Next best show recommendation for exhibitors and how to score candidates
A next best show recommendation scores each account against each show it does not hold, on category fit, buyer overlap with the shows it already does, geography and budget capacity implied by current spend. Rank the candidates, then ship the components as a sentence a rep can say, because a score alone gets ignored.
A rep gets a cross-sell list of 640 accounts and a fortnight to work it. Every pitch she makes is for the flagship. The flagship suits almost none of those accounts, and it has 900 square metres unsold in hall four, and the show director raised that in Monday's meeting.
Two weeks later the list has produced four meetings and one signed contract. The programme gets written off as a cross-sell problem. What actually happened is that nobody told the rep which show to pitch to which account, so she used the only signal available to her, which was internal inventory pressure.
The default recommendation is whatever has space left
Inventory-led recommendation has a specific failure pattern and it repeats across every portfolio I have seen it in.
The shows with unsold space are usually the ones with a weaker proposition this cycle, so you are pitching your least persuasive product. The pitch lands on accounts chosen for their spend on a different show, so the category argument is thin. And the rep, who is not stupid, works out after eight calls that the list is not really a list, and starts working her own accounts instead.
A recommendation has to answer a narrower question than "who might buy something". It has to answer: for this specific account, which of my other shows is the one most likely to convert, and what is the sentence the rep says first.
That is scoreable. Not perfectly, and not without data you probably have to go and get, but scoreable in a way that beats hall four.
What should a next best show recommendation score?
Score every account against every show it does not currently hold, sitting on top of the resolved parent view that G25 builds. Four components carry most of the signal, and each one is measuring a different reason a deal dies.
Category fit. Does this exhibitor's product set appear in the candidate show's exhibitor taxonomy, and how much of it. This is the component sales teams already reason about, and it is the one they overweight.
Buyer overlap. How much of the candidate show's audience resembles the buyers this exhibitor actually met at your show. Not the two shows' audience overlap in general, which is a portfolio statistic. The overlap for this account, computed from the badges their staff scanned.
Geography. Can they service the market the candidate show sells into. An exhibitor with no distribution in a country is not going to take a stand there, however well the category matches.
Budget capacity. Can they afford the smallest realistic package on the candidate show, given what they currently spend with you. What they spend with other organisers is a separate estimate and belongs with G27.
Weights are a judgement and should be written down and argued about once rather than tuned quietly. I would start at category fit 0.35, buyer overlap 0.30, geography 0.20, budget capacity 0.15, and I would defend the buyer overlap weight hardest, because it is the component that distinguishes a recommendation from a guess and the one most teams do not compute at all.
Scoring one account against three candidates
Take a fastener manufacturer holding 24 square metres on your industrial automation show. Space at 1,300 per square metre is 31,200, plus 4,800 of sponsorship, so 36,000 with you this cycle. The portfolio has nine shows. Five of the eight candidates clear a category fit floor of 0.30. Here are three of them.
Buyer overlap needs a denominator that is comparable across accounts, so divide by the highest exhibitor-to-show overlap seen anywhere in the portfolio, which is 0.24. Across the last two editions this exhibitor's staff scanned 1,180 unique verified buyers.
The packaging and processing show. Category fit 0.62, because two of their three coded product categories appear in that show's taxonomy. Of the 1,180 buyers they scanned, 214 also attended the packaging show, so raw overlap is 0.181 and the normalised score is 0.181 divided by 0.24, which is 0.75. Same country, so geography 1.00. The entry stand is 18 square metres at 1,450, or 26,100, against current spend of 36,000, so budget capacity caps at 1.00. Weighted: 0.35 times 0.62 is 0.217, plus 0.30 times 0.75 is 0.225, plus 0.20, plus 0.15. Total 0.79.
The flagship manufacturing show. Category fit 0.88, the best of the five. Overlap is 96 of 1,180, or 0.081, normalised to 0.34. Geography 1.00. The smallest stand anyone actually takes is 36 square metres at 2,100, or 75,600, so budget capacity is 36,000 divided by 75,600, which is 0.48. Weighted: 0.308 plus 0.102 plus 0.200 plus 0.072. Total 0.68.
The regional show in a neighbouring country. Category fit 0.80. Overlap 106 of 1,180, normalised to 0.375. Geography 0.35, because their distribution there runs through a single agent. Entry at 12 square metres and 1,100 is 13,200, so budget capacity caps at 1.00. Weighted: 0.280 plus 0.113 plus 0.070 plus 0.150. Total 0.61.
The packaging show wins on 0.79 with the second-worst category fit of the three. It wins because a fifth of the buyers this exhibitor already meets are also in the packaging hall, and because the entry price is inside what they already spend. The flagship, which is the show every rep in the building would have pitched, comes second on a component the exhibitor cannot act on.
Shows have an order, and it is not the same for everyone
Scoring each candidate independently treats the eight shows as eight unrelated products. In most portfolios they are not.
Li, Sun and Wilcox set out the machinery for this in the Journal of Marketing Research in 2005, working on consumer banking. Their structural multivariate probit model handles products that are naturally ordered, where customers move along what they call a financial maturity continuum and acquire in sequence. The useful part for an exhibition portfolio is the framing: what a customer already holds, and how long they have held it, changes which product is next, and a model that ignores the sequence recommends the wrong one.
Exhibition portfolios have exactly this shape when the shows differ in size and commitment. A regional show at 12 square metres is an entry product. The flagship at 36 square metres with a two-year waiting list is not something an account buys second. Pitching it to a two-edition exhibitor on a small regional stand is asking them to skip three steps, and the rep will be told the price is too high when the real objection is that the account has not yet had the year of evidence it needs.
The counter-case matters too. A large multinational entering your region for the first time can start at the flagship, because the sequence in their head is not your sequence. Which is why the order should come out of the data instead of out of a portfolio strategy deck.
How do you read the buying order off your own file?
You already hold the evidence. Take every parent company on two or more shows and look at which show they bought first and which they bought second.
Say 528 parents hold two or more shows across the portfolio. Of those, 212 started on the industrial automation show. Their second purchase splits: 96 went to packaging, 44 to the flagship, 31 to the regional show, and 41 across the rest. So 45 per cent of the accounts that ever started where this exhibitor started went to packaging next, which agrees with the score and gives the rep a sentence with a number in it.
Where the transition counts and the component score disagree, look at the disagreement rather than averaging them. A candidate show with high category fit and almost no historical transitions is either a genuine gap in your selling or a pairing that has been tried and quietly fails.
One reality check on the whole exercise. CEIR's 2026 Marketing Spend Decision Report found 28 per cent of exhibitors expect to add one or more shows to their programme, against 47 per cent expecting to hold the number they already run. That 28 per cent is measured across everyone's shows, so it is a ceiling on what your own portfolio can claim. On a list of 640 accounts, the number in market for an additional stand anywhere is unlikely to be much above 180, and a recommendation engine that ranks well is mostly buying you a better first 180 calls. Sell that second stand on its own revenue, because the retention benefit on the first show is far smaller than the raw comparison implies, which is G28's finding.
The reason line is the whole deliverable
A ranked list with a score on it gets ignored. A rep who is handed "packaging show, 0.79" has no idea what to say and will not use it twice.
Ship the components as a sentence instead. Something close to: 214 of the 1,180 buyers your staff scanned at automation also attended packaging, entry space there is 26,100 against the 36,000 you spend with us, and 96 of the 212 accounts that started where you started went to packaging next.
Every clause in that is a number from a table somebody can open. If the rep gets challenged on any of it, the answer exists. And when the account says no, the reason they give maps onto a component, so the objection can be logged against geography or budget capacity and the weights can be revisited with evidence at the end of the cycle.
Where this stops
The buyer overlap component is the strongest signal in the score and it needs the thing most portfolios do not have, which is attendee identity resolved across shows. Without it you are matching on email string equality and undercounting overlap badly, which pushes every recommendation in your renewal and cross-sell scoring back towards category fit and back towards the flagship.
The category taxonomy is your own opinion written down. Two shows can look adjacent in it and draw completely different buyers, and the taxonomy will never tell you that because it was built by the marketing team of one show at a time.
The honest limit is bigger than either. This score predicts fit. It says nothing about whether the account has budget this year, whether the person who decides has changed, or whether they were burned by a bad hall position in 2023 and have not mentioned it. Fit is a filter that stops you wasting the good calls. The conversion evidence only arrives after you have called, and the first cycle of a recommendation engine is a data collection exercise you should budget for as one.
Take one show, pull the exhibitors who hold only that show, and join their scanned buyers against the verified attendee list of one other show in the portfolio. Count the overlap per exhibitor and sort descending. The top of that list is a call sheet you can use this week, and it took one query.
Questions people ask about next best show recommendation
- What should a next best show recommendation be scored on?
- Four components carry most of the signal. Category fit against the candidate show's exhibitor taxonomy, buyer overlap computed from the badges this exhibitor's staff actually scanned, geography meaning whether they can service that market, and budget capacity against the smallest realistic package. Weights of 0.35, 0.30, 0.20 and 0.15 are a defensible starting point.
- Why do cross sell reps default to pitching the flagship?
- Because nobody tells them which show suits which account, so they use the only signal available, which is internal inventory pressure. The shows with unsold space are usually the weaker proposition that cycle, the category argument is thin, and after eight calls the rep works her own accounts instead. A scored recommendation replaces that signal.
- How many exhibitors are actually in market for another show?
- Fewer than a cross sell list implies. CEIR's 2026 Marketing Spend Decision Report found 28 per cent of exhibitors expect to add one or more shows, against 47 per cent holding the number they already run. That 28 per cent covers every organiser's shows, so on a list of 640 accounts the ceiling is around 180.
Related reading
- Cross sell across an event portfolio starts with one account view
- Measuring exhibitor share of wallet when you only see your shows
- Multi show exhibitor retention and whether the second show protects the first