Double counting sponsorship impressions is how a package reaches unbelievable reach

The platinum package fulfilment report lands with a headline figure of 480,000 impressions. Nine assets, each with its own line, summed at the bottom. The sponsor's marketing lead reads it, nods, and forwards it to a finance director who divides 480,000 by the 6,000 attendees on the show's own website and sends back a one-line email asking whether that really means eighty.

Double counting sponsorship impressions is what produced the 480,000, and the finance director is right to ask. The arithmetic in the fulfilment report is correct on every line and wrong at the total, because adding impressions across assets sold to the same sponsor at the same event counts one attendee once per asset they walked past.

Where the 480,000 comes from

The nine lines in the package, each estimated the way its own asset type is normally estimated.

Lanyard, 150,000. Aisle sign package, 132,000. Registration area banner, 48,000. App banner, 36,000. Confirmation email footer, 32,000. Sponsored email to the database, 29,000. Hall entrance banner, 26,000. Charging lounge signage, 18,000. App splash screen, 9,000. Those nine sum to 480,000, and each one was produced by somebody who was not thinking about the other eight.

Nothing on the list is fabricated. Every figure is an estimate of exposure events for that asset alone, and if the sponsor had bought only the lanyard, 150,000 would be a reasonable thing to report. The defect appears at the summation, which treats nine overlapping populations as though they were nine separate audiences.

What does the total imply if you divide it by attendance?

This is the sanity check the finance director ran, and it takes ten seconds.

480,000 divided by 6,000 attendees is 80 exposures per person across a three day show. Show hours run eight hours a day, so 24 hours of open floor, which puts it at 3.3 exposures an hour, or one branded contact every eighteen minutes for the entire duration of the event, sustained for three days, for every single person in the building.

Say it out loud and it is obviously wrong. Not impossible, exactly, since the lanyard alone is around the neck of everyone in view, but wrong as a claim about attention and wrong as a basis for a renewal price. Any impressions total that implies an implausible per-person frequency is a sum of overlapping counts, and the division is the fastest way to find out.

Gross, net and the ratio between them

The three-figure convention is settled in digital media and it transfers cleanly.

The MRC Digital Audience-Based Measurement Standards, published in December 2017 and sponsored jointly by the Media Rating Council and the IAB, define reach as the unique users or unduplicated audience exposed at least once in a period, and state explicitly that unique audience reporting requires de-duplicating individuals with multiple exposures over the measured time period. Frequency in the same document is the sum of viewable impressions divided by the sum of unique audience with a viewable impression.

So there are three numbers, and each answers a different question. Gross impressions asks how much exposure was delivered. Net reach asks how many distinct people it touched. Frequency asks how heavily it landed on each of them. A package report carrying only the first of the three invites exactly the misreading the finance director caught.

Working the split by hand

Back to the nine assets. Split them by where they land.

Seven are on site: lanyard, aisle signs, registration area banner, entrance banner, charging lounge, app splash, app banner. Their gross impressions sum to 150,000 plus 132,000 plus 48,000 plus 26,000 plus 18,000 plus 9,000 plus 36,000, which is 419,000. The unique population those seven can possibly touch is capped at the 6,000 people who came, and in practice it is 6,000, because the lanyard and the entrance banner between them cover everybody. On-site frequency is 419,000 divided by 6,000, or 69.8.

Two are off site: the sponsored email at 29,000 opens and the confirmation email footer at 32,000 views. Those overlap heavily with each other and partially with the attendee population. Suppose 31,000 distinct people opened at least one of the two, of whom 6,000 also attended. The incremental unique reach from the email pair is 25,000.

Net reach for the whole package is 6,000 plus 25,000, or 31,000. Gross is 480,000. Frequency is 480,000 divided by 31,000, which is 15.5.

Report those three. Gross impressions 480,000, net reach 31,000, average frequency 15.5. Add the on-site subtotal if the sponsor cares about the floor specifically, because 419,000 gross against 6,000 reach at a frequency of 69.8 is a different and more interesting statement than the blended figure.

Why not just report the biggest number?

Because the biggest number is the one the sponsor will test, and it fails the test in about ten seconds using data on your own website.

There is a longer argument underneath that. The AMEC Barcelona Principles 3.0, published in 2020, set out as their second principle that measurement and evaluation should identify outputs, outcomes and potential impact. An impressions total is an output, and stacking outputs higher does not move anything toward the outcome column. A sponsor who is being sold on volume alone is a sponsor whose renewal depends on next year's volume being larger, which is a treadmill you have to keep running while the floor plan stays the same size.

The commercial reason is sharper. A frequency of 69.8 on the show floor is a statement that your package has saturated its audience several times over, and that the ninth asset added almost no new people. If you know that and the sponsor does not, you are selling incremental assets at full price into a market you can prove is saturated, and the day the sponsor works it out is the day the whole package gets renegotiated. If you know it and act on it, you sell the ninth asset to a second sponsor and the first one gets a lower frequency and a smaller invoice they can defend internally.

What a sponsor actually does with a frequency of 15.5

Frequency is the figure a media buyer knows how to use, which is why handing it over changes the conversation.

At a blended frequency of 15.5 against a reach of 31,000, a buyer can compare your package against an out-of-home buy or a trade publication schedule on terms they use every day. They can also make the argument internally that the show delivered depth on a narrow, qualified audience, which is the actual case for exhibiting and is much stronger than a large flat number that dissolves under division.

The one thing to avoid is quietly changing the reach definition between years to keep the trend line pointing up. If this year's reach counts registrants who did not attend and last year's did not, the comparison is fiction, and the choice of audience base under all of this is a decision worth making once and documenting, which is the subject of choosing the attendance figure everything else divides by. Once the base is fixed, the per-asset ranking work in one CPM column across the card becomes comparable year on year too.

Where this stops

Splitting gross from net requires knowing which people each asset touched, and for most physical assets you do not know that, you assume it.

I assumed the lanyard and entrance banner covered all 6,000 attendees, which is close enough to true that the error is small. I assumed 31,000 distinct openers across two emails, which is knowable from your own sending platform and probably was not looked up. The on-site overlap is the easy case precisely because the cap is hard: no on-site asset can reach anybody who did not come, so the net reach of any all-physical package is bounded by attendance no matter how many assets are in it. That single fact does most of the work, and it means the reach figure for a floor-only package is not really an estimate at all.

Where it genuinely breaks down is a portfolio deal spanning several shows, where the same buyer attends two of your events and the same de-duplication problem reappears one level up, without a badge scan that links the two records. That is an identity resolution problem before it is a sponsorship measurement problem, and no amount of care with the impression arithmetic will fix it. Each figure in the split also carries its own uncertainty, which the band around an impression estimate treats properly, and both are part of the same exhibitor analytics discipline.

Open your largest sponsor's last fulfilment report, sum the impression lines, and divide by the attendance figure on your own website. Whatever per-person frequency comes back is the number the sponsor will compute if they ever bother. If it is above about twenty for an on-site package, add a reach line and a frequency line to the template before the next report goes out.