Exhibitor tenure and lead volume rarely move in the direction you expect
Exhibitor tenure and lead volume do not rise together indefinitely. Plotted as median unique leads per square metre against consecutive editions exhibited, the curve typically climbs for a few editions and then drifts down. Much of the rise is survivorship, because each point is computed on the companies that chose to come back.
The slide was called loyalty and it had two bars on it. Exhibitors with ten or more editions behind them, and everybody else. The long tenured bar was taller, and the slide was there to justify weighting tenure more heavily in the priority points formula.
The bars were totals. Long tenured exhibitors hold more space, so the slide was a picture of the floorplan with a different label on it. Exhibitor tenure and lead volume were being read off a chart that was mostly measuring square metres.
Redrawn on unique leads per square metre, the shape was not two bars. It was a curve that went up for four editions and then came back down, and the companies on the loyalty slide were sitting on the wrong side of it.
Drawing the curve on your own floor
Take every exhibiting company at your last edition, attach the count of consecutive editions it has exhibited including this one, and compute the median unique leads per square metre for each tenure value. The space denominator matters here and E25 sets out why the contracted figure is the defensible one, which is worth reading before you divide anything. The curve belongs near the front of an exhibitor analytics pack, because it settles an argument the sales team is going to have anyway.
Medians, not means, and the count printed beside every point. A tenure curve without the counts underneath it is unreadable, for a reason that becomes the main argument of this post.
On a floor of 600 exhibiting companies it comes out something like this.
| Editions exhibited | Median unique leads per square metre | Exhibiting companies |
|---|---|---|
| 1 | 4.1 | 118 |
| 2 | 5.2 | 96 |
| 3 | 5.9 | 81 |
| 4 | 6.1 | 66 |
| 5 | 6.0 | 54 |
| 6 | 5.7 | 44 |
| 7 | 5.5 | 38 |
| 8 | 5.2 | 31 |
| 9 or more | 4.9 | 72 |
The peak is at four editions, at 6.1. First edition stands sit at 4.1, so the rise is 2.0, which is 48.8 per cent. From the peak to the long tenured group the fall is 6.1 down to 4.9, a drop of 1.2, or 19.7 per cent off the peak.
Two things follow immediately. The stands doing the most with their space are in their third to fifth edition, and the group the loyalty slide was built to reward is capturing less per square metre than a stand in its second year.
The rising limb is a learning curve and it flattens
The left hand side of that curve is not mysterious. A stand gets better at the show by having been to the show.
Wright described the general form in the Journal of the Aeronautical Sciences in 1936, working on airframe production costs, and reported labour input per unit falling by a roughly constant proportion each time cumulative output doubled. The gains arrive on a doubling schedule, so they are large early and small later.
Test it against the curve. From one edition to two is a doubling of cumulative editions and the median gains 1.1, from 4.1 to 5.2. From two to four is the next doubling and the median gains 0.9, from 5.2 to 6.1. Decaying gains on successive doublings, which is what the model predicts.
From four to eight is the next doubling. On the learning story it should return something small and positive. It returns minus 0.9, from 6.1 down to 5.2. The learning curve describes two doublings on this floor and then stops describing anything, and whatever is happening on the right hand side of the curve needs a different explanation.
Why is the right hand end of the curve measured on survivors?
The explanation most people reach for is that long tenured exhibitors get complacent. Some of that is true. Most of the shape is an artefact of who is left.
Every point on that curve is computed on a different set of companies, and the sets are nested. The companies at edition four are a subset of the ones that started, filtered by four rounds of a decision about whether the show was worth repeating. The filter is not random. The stands that did badly are the ones that left.
Work through what that does. Take the 118 companies that first exhibited at the 2021 edition. Of those, 71 came back for a second edition, 48 for a third and 39 for a fourth.
The 4.1 at edition one is computed on all 118. The 6.1 at edition four is computed on 39 companies that had already survived three rounds of selection. Those are not comparable populations and the difference between them is being read as improvement.
Go back to the 2021 file and recompute the first edition median using only the 39 companies that made it to a fourth edition. On this floor it comes out at 5.4, well above the 4.1 for the full cohort, because the companies that were going to keep exhibiting were the ones that had done reasonably well the first time.
Now the same comparison, honestly. The apparent gain from edition one to edition four is 6.1 minus 4.1, which is 2.0. The gain for the companies that actually lived through it is 6.1 minus 5.4, which is 0.7. Divide: 0.7 over 2.0 is 0.35. About a third of the apparent improvement is real learning and about two thirds is the weak stands having left.
The same mechanism runs the other way at the far end. Companies with nine or more editions include a long tail that has been coming for fifteen years out of habit, in a category that has moved on, and nothing removes them because nobody has ever given them a reason to stop.
Cohort curves, which cost more and tell the truth
The correction is to stop drawing a cross section and start following cohorts.
Fix a cohort by the edition at which a company first exhibited. Follow that cohort forward and plot its median at each subsequent edition, using only companies from that cohort. You now have a curve where the population changes for one reason you can see, attrition, and you can print the surviving count at every point.
The stricter version is a balanced panel: restrict the whole curve to the companies present at every edition from the first through to the last one you are plotting. That removes attrition entirely and produces the cleanest read of what actually happens to a stand over time. It also shrinks fast. On the numbers above, a balanced panel through four editions has 39 companies in it, and by seven editions you may have a dozen, at which point the median is being set by two or three stands and should be labelled as such.
Run both. The cohort curve tells you what happened to the group. The balanced panel tells you what happened to a stand. The gap between them is the size of the survivorship effect, and it is the number worth quoting when somebody puts the loyalty slide back up.
Plot two or three entry cohorts on the same axes and the show's own history appears. A cohort that entered at a strong edition and one that entered at a weak one behave differently for years afterwards, which is a finding about your recruitment rather than about exhibitor behaviour.
Sentiment and productivity move on separate instruments
The awkward part of the tenure curve is that the exhibitors on its declining limb are usually happy.
Explori and UFI's 2025 Channel Insights Report, drawing on more than 3,000 events between 2017 and 2025, tracked exhibitor sentiment across that period and found overall satisfaction moving from 3.55 in 2019 to 3.80 in 2022, likelihood of return from 3.88 to 4.08, importance from 3.68 to 3.93, and average net promoter score up 27 points from before the pandemic to after it. Those are survey measures of how exhibiting felt, collected from exhibitors, and they moved sharply upward across the population.
Captured lead volume per square metre is a different instrument measuring a different thing, and the two can move in opposite directions on the same floor without either being wrong. A long tenured exhibitor who comes to meet existing customers, hold a distributor dinner and hold their position in the aisle can be highly likely to return and steadily less productive on any scan based measure, year after year, permanently.
Two mechanisms do most of the work on the declining limb, and both are checkable. The stand's returning visitor share rises as its customer base gets used to finding it there, so a growing fraction of its scans are people it already knows, and what those repeat scans are worth is E13's ground. The stand also stops being a launch venue: a company in its ninth edition rarely has a new product to put on it, and the pre-show promotion that comes with a launch is what drives strangers to a stand in the first place.
What should exhibitor tenure decide, and what should it not?
Tenure should not be used as a proxy for exhibitor quality in any analytics output. The curve above is enough to show why: on this floor, a stand in its second edition outperforms a stand in its ninth on the metric the sales team cares about most.
Tenure is a perfectly good input to space allocation, and that is a different question answering to different objectives. A priority points formula rewards commitment, smooths the floorplan and gives the sales team something to sell against, and none of those objectives requires long tenured exhibitors to be more productive. The system itself is F13's subject and designing the formula so it rewards the behaviour you want is F15's, and the honest position is that tenure earns points for being tenure, not for a lead yield claim the data does not support.
The commercially useful reading of the curve is the middle of it. Stands in their third to fifth edition are the most productive on the floor and the least attended to, because the sales team worries about first year stands and about large long standing accounts, and the group in between is quietly doing the best work in the building.
The limit
Editions exhibited is not the same as years, and the two get conflated the moment anyone talks about the curve out loud. On a biennial show, edition four is eight years in. On a show that added a second annual edition three years ago, tenure counts are not comparable across companies. Say which unit you used and use it consistently.
The curve is specific to your floor. The peak at four editions on this example is an artefact of this example, and importing a shape from somebody else's show, or from this post, would be exactly the error the whole method exists to avoid. Draw your own.
Tenure is also confounded with size, tier and position, because long tenured exhibitors hold better space. Dividing by contracted square metres removes one of those and leaves the others, and if your tiers are unevenly distributed across tenure then standardise the way E27 sets out before reading anything into the curve. The tenure field itself is unreliable for the reasons E28 covers on first time stands, so run the long tenured list past your sales team before publishing. Reading a single stand's movement between two editions is a separate calculation, and indexing it against the show's own change is E29's method.
Add an editions exhibited column to last edition's exhibitor export, compute the median unique leads per square metre for each tenure value, and print the company count beside every point. Then take one entry cohort, find the companies from it still exhibiting four editions later, and recompute that cohort's first edition median restricted to those survivors. The distance between that number and the one you started with is how much of your tenure curve was never there.
Questions people ask about exhibitor tenure and lead volume
- Do long standing exhibitors capture more leads than newer ones?
- Not per square metre. On the floor worked through here the peak sits at four editions, at 6.1 unique leads per square metre, while companies with nine or more editions return 4.9. A stand in its second edition outperforms a stand in its ninth, which is the opposite of what a loyalty chart built on totals shows.
- Why is a cross-sectional tenure curve misleading?
- Because every point is computed on a different set of companies, filtered by rounds of a decision that is not random. Recomputing the first edition median using only the companies that survived to a fourth edition raises it from 4.1 to 5.4, which means about two thirds of the apparent improvement is the weak stands having left.
- How do you draw an honest exhibitor tenure curve?
- Follow cohorts rather than a cross section. Fix a cohort by the edition at which a company first exhibited, plot its median at each subsequent edition using only that cohort, and print the surviving count at every point. A balanced panel restricted to companies present at every edition is stricter still, and it shrinks fast.
Related reading
- First time exhibitor performance is worse than you think and measurable from day one
- Returning exhibitor lead trends across editions and how to separate show effects
- Designing a priority points formula that rewards the behaviour you actually want