards for video
  • January 2023: Nielsen ONE Ads launches market-wide
  • 15 July 2025: Comscore extends Social Incremental Audiences to nine additional markets
  • 5 November 2025: Truthset research for CIMM and Go Addressable reports weak probabilistic IP linkage
  • 15 January 2026: CIMM and the WFA launch a strategic review of European cross-media measurement
  • 2 March 2026: AudienceProject opens total and incremental reach reporting in Poland
  • April 2026: Display and Video 360 adds campaign-level reach overlap dimensions
  • 22 April 2026: AudienceProject publishes its cross-media measurement white paper
  • 27 May 2026: Omnicom Media Italy integrates total and incremental reach reporting
  • 3 August 2026: Nielsen extends four-screen deduplication in Japan beyond YouTube
  • 3 September 2026: DAZN Media+ contracts AudienceProject for incremental reach across Kayo SPORTS, BINGE, Foxtel Go and linear television, available from 2027

Summary

Who: Advertisers and agency planners consume the metric; measurement companies including Nielsen, Comscore, AudienceProject, iSpot and Innovid produce it; platforms including Google, Amazon and TikTok publish their own versions. The Media Rating Council sets standards, and CIMM and the WFA are reviewing competing European approaches.

What: The audience a channel, publisher or placement reached that no other component of the same campaign reached, calculated as deduplicated combined reach minus the reach of the base channel, and expressed in people or as a percentage of a defined universe.

When: Random duplication modelling dates to the middle of the twentieth century. The MRC issued cross-media video standards in September 2019, Google Ads added Cross-Media Reach in June 2024, and independent hybrid measurement expanded through 2025 and 2026, with Australian coverage due in 2027.

Where: Across linear television, connected television, online video, social media, audio and the open web, wherever a campaign runs in more than one environment.

Why: Reach figures reported channel by channel cannot be added together, so a plan that sums them overstates its audience by the size of its overlap. Incremental reach is the correction, and it determines whether a channel is buying new people or repeating the same ones. Its weakness is that most published figures come from the sellers of the inventory, and its accuracy depends on identity linkage that independent research has found unreliable.