Post show reporting stakeholders and why one report cannot serve all of them

Four people asked the data team for the post-show numbers in the same week. The exhibition sales director wanted them by Thursday because renewal calls started Monday. The marketing lead wanted them by channel and by day. The finance director wanted to know why the revenue figure in the draft did not match the ledger. The chief executive wanted one page.

The data team, reasonably, tried to build one document that answered all four. It took five weeks, ran to sixty pages, and none of the four used it. The sales director had already made her own spreadsheet.

That failure is structural. The four readers want different grains, different time windows, and different lead metrics, and a document that averages across those requirements produces something with no reader at all. Deciding what belongs in the report and what gets cut is the prior question, and it gets answered once for the whole family of documents.

Who reads a post show report, and what does each one want?

Sales wants material for a renewal conversation. The grain is the account. The useful facts are what this exhibitor got, how that compares to the stands around them, and what changed since last edition. A show-level attendance figure is close to useless in that conversation, because the exhibitor's response is always that the crowd never reached their aisle. The pack that goes to the exhibitor itself is a fifth document with an audience outside the building, and the exhibitor report pack covers what belongs in it.

Marketing wants to know which acquisition routes worked. The grain is the channel and the campaign, the window starts months before the show opens, and the metric that leads is registrations, with attendance appearing as a conversion denominator. The Freeman Trends Report published in 2025 reports that only about 30 per cent of first-time attendees return to an event, and that 51 per cent of attendees name successful networking as a reason to come back. A marketing cut that cannot separate first-timers from returners cannot act on either finding.

Finance wants the number that will survive an audit of the management accounts. The grain is the accounting period, which almost never matches the show. Revenue recognised, costs booked, accruals for services invoiced after close. UFI's Global Exhibition Barometer, 36th edition, published in January 2026 from 378 companies across 57 countries, has 31 per cent of respondents reporting an annual increase in operating profit of more than 10 per cent for 2025, and the finance reader's whole interest is whether this show's contribution to that line is what was forecast.

The board wants the trend. The grain is the show and the window is five editions. The four components of the CEIR Index, published through IAEE, are the right shape for that cut: net square feet of exhibit space sold, professional attendance, number of exhibiting companies, and gross revenue, each against its own history.

Can one warehouse serve four different cuts?

What works is one warehouse and four published cuts, with the discipline sitting in how precise you are about what each cut is allowed to change. Four teams building four extracts from four systems gives you the same four documents and no way to reconcile them.

Three things vary between cuts: the grain, the time window, and the lead metric. Nothing else. A cut may aggregate the same rows differently, restrict them to a different window, and put a different metric at the top. A cut may never apply a different filter to the base population, and it may never rename a metric.

That distinction is what keeps four documents reconcilable. The 2025 Channel Insights Report, produced by Explori with UFI and supported by SISO across more than 3,000 events surveyed between 2017 and 2025, tracks exhibitor sentiment on four named measures and reports exhibitor net promoter score up 27 points since 2019. The sales cut and the board cut can both carry that score, at account grain and at show grain, because it is one measure with one definition, aggregated twice.

Kimball and Ross made the general version of this argument for dimensional models in 2013, and the report specification that keeps editions comparable puts it to work on the artefact. The point that carries over here is that shared definitions are the mechanism that lets separately built outputs be compared, and they have to be managed once rather than agreed four times.

The same edition, four top lines, all correct

Take one edition. The warehouse holds 31,600 unique badges scanned at entry, excluding exhibitor staff and contractors, across three show days. That is the attendance definition, fixed, and every cut inherits it. Fixing it once is what makes the attendee analytics in four separate documents comparable to each other.

The board cut leads on 31,600, alongside 1,240 exhibiting companies and gross revenue of 9,840,000, each with five editions of history behind it.

The sales cut runs at account grain and reports the audience each exhibitor's category actually saw. Summed across the eleven product categories, that comes to 34,900, which is 3,300 higher than the show total. The gap is not an error. 3,100 attendees registered interest in two categories and 100 in three, giving 3,100 plus 200, which is 3,300 double counts. Each account gets a category figure it can use, and the appendix says in one line that category figures sum above the show total by design.

The marketing cut runs at channel grain over a window that opens 190 days before the show. It leads on 44,100 registrations across nine channels, with attendance as the conversion denominator: 31,600 divided by 44,100 is 71.7 per cent. That cut also carries 2,900 registrations that arrived in the four days after the finance cut-off, which is why marketing's registration total will never tie to a figure computed from the finance close.

The finance cut runs at accounting period grain and leads on 9,840,000 of gross revenue, split across two months because the show straddles a period end. Attendance appears once, as context, and is labelled as the board cut's figure.

Four documents. One base population. Every difference explainable in a sentence.

The definitions appendix is the contract between the cuts

One appendix, shared by all four cuts, published as a section of each. It carries every metric that appears anywhere in the family, and for each one: the population, the filters, the time window, the source system, and the cut or cuts it appears in.

The appendix is what makes the sales director's 34,900 defensible. Without it, the first person who notices that sales is quoting a bigger audience than the board has found a scandal. With it, they have found a footnote, which is what it is.

Writing a good metric definition has a method of its own and D5 sets it out. The appendix here is the distribution mechanism, and the only rule specific to it is that it appears in full in every cut. Not a link. Not a reference to a shared drive. In the document, every time, because the cut will be forwarded to somebody who does not have access to the drive and that person will still need to know what attendance means.

The four cuts have four deadlines, and that is the real constraint

Grain and window get all the attention in a design like this, and the thing that actually decides whether it survives is timing.

The sales cut is needed within about five working days of close, because renewal conversations start while the show is still fresh and an account manager with no data is an account manager improvising. The finance cut cannot exist until the accounting period closes, which is usually three to five weeks later. The board cut goes to whichever meeting comes next, which might be six days or ten weeks. The marketing cut is the least urgent and, in practice, the one most often skipped entirely.

Those deadlines are irreconcilable, and trying to publish all four together means the sales cut waits for finance. So publish them separately and stamp each one with the data as-of date on its cover. The sales cut says its figures are as of day five and will be superseded. The finance cut says it is final. A reader who knows which of the two they are holding will not treat a preliminary number as settled, and D34 works through how to publish preliminary and final figures without spending credibility each time.

The request I would refuse

Somebody will ask for a single number that means the show did well. Usually it is framed as wanting one headline metric so everyone is talking about the same thing.

I would not give it. A composite is easy enough to build and it hides which of its components moved, when the whole reason four readers exist is that they act on different components. The CEIR Index publishes a Total Index and also publishes its four parts, and the parts are where the information is. In the second quarter of 2025 CEIR had attendees 3.7 per cent below their 2019 level, net square feet 4.9 per cent below, exhibiting companies 8.8 per cent below, and real revenues 15.6 per cent below. A single composite for that quarter would have told you the industry was down and told you nothing you could act on.

What I would give instead is the shared summary page, five numbers with prior-year figures, appearing identically at the front of all four cuts and built the way a post event report executive summary is built. Everyone is then looking at the same numbers, and each reader turns to a different page after that.

Where this stops

Four cuts is four documents to maintain, and the maintenance is the failure mode. In the second year, somebody adds a filter to the sales cut because an account manager asked for it, does not update the appendix, and the family quietly stops reconciling. Nobody notices until a board meeting.

The only defence I have seen work is a reconciliation check that runs before publication and compares each cut's base population count to the warehouse figure, with any difference having to be explained by a named rule. If the sales cut's underlying badge count is not 31,600, publication stops. That check is cheap to build and it is the thing that gets dropped when the report is late.

The second limit is that this design assumes a warehouse exists. Plenty of organisers are assembling the post-show report from four system exports joined by hand in a spreadsheet, and for them the four cuts will genuinely be four different numbers, because the joins get done slightly differently each time. D33 covers where those joins break.

Take the last post-show report you published and count the distinct readers who acted on it. If the honest answer is one or two, you have been writing four documents' worth of material for a fraction of its audience, and the cheapest fix is to ask the readers who ignored it what grain they needed.