Cross-media reach looks solid on the planning slide and dissolves in the post-campaign report.
Linear TV, connected TV, walled gardens and retail media each report their own audience in their own units. No single measurement layer can verify the total after the budget has been spent.
The pattern is structural.
In an Affinity Solutions survey of 210 senior marketing leaders conducted in March 2026, more than four in five said they rely primarily on attributed digital conversions, modeled outputs or brand lift when optimizing in flight, rather than verified purchase data.
Those are channel-level signals, and none of them can say how many unique people a plan actually reached. Affinity sells purchase data, so read the finding as a vendor's survey.
This piece covers:
Our view: reach is both a planning promise and a measurement claim, and in most organizations two different teams own those two things with two different definitions.
A measurement layer earns its place only when it changes a decision. So the first job is agreeing what a reached person means and reconciling one campaign honestly.
Media and creative teams should then read the same report, because a segment reached at high frequency with no response usually points to the message.
Cross-media reach is the number of unique people a campaign reached across all media, each person counted once. It's the deduplicated total across linear TV, radio, out-of-home, connected TV, digital and retail media, and it's always smaller than the sum of what each channel reports.
Two related terms get mixed up with it:
The distinction matters because a cross-channel dashboard can look complete while the largest reach-driving channel sits outside it.
A reporting layer covering the major search, social and video platforms is cross-channel. The same plan with linear TV and retail media deduplicated into one audience estimate is cross-media.
Most of the confusion is inherited from tools.
Reporting stacks were built channel by channel, so executives receive a cross-channel view labeled as the whole picture.
Nielsen's 2025 Annual Marketing Report, a survey of 1,400 marketers, found that only 32% measure their traditional and digital media spend together.
A useful first step: classify the current reporting stack as cross-channel or cross-media, and name every channel that sits outside it.
Plan against the most comparable reach number available, and record its source and deduplication method next to every figure, even when a less comparable number is larger.
Three kinds of reach figure show up in most plans, and they behave very differently.
Platform-reported reach is controlled by the seller and deduplicated only inside that platform.
It's usually the largest number on the page and the least comparable, because walled gardens, connected TV, retail media and linear TV each report in different units.
Panel-based reach comes from TV panels, cross-media panels and standards such as the Media Rating Council's cross-media audience measurement standards for video.
It's comparable across media because one methodology describes everyone. Panels are also sample-limited, slow and weak on small or local audiences.
Modeled reach uses virtual IDs, probabilistic matching or a platform's own cross-media tool to fill gaps across walled gardens.
Google's Cross-Media Reach tool is one example. It deduplicates across what Google can observe, and the buyer can't audit the model behind it. The Google Ads Cross-Media Reach section below covers what it shows.
| Reach source | Who controls it | What it deduplicates | What it misses | Best use in planning |
|---|---|---|---|---|
| Platform-reported | The seller | Its own logged-in users or devices | Everyone reached on other platforms and offline | In-platform frequency and pacing |
| Panel-based | Independent measurement body or vendor | People across all measured media | Small, local and niche audiences; slow to report | Cross-media comparison and unique reach estimates |
| Modeled or blended | Vendor or platform running the model | Overlap across inputs, by assumption | Whatever the inputs can't see; assumptions are rarely auditable | Filling gaps between walled gardens, with stated confidence |
A hypothetical plan shows the stakes: 30 million reach from TV, 20 million from social and 10 million from CTV, reported as 60 million total.
A panel or model that accounts for overlap would return a smaller unique figure, and the difference is the planning error, carried into every frequency and cost-per-reach calculation downstream.
The rule for media planning follows from that.
Plan against the most comparable number, then:
The later sections take each layer in turn.
Reach totals double count because three things overlap at once: identity graphs, attribution windows and the populations each data source describes.
Each walled garden deduplicates against its own logged-in users. The same person on YouTube, Meta and a retailer's media network is three people in three reports, and no shared currency exists to collapse them back into one.
Seven-day, 28-day and view-through windows overlap across platforms, so one conversion can be claimed by several of them.
The result inflates attributed outcomes, so the channels that over-claim look efficient and win more budget. That budget then buys reach where the overlap is already highest.
Linear TV panels estimate people in households, CTV logs count devices or impressions, and digital platforms count cookies or logged-in IDs.
Summing them adds households to devices to IDs, which is a unit error before it's a measurement error.
A hypothetical makes it concrete: one household reached on linear TV, served the same creative on CTV and shown it again in a retail media app is counted once by a panel and three times by summed platform reports.
The commercial consequence runs in three directions:
The check: pull one campaign's platform reports and confirm whether attribution windows and unit definitions were aligned before anyone compared totals.
Video: Bruno Furnari on overlapping reach, cross-media measurement and why brands won't wait (Adwanted UK, YouTube)
Privacy constraints change the fix that's available. Deterministic cross-media matching now covers a shrinking share of the audience, so planning has to work with modeled ranges and stated confidence.
Fewer third-party cookies and mobile identifiers mean fewer people can be matched deterministically across platforms. The deterministic share of any cross-platform estimate has been falling, and plans should assume it stays low.
Server-side tracking and a first-party data warehouse improve a brand's own measurement.
They can deduplicate site, app and CRM touches, but they do not reach into walled gardens or offline media, so a first-party warehouse still can't tell you whether a customer was reached on linear TV.
Probabilistic matching, platform cross-media reach tools and panels fill the gap, each with assumptions the buyer must understand.
Ask every measurement partner which share of reported reach is deterministic and which is modeled, and how that share has moved over the last few quarters.
Deterministic cross-media reach isn't coming back.
The practical response is to plan against ranges, with the confidence level written next to each figure.
Google Ads Cross-Media Reach is a report inside Google Ads that deduplicates reach across your Google video campaigns and, in some regions, estimates how much reach YouTube added on top of linear TV.
It's useful, and it's still one seller's view of the plan.
The report sits under Measurement in Google Ads. It combines video campaigns into one total, deduplicated reach and reports:
Google recommends it for awareness and consideration campaigns. Since October 2025, digital and YouTube video reports have moved to Brand report, while TV campaign reports stay on the Cross-Media Reach page (Google Ads Help).
In select regions, the report compares your digital campaigns with your linear TV campaigns and estimates incremental reach using licensed third-party TV data.
Where single-source cross-media panel data isn't available, Google uses a method it calls a Conditional Independence Approximation. In plain terms, it models how much the two audiences overlap instead of observing it.
Use it for in-platform reach and frequency and to size YouTube's incremental reach over TV. Keep it out of the cross-media total, or add it with its boundary written next to the number.
Advertiser bodies are building shared, deduplicated reach measurement market by market. In the UK and the US it's now usable for some channels, and coverage is still partial everywhere.
| Program | Market | What it measures | Status (November 2026) |
|---|---|---|---|
| MRC Cross-Media Audience Measurement Standards | US | Requirements for measuring deduplicated video reach across TV, OTT and digital | Phase I (video) final. A standard to audit against, not a product |
| WFA North Star framework | Global | Principles for privacy, infrastructure, consistent metrics and governance | The blueprint Origin and Aquila follow |
| Origin (ISBA) | UK | Deduplicated reach and frequency across linear TV, online video and online display, using a virtual ID model | Open to all advertisers since mid-2025; about 70 supporting organisations; investment talks with Fifty5Blue in 2026 |
| Aquila (ANA) | US | Deduplicated reach and frequency across linear TV, CTV and online video, with inventory from Amazon, Google, Meta, iHeartRadio and TikTok | Commercial rollout in spring 2026; display and audio planned |
| Nielsen four-screen deduplication | Japan, Italy, Malaysia and other markets | Deduplicated ad reach across computer, mobile, CTV and linear TV | Live in selected markets |
For planning today, use Origin or Aquila for the channels they cover and treat MRC standards as the bar for TV and video measurement.
Everything outside their coverage still needs a stated overlap assumption. Aquila also says it doesn't intend to become a trading currency, so plans will still be bought on existing currencies.
A data clean room lets a brand and its media partners match audiences without handing over raw user data, so reach can be deduplicated across every partner inside the room.
The total is only as complete as the list of partners who join.
Each partner brings its exposure data into a shared, privacy-controlled environment. The room matches people across all sources at once and returns one deduplicated total, which comes back lower than the sum of the individual reports.
As Decentriq explains, that simultaneous reconciliation is what makes reach a harder computation than lift.
Publicis Media France used Amazon Publisher Cloud to deduplicate audiences for Mugler Angel across Prime Video and other premium video, including the broadcasters TF1 and M6.
The campaign reached 3.9 million unique users, and 79% saw ads from only one inventory source, so overlap was low.
L'Oréal used the result to plan raising its broadcaster video-on-demand share from 16% to 30% for 2026 (Amazon Ads case study).
Plan each channel by the new people it adds to what the plan already reaches. Four steps turn that into a budget split.
Write the target audience, the share of it to reach and the effective frequency for each objective before any channel gets budget. Awareness and consideration goals usually need different frequencies.
For each channel, estimate how many people it reaches who aren't already reached by the channels above it in the plan.
Google's research on the TV and online mix models exactly this: how a campaign's reach changes when budget moves from TV to online. Cross-media panels answer the question directly, but they're expensive to scale across all media, so most planners use a model or a stated overlap assumption.
Platform caps only work inside each platform. Set the cap at plan level, in people, and translate it into each channel's own cap using the overlap from step 2.
Reach curves flatten: each extra dollar in a channel buys fewer new people. When a channel's curve flattens, the next dollar usually adds more new people in a channel with less overlap.
Fieldtrip's guide to the media planning process covers where this step sits in the wider plan.
Build the baseline in four steps using existing reports, in one quarter and without a new vendor.
List each channel's reach figure, who produces it, the unit it uses (household, device, ID) and the deduplication method behind it. The output is one sheet with a row per channel.
Write a single definition of a reached person and a frequency cap, and have planning, media buying and measurement sign it. The definition should state the unit, the exposure threshold that counts as reached, and how units will be converted when channels disagree.
Take one finished campaign and compare its planned reach against the best available deduplicated estimate, whether from a panel, a model or a stated overlap assumption. Document the gap and its causes row by row.
A hypothetical illustrates the output: a team that planned 60 million summed reach finds a reconciled estimate of materially fewer unique people and higher real frequency. The documented gap, with its causes, becomes the business case for the next measurement layer.
Name the decision the next investment must change: a budget shift, a frequency cap, a channel exit. If no decision would change, the tool isn't needed yet, however good the demo.
This baseline converts proxy-driven in-flight habits into a reviewable record. It costs little and uses data already in the building.
It also turns the later conversation about MMM or a clean room into a conversation about a specific, documented gap.
Pro tip: put the reconciliation sheet in front of the channel owners before the measurement team presents it.
The causes of the gap are usually known to the people buying each channel, and the sheet lands better as shared work than as an audit.
MMM, MTA and incrementality tests answer different questions, and the useful move is to map each one to the decision it should drive.
Neither MMM nor MTA fixes reach deduplication on its own. Both measure outcomes against reach inputs, so bad reach inputs degrade both.
Fieldtrip's guides to media mix modeling for enterprise marketers and multi-touch attribution and marketing mix modeling cover each method in depth. This section only covers how they depend on reach.
MMM works on aggregate data, which makes it privacy-safe and lets it cover offline media and walled gardens alike. It answers budget allocation questions over months and is blind to creative-level and short-horizon decisions.
MTA works at user level and runs fast, which makes it useful for digital path analysis.
It's structurally weak on walled gardens and offline media, and it inherits the same inflation as platform reports because it depends on the same exposure logs. The signal loss described in the privacy section narrows its coverage further.
Geo or holdout tests are the only method that measures causal lift directly. Use them to calibrate both MMM and MTA, since a model that disagrees with a clean holdout needs correcting, whichever direction the gap runs.
| Method | Question it answers | Data it needs | Walled-garden and offline coverage | Main blind spot |
|---|---|---|---|---|
| Media mix modeling | How should budget be allocated across channels over months | Aggregate spend, exposure and outcome time series | Strong; works on totals from any channel | Creative-level and short-horizon decisions |
| Multi-touch attribution | Which digital touchpoints preceded conversions | User-level exposure and conversion logs | Weak; can't see inside walled gardens or offline media | Inflation from shared logs and overlapping windows |
| Incrementality tests | Did this spend cause lift | A holdout or geo split with matched outcome data | Good where exposure can be withheld by region or audience | Slow to run, limited to the tested channel and period |
A conceptual pairing shows the blind spots in practice. An MMM showing TV drives sales can't tell a creative team which spot worked; an MTA showing paid social drove conversions can't see the TV exposure that preceded them.
Verified purchase data means transactions matched to ad exposure, and all three methods get more reliable with it.
In the same Affinity Solutions survey, roughly one in five marketers already use it as their main in-flight signal, and nearly half expect double-digit ROAS gains from optimizing on it in real time.
Treat that as an expectation to test.
Brand lift studies still cover what purchase data can't:
Action: map each method currently in use to the decision it actually informs, and note any decision that relies on a method structurally blind to it.
A cross-media reach report shows planned and delivered reach side by side, channel by channel, with the source and deduplication method next to every number. The hypothetical example below continues the 60 million plan from earlier.
| Channel | Planned reach | Delivered reach | Also reached elsewhere | Average frequency | Source and dedup method |
|---|---|---|---|---|---|
| Linear TV | 30M | 27M | 35% | 4.2 | TV panel, people |
| Social | 20M | 19M | 50% | 6.1 | Platform-reported, logged-in users |
| CTV | 10M | 9M | 60% | 3.8 | Census logs, devices |
| Total | 60M summed | 41M deduplicated | 6.0 | Modeled overlap, stated confidence |
Report the deduplicated total next to the summed total, and how much of each channel's audience was also reached elsewhere. The gap between the two totals is the planning error.
An average hides the people who saw an ad 15 times. Show the share of people reached at 1, 2 to 3, 4 to 6 and 7 or more exposures, so the frequency cap can be checked against real people.
Show both numbers, the gap and its likely causes by channel, and name the decision the gap changes.
A single blended reach figure hides the problem. A documented gap becomes the template for the next plan and the evidence for whether another measurement layer is worth adding.
Every number carries its source type (platform, panel, model or clean room) and its deduplication method, so a reader can tell which figures can be compared and which can't.
Report reach and frequency per creative unit, so creative teams can see where exposure concentrated and respond to it.
Channel-level reach hides where the pressure landed. A per-creative view shows which units carried the frequency and in which segments, which is the information a creative team needs to decide what to change.
Treat a segment with verified reach and flat response as a creative or message problem before treating it as a media problem.
A segment reached at high frequency with flat response across two flights signals a message test; another budget shift would be the wrong lever.
After each flight the media team passes reached segments, frequency and response by segment to the creative team, and creative returns hypotheses to test in the next flight.
That loop is how craft and performance inform each other through evidence. Fieldtrip's guide to creative testing measurement for large brands covers how to set up those tests.
Next step: add a creative-level reach and response view to the next post-campaign report and schedule one joint media and creative review.
Reach breaks down between planning and measurement when the two use different definitions and nobody owns the reconciliation.
The sequence that fixes it:
Fieldtrip is a group of specialist agencies connecting strategy, creative, media and measurement. Small autonomous teams own execution with shared strategy and quality standards, and engagements usually begin with one defined problem, working alongside existing internal teams and agency partners.
If a recent campaign's planned and reported reach don't agree, discuss your cross-media measurement with Fieldtrip and bring that gap along. It's the most useful starting point for a first conversation.
Yes. The garden deduplicates its own users, but it still can't see the TV, retail media or out-of-home exposure the same people receive.
Even a concentrated plan needs an agreed definition of a reached person and an overlap assumption for the channels outside the garden, or frequency against real people will run above plan.
Apply the principles at any budget: discount platform attribution, agree one reach definition, and reconcile one campaign against a stated overlap assumption.
Full MMM and clean rooms are justified when they'd change a budget, frequency or channel decision large enough to pay for them; before then, a documented baseline and occasional geo holdouts do most of the work.
No. AI can improve models and speed up reconciliation, but it can't create cross-platform identity that platforms don't share or verified purchase data that isn't collected.
Fed the same fragmented inputs, an automated optimizer reproduces the same overcounts faster, so data quality decides what any AI layer can deliver.
Convert units explicitly and write the assumption down: households to people using an agreed persons-per-household factor, and impressions to estimated unique devices using a stated frequency assumption or a modeled dedup.
Record the confidence of each conversion next to the figure, and keep panel-derived and census-derived numbers in separate columns until the conversion has been applied.
