Double counting across channels
A plan using several channels reaches some people through more than one. Adding the channels' reach figures counts those people multiple times, and the resulting total is larger than the real audience — sometimes substantially.
Every combined reach figure is therefore an estimate of overlap as much as of reach.
Where does the overlap come from?
The same people consuming several media. A shopper who uses a social platform, listens to radio and visits a corner store is in three channels' audiences, and any plan using all three reaches them once.
The more channels a plan uses and the broader each is, the larger the overlap and the more misleading a simple sum becomes.
How is it handled properly?
With a de-duplication model that estimates joint audiences rather than summing marginals. Doing that well requires data on cross-media consumption, which is why the figures come from research panels rather than from arithmetic.
In the absence of such data, the honest presentation is channel-by-channel reach with a stated caution that they are not additive.
What is the specific risk with in-store media?
That its audience is presented as fully incremental when it is partly the same people. Store shoppers use other media, and the incremental portion is an empirical question rather than the whole audience.
Overstating incrementality is the version of this error that a store network is most likely to commit, and it is the one most likely to be checked by a sophisticated planner.
How should a combined figure be presented?
With the de-duplication method named, or not presented at all. A total that is the sum of parts should be labelled as a sum of parts, which immediately tells a reader what it is.
Presenting an undeduplicated sum as a plan's reach is the kind of claim that damages credibility when somebody checks it.
Does this affect frequency too?
Yes, in the other direction. A person reached by several channels receives higher total frequency than any single channel's figure suggests, which matters for wear-out and for budget efficiency.
A plan that reports reach per channel and never considers combined frequency is missing half the interaction.
What can be measured rather than modelled?
Outcomes. A test that varies channel combinations and measures the result sidesteps the audience modelling entirely, and it answers the commercial question directly.
That is more work than a reach calculation and it produces a considerably stronger claim.
What should a buyer ask?
Whether a combined figure is deduplicated, by what method, and against what population. Three questions, and they are routinely unasked.
Frequently asked questions
Can channel reach figures ever be added?
Only where the audiences are genuinely disjoint, which is rare. In practice a sum should always be treated as an upper bound rather than an estimate.
How large is the overlap typically?
It depends entirely on the channels and the population and there is no general figure. Two broad channels in the same market will overlap heavily; a narrow specialist channel and a broad one less so.
Does in-store media overlap with retail media online?
Frequently, since both target shoppers in the same categories. Whether the same individuals are reached depends on where they shop and what they use.
What is a currency, and does this channel have one?
A currency is an agreed measurement standard a market transacts on. Out-of-home measurement practice is developed by bodies such as Geopath, and in-store networks vary in how closely they align to it.
Should I discount a network's reach claim?
Not automatically, and you should understand its population and method. A claim about a defined store audience is a different thing from a claim about a market.
How do I compare two channels fairly?
On cost per unit of a defined outcome rather than on reach, where an outcome measure is available. Reach comparisons across differently measured channels compare methods as much as audiences.
Does double counting affect frequency capping?
It means the effective frequency across a plan is higher than any channel's figure suggests, which is a reason to consider total exposure rather than channel-level caps.
Is a modelled deduplication trustworthy?
As trustworthy as its inputs, which should be stated. A deduplication based on a research panel is a real estimate; one based on an assumed overlap coefficient is an assumption.
What is the safest presentation?
Reach by channel, clearly not summed, with the overlap question named. Anything more requires evidence most plans do not have.
### What should a plan document say about overlap?
That reach figures are not additive, which channels were estimated by which method, and what the combined figure means if one is given. Three lines, and they prevent a table being read as arithmetic.
Where no deduplication is available, saying so is better than producing a total. A plan that acknowledges the limit is easier to defend than one that quietly ignores it.
Does this matter for a small plan?
It matters proportionally more, because a small plan's channels are more likely to be concentrated in the same population. Two channels reaching the same neighbourhood overlap heavily.
### What should a plan present instead of a combined figure?
Reach by channel, the largest single channel's reach as a floor, and a note that the combined figure is unknown without deduplication. That is more honest and it is usually sufficient for the decision.
Where a combined number is genuinely required, a deduplicated estimate with a named method is the only defensible version.
Does frequency double count too?
In the opposite direction: combined frequency is higher than any channel's figure, which matters for wear-out and for whether additional weight is useful.
How should a sceptical planner test a claim?
By asking for the overlap assumption. A combined reach figure with no overlap assumption has assumed zero overlap, which is never true.
Considering NRS Digital Media? Start at nrsdigitalmedia.com, read the media deck, or contact the team.