A media director is asked to approve another audience package. The presentation offers a larger addressable pool, better shopping signals and a route into inventory the brand was not buying last month. The price is reasonable. The spreadsheet still cannot answer the one question that matters: will this purchase reach people the current plan misses, or keep talking to the same people in another place?
That question is usually called audience overlap analysis. It sounds like a deduplication exercise, but treating every repeated exposure as waste is too blunt. A brand trying to introduce itself to new households needs a different answer from a retailer trying to stay on a shopper’s shortlist until a promotion arrives. Both may buy the same channels. Only one is paying chiefly for new reach.
Recent ad launches show how easily the distinction gets lost. Pinterest is opening selected inventory to programmatic buyers, Walmart is putting shopper-derived targeting into streaming television, and new Southeast Asian shopping data shows consideration unfolding over weeks rather than on a single sale day. The platforms bring more ways to find a buyer. None can tell an advertiser, by itself, how much of the resulting demand was truly additional.
Key Takeaways
- Audience overlap is neither automatically waste nor evidence of valuable reinforcement; its value depends on the campaign’s job.
- Incremental reach, new-to-brand buyers and incremental sales measure different things and should not be reported as interchangeable results.
- Cross-channel budget decisions need consistent audience definitions, an agreed baseline and causal testing when the investment is consequential.
Table of contents
Jump to section:
- The same person can be a waste or an opportunity
- A reach result is not a sales result
- Shopping data makes overlap harder to interpret
- Buy against a decision, not an overlap percentage
The same person can be a waste or an opportunity
In an awareness campaign, buying someone who has already seen a sufficient number of ads may add little. When the goal is to introduce a new product to a genuinely new audience, deduplicated reach belongs near the top of the scorecard. Repeated delivery can be expensive because a media plan pays for impressions, not only distinct people who remember the message.
The same logic does not apply without qualification to consideration. A shopper comparing several products may need an initial encounter, independent proof and a reminder that the product remains available. Those exposures can happen in different environments and across several days. A repeated impression is worth something if it performs a new job. It is waste if it is an expensive copy of an impression that already accomplished that job.
This is why the unit of analysis has to be explicit. Overlap between households reached on connected TV is not identical to overlap between browser IDs, logged-in accounts or people in an advertiser’s CRM. A household could include multiple buyers; one person could use several devices. Matching procedures also cover only parts of the reachable population. A dashboard showing low measured overlap cannot establish that the audience itself never overlaps.
A practical definition therefore starts with the buying question. For prospecting, the team might want unduplicated qualified reach among people who have not purchased from the brand. For retention, it may want repeated reach among known customers with a replenishment reason. For a high-consideration purchase, it may want a deliberately sequenced exposure to a relevant audience across independent contexts. Calling all those jobs “more reach” conceals the difference.
Pinterest’s expansion is a live example. Its October 26 programmatic announcement lets enterprise buyers add selected Pinterest display and video inventory through established DSP routes, including Amazon Ads DVA+ and Google Display & Video 360. Magnite is the primary supply-side partner. The benefit is less buying friction, not automatically less duplication.
For media teams, access to a familiar buying interface makes comparisons easier to attempt. It does not solve the harder problem of comparing identity coverage, reach definitions and outcomes. A plan may now contain more inventory while offering exactly the same number of independently measured buyers. More supply is only a gain when it changes what the investment can accomplish.
A reach result is not a sales result
Pinterest highlighted one Moroccanoil campaign that added its inventory to existing video and display spending. The early numbers are encouraging, but they describe several different outcomes. They should not be rolled into a single claim that Pinterest caused more sales.
15% incremental reach and 1% audience overlap were reported for Moroccanoil after Pinterest was added through a programmatic deal using Amazon Ads and Magnite.
The reach finding suggests a useful new group was reached within the campaign’s measurement setup. Yet the public announcement does not provide all the underlying identity assumptions, audience universes, statistical uncertainty or costs needed to predict a similar result for another buyer. A measured overlap of one percent is not a universal characteristic of Pinterest audiences.
63% more new-to-brand buyers were reported among users whose purchase paths included Pinterest; this is a journey association, not demonstrated sales causation.
The distinction is easy to miss in a campaign review. Incremental reach describes additional people exposed within a defined comparison. A new-to-brand buyer may be someone who has no recorded purchase history in a particular retailer or measurement environment. Incremental sales describe purchases that would not have happened without the advertising intervention. None of those is a substitute for the others.
A brand can increase unique exposure without changing sales. It can increase recorded new-to-brand purchases because an existing customer bought through a different merchant. It can also generate incremental sales by reaching familiar shoppers at a moment when the message resolves a real doubt. The campaign’s objective determines which result counts. The measurement method determines what can be claimed.
The IAB and IAB Europe’s November 2025 incrementality guidance draws a similar boundary. It asks advertisers to build a credible counterfactual, control bias and separate measurable effects from noise. Randomized experiments, model-based counterfactuals and econometric approaches can all be useful, but their evidentiary strength depends on design and assumptions.
This matters whenever a platform offers an impressive result from one test. If the counterfactual is not clear, the lift may reflect creative, placement, seasonality, discounts or a difference in the people selected for exposure. A campaign can still be worth expanding, but the reason for expansion should be written as a hypothesis, not as a proven causal mechanism.
A CMO defending new budget therefore needs two columns that should not be merged. One records whether the buy delivered people or contexts the base plan lacked. The other records whether it improved a business outcome beyond what the base plan would have delivered. The first can guide media mix exploration. The second justifies scaling with greater confidence.
That is an uncomfortable standard because the second answer often takes longer to produce. But the absence of a reliable sales-lift result does not turn a reach test into a sales-lift result.
Shopping data makes overlap harder to interpret
Walmart’s new connection with Vibe changes a different part of the calculation. Advertisers can use Walmart-derived shopper audiences while buying connected-TV inventory through Vibe. The segments reflect shopping behavior, life stages and category interest rather than viewing history alone. That is a meaningful difference in signal source.
Walmart Connect’s October 29 explanation says advertisers can select recent category buyers or households showing signs of future purchase interest. Vibe retains campaign buying, optimization and reporting. Existing tools can measure purchases on advertiser websites and run incrementality tests, but the announcement does not establish Walmart in-store sales attribution for this particular integration.
More than 3,000 shopper segments are listed in the Vibe audience catalog, describing available targeting choices rather than proven campaign lift.
A new targeting catalog presents a choice, not a result. A household that bought pet food at Walmart may be valuable to a pet brand, but it could already know the brand, have bought its product elsewhere, or be reached through the brand’s existing streaming buy. Shopping data makes a prediction more relevant. It does not remove the need to measure overlap or causality.
The same household might also be an excellent candidate for reinforcement. Someone who recently considered a category may be more responsive to a demonstration than a cold prospect. For that campaign, overlap could be planned: an earlier ad introduces the product and a later CTV spot supplies a new proof point. The failure is not that the household saw the brand twice. It is paying for the second exposure without knowing what job it was meant to do.
Southeast Asian shopping data makes the timing issue even clearer. Reporting on Criteo’s October 2026 regional findings describes purchases made around Singles’ Day in the previous year. Shoppers often compared brands across retailers for weeks before buying. That is observational commerce data, not a controlled advertising experiment.
22 days was the average period from first product view to purchase around Singles’ Day in Southeast Asia; shoppers browsed an average of 5.8 brands across retailers beforehand.
A buyer who keeps comparing brands is not necessarily an audience a media team should stop reaching after the first impression. But extending frequency cannot be the default response either. A crowded category may require more proof, while a familiar product bought routinely may need only one timely prompt. Media teams should test which message adds information at which stage, rather than interpreting a long journey as permission for unlimited repetition.
The more detailed the audience signal becomes, the more carefully the advertiser must distinguish who appears eligible from who actually responds because of the campaign. Data can make targeting precise without making the resulting explanation true.
Buy against a decision, not an overlap percentage
Suppose a brand has an established social and streaming plan, and a new partner offers an audience derived from retail purchases. The usual procurement conversation asks for reach, cost per thousand impressions and a projected return. A better buying conversation first asks what the new line item is expected to add. The answer changes the measurement design.
If the job is finding people the brand has not reached, the useful baseline is unduplicated qualified reach in the existing plan. The new source should be evaluated for additional people reached, after considering frequency and fees. If the job is converting a shortlist already forming, a campaign needs an outcome test that can distinguish additional purchases from purchases that would have occurred anyway. The same media inventory can be valuable in both roles, but the same evidence cannot prove both.
A cross-channel review can stay disciplined without pretending every platform supplies person-level truth. The team should document the audience unit used by each partner, the coverage of identity matching, the period over which overlap is counted, and the exact definition of new-to-brand. Where deduplication is modeled rather than observed, the uncertainty belongs on the slide beside the result.
The following sequence is a decision aid, not a universal test design:
- State the campaign job. Choose among prospecting, reinforcing consideration and generating incremental transactions. Do not let a blended objective become an excuse for blended evidence.
- Freeze the base plan. Record its audience, creative, spend, frequency, markets, promotions and measurement window before adding a new source.
- Define the comparison. Where feasible, use randomized or carefully matched holdout groups. Keep promotions and timing comparable, and document any changes that could move results independently of media.
- Report reach separately from outcomes. Show deduplicated exposure, frequency, new-to-brand definitions and incremental conversion estimates with uncertainty. A platform attribution total is not a causal lift test.
- Decide what happens if the result is weak. Agree in advance whether a low-overlap but low-lift channel will be optimized, retested or removed. A test without a budget decision is a presentation exercise.
In an earlier ContentGrip analysis of agency and software accountability, Enricko Lukman argued that human judgment still matters when the platform’s objective may be wrong. His published statement concerned agency value, not this specific campaign or a new interview for this article.
Enricko Lukman, CEO at ContentGrow, an AI-powered content marketing agency:
“The agency still earns its fee when it can question the system itself.”
The principle applies in a budget meeting. A platform can optimize toward a reported audience or conversion target. The advertiser must decide whether the target represents growth, defensive coverage, useful repetition or an easier way to claim credit for demand already present. An AI-recommended audience is still a buying proposal, not an explanation of business impact.
No single overlap threshold can make that decision across brands and objectives. Very low overlap may indicate access to unfamiliar people, but those people may be poor prospects. High overlap may be costly saturation, or it may support a deliberate sequence while a shopper decides. The only defensible rule is to specify what repeated contact is supposed to change, then measure whether it changed.
Stop treating overlap as the verdict. Make it the question the budget has to answer.
