Exhibitor scorecard weighting quietly decides which stands look successful

An exhibitor rang about a score of 59 out of 100. She wanted to know which part of it she should fix.

Nobody on the call could answer, because the score had been assembled in a spreadsheet the previous year by somebody who had since left, and the weights lived in a hidden column. We could tell her the number. We could not tell her what it was made of, which meant we could not tell her what would change it. Exhibitor scorecard weighting had already decided her result, and nobody in the building could show her the arithmetic.

A composite score whose weights are not published is an opinion with a decimal point on it. Somebody chose which stand types would come out well, and the exhibitors on the receiving end have no way to argue.

Raw units cannot be added

The first thing a composite has to do is get its components onto a common scale, and this is where most scorecards go wrong before any weighting question arises.

Unique leads is a count that runs from zero into the thousands. Grade mix is a proportion between zero and one. Scanning coverage is a proportion of open hours. Repeat visitor share is another proportion, usually in single-digit percentages. Add those four together in any combination and unique leads decides the answer entirely, because it is three orders of magnitude larger than everything else. Weights applied to raw units do not do what the person setting them thinks they do.

The fix is to convert each component to a percentile inside the exhibitor's own peer group first, then weight. Every component now runs from 0 to 100 and one point of grade mix is worth the same as one point of lead volume before the weights are applied. The weights then carry the whole of your judgement, which is where you want it, because judgement you can see is judgement people can dispute.

Percentile inside the category, not across the whole floor. A software stand and a bearings stand do not capture leads at the same rate, and a composite that ranks them against each other is measuring category, then everything else. Cell size and fallback rules are a separate problem and sit with E24.

The weights I would start from

Four components, on the assumption that the choice of what belongs on the scorecard has already been settled, which is E19's argument rather than this one. Unique leads 0.40. Grade mix 0.25. Scanning coverage 0.20. Repeat visitors 0.15.

Lead volume takes the largest share because it is the objective exhibitors themselves name first. The CEIR 2015 study on exhibitor ROI and performance metric practices put lead generation at the top of stated exhibiting objectives, ahead of brand awareness and reinforcement. A scorecard that buries the thing exhibitors came for under process metrics will be ignored, correctly.

It does not take a majority, though, and that is deliberate. Explori and UFI's 2025 Channel Insights Report, drawing on feedback from more than 3,000 events between 2017 and 2025, measured how well exhibitors felt their objectives had been met and found the largest post-pandemic improvement on meeting existing customers, up 0.68 on their scale, with generating leads and collecting target customer data up 0.40. Launching a new brand, product or service improved 0.56. A stand doing well on the objectives that improved most would score badly on a lead-count-only view, and the exhibitor would know it even if your report did not.

Grade mix at 0.25 is the closest thing an organiser can measure to lead usefulness without asking the exhibitor to hand over CRM outcomes. Scanning coverage at 0.20 is a process metric that belongs in the score because it is the one thing an exhibitor can definitely act on next edition. Repeat visitors at 0.15 gets the smallest share because it is the noisiest of the four on a small stand, where the difference between 6 per cent and 11 per cent can be four badges.

What does changing one weight do to the ranking?

Two stands in the same category, both with percentiles already computed against the 47 companies in that category.

Stand A: unique leads 61st percentile, grade mix 78th, scanning coverage 34th, repeat visitors 55th.

Stand B: unique leads 95th percentile, grade mix 22nd, scanning coverage 88th, repeat visitors 41st.

Under the weights above, Stand A scores 0.40 times 61, plus 0.25 times 78, plus 0.20 times 34, plus 0.15 times 55. That is 24.4 plus 19.5 plus 6.8 plus 8.25, which comes to 58.95, so 59.

Stand B scores 0.40 times 95, plus 0.25 times 22, plus 0.20 times 88, plus 0.15 times 41. That is 38.0 plus 5.5 plus 17.6 plus 6.15, which comes to 67.25, so 67. Stand B is comfortably ahead.

Now swap two weights. Put lead volume at 0.25 and grade mix at 0.40, leaving the other two alone. This is a defensible change and an internal argument you will actually have, because somebody in the room believes lead quality should count for more than lead volume.

Stand A becomes 15.25 plus 31.2 plus 6.8 plus 8.25, which is 61.5.

Stand B becomes 23.75 plus 8.8 plus 17.6 plus 6.15, which is 56.3.

Stand A has moved up 2.5 points and Stand B has fallen 11 points, and the ranking has reversed. Neither stand did anything differently. One line in a configuration file changed, and the exhibitor who was ninth in the category is now nineteenth. If your scorecard is attached to a rebooking conversation or an awards shortlist, that reversal has commercial consequences and no audit trail.

Publish the components beside the total

The single control that makes all of this survivable is showing the four sub-scores next to the composite, with the weight printed against each one.

Then the call with the exhibitor goes differently. She is at 59, made of 61 on volume, 78 on grade mix, 34 on coverage and 55 on repeats, and the 34 is dragging roughly seven points out of her total. Her scanners were active in 22 of the 96 quarter-hour bins the hall was open. That is the thing to fix, it costs her nothing but a briefing to her stand team, and she can see the mechanism rather than taking your word for it.

There is a second reason to publish components, which is that it constrains you. A weighting you have to show to 600 exhibitors is a weighting you have to be able to justify, and the ones that cannot survive that are usually the ones quietly tuned until the accounts you wanted to look good looked good. The same test applies to everything else in an exhibitor analytics pack. If you would not show the reader how a number was built, do not print the number.

Keep the weights fixed for a full edition and change them only between editions, with the change noted in the report. The CEIR 2026 Marketing Spend Decision Report describes exhibitors demanding clearer business impact and stronger post-event reporting from organisers, with exhibitions holding 40.8 per cent of participant marketing spend. An exhibitor comparing this year's score to last year's needs to know whether the movement is theirs or yours.

Should you publish a composite exhibitor score at all?

I would still ship one, and the reason is narrow.

Six hundred exhibitors is too many for anyone on your team to read individually, and a single ordered column lets a sales lead work down a call list in priority order. That is a triage function, and a composite is good at it. Nobody should be making a case-by-case judgement about a specific account from the composite, and if your renewal team is quoting scores back to exhibitors as though they were measurements, the tool has been misused.

So the composite exists to sort. The components exist to explain. Do not let the sorted column become the thing you discuss.

The other legitimate use is tracking a single account across editions, and it works only if the weights held still. An exhibitor moving from 48 to 63 over two editions has improved against their category peers on the same measuring stick, and that is a sentence worth having in a renewal pack. The same movement produced by a weighting change is a sentence you should not write. Version the weights, store the version with each computed score, and refuse to draw an arrow between two scores computed under different versions.

A fifth component gets proposed at this point, usually something about the exhibitor's own reported outcomes or their satisfaction survey response. I would keep it out. A measure covering the whole cycle, from contracting through portal activity to whether the report was ever opened, is a different object with a different purpose, and scoring engagement across that cycle belongs to E34. Survey response is not available for the stands that did not respond, which is most of them, and a component missing for half the population cannot sit inside a weighted sum without silently reweighting everything else for the stands where it is absent. If you want survey data in the picture, report it beside the composite as its own line.

The limit

Percentile normalisation destroys magnitude, and that matters more than it sounds.

If every stand in a category captured between 180 and 210 unique leads, the difference between the 20th percentile and the 80th is thirty leads, and your composite will present that as a sixty-point spread. The scorecard will look decisive about a floor where nothing much distinguished anybody. The reverse happens too. In a category with one enormous stand and forty small ones, the percentile spread understates a genuine gap.

The honest patch is to print the raw value alongside the percentile in every component, so a reader can see that 61st percentile means 240 unique leads against a category median of 205, and decide for themselves whether the gap is worth a conversation. It makes the report denser and it removes the main way these things mislead.

The deeper limit is that a weighted sum assumes the components trade off linearly, and they do not. A stand with excellent grade mix and almost no scanning coverage graded a handful of leads carefully. That is not two thirds of a good show and a third of a bad one. It is a stand that barely captured anything, described flatteringly by an average.

Open whatever produces your current exhibitor score and find the weights. If they are not written down in a place an exhibitor could be shown, write them down this week, then recompute last edition's scores with grade mix and lead volume swapped and count how many stands move more than ten places. That count is how much of your scorecard is judgement rather than measurement.