Repeat visitor scans across a show floor and what they say about intent

At a review after a four day exhibition, an exhibitor's marketing lead asked the show director a question that sounded like small talk. Out of everyone we scanned, how many came back?

The show director did not know. The exhibitor report had unique leads, total scans, and a grade breakdown from the qualifying questions. It had no count of repeat visitor scans at all, because everything in it had been aggregated to the exhibitor level and the day column had been dropped on the way.

That question is the best one an exhibitor asks all year, and answering it takes a group-by rather than a project.

Why count distinct scan days instead of scans?

Two scans an hour apart mean very little. The visitor was standing near the stand, somebody read the badge, somebody else read it again, and depending on the stand design that may have been two conversations or one person walking past a kiosk twice. The suppression rule for near duplicate scans is E11's, and this measure sits on top of whatever that rule leaves behind.

Two scans on separate days mean something quite specific. The visitor left the building, slept, decided how to spend the next day out of an exhibition full of alternatives, and walked back to that stand. Nobody does that by accident.

So the measure is distinct scan days per badge per exhibitor. For each badge scanned by a given exhibitor, count the number of separate show days on which at least one scan occurred, then report the share of that exhibitor's unique badges reaching two or more.

The implementation is trivial and the discipline is where teams go wrong. The date must be derived from the show day, because a late night networking session that runs past midnight will otherwise split across two calendar dates and manufacture returns that never happened. Define show day boundaries once against the published opening hours and apply them everywhere. What those timestamps say about dwell and stand rhythm is E2's subject.

Which denominator makes the number honest?

Here is the part that changes the number by a large factor.

If you divide multi-day badges by all unique badges, you are including every visitor who only ever attended one day of the show. Those people could not have returned. They were not in the building. Including them in the denominator means the measure partly reports how many day passes your registration team sold, which has nothing to do with the exhibitor.

The correct denominator is the badges that were physically present on more than one show day, which you know from entrance scans, session scans or any other floor-wide capture you run. Restrict the population to visitors who had the opportunity to return, then measure how many of them chose to return to that particular stand.

Both figures are worth having. The unrestricted one answers what share of my leads came back. The restricted one answers what share of the people who could come back did. The second is the intent measure, and it is the one that compares fairly between an exhibitor near the entrance in a hall that empties on day three and an exhibitor at the far end of a hall hosting the closing keynote.

Working it through on 980 badges

One exhibitor, four day show, 980 unique badges scanned.

Group their scan rows by badge and count distinct show days:

• one day only: 896
• two days: 71
• three days: 11
• all four days: 2

That adds to 980. Badges reaching two or more days: 71 plus 11 plus 2, so 84. Against all unique badges that is 84 divided by 980, or 8.6 per cent.

Now bring in the entrance data. Of those 980 badges, 588 were on site on more than one show day. Recomputed against the population that could have returned, the figure is 84 divided by 588, or 14.3 per cent.

The difference is not cosmetic. Reported one way the exhibitor is told that fewer than one lead in eleven came back. Reported the other way it is closer to one in seven of the people who had the chance. Publish both, label both, and put the eligible population count on the line so nobody has to guess which denominator produced which number.

It is also worth separating the within-day repeats so the exhibitor can see the contrast. In this file, 210 badges were scanned more than once on a single day and never appeared on another one. So 21.4 per cent of the exhibitor's leads were scanned twice in a day, and 8.6 per cent came back on a different day. Those two populations behave differently in follow-up and treating them as one bucket of repeat scans loses the distinction entirely. How many scans a stand records per unique badge is a separate diagnostic with its own healthy range, and the scan to unique lead ratio sets that range out in E12.

Why the second day is evidence and the second scan is not

The CEIR Industry Insight Report on improving lead quality and sales conversion, written by Jefferson Davis of Competitive Edge and published by CEIR in 2019, is blunt about what a scan on its own is worth. If booth staff believe that scanning as many visitor badges as possible creates leads, Davis treats that behaviour as a problem in itself. A badge read is a record that a scanner fired, and it carries no information about why.

A return visit is a cheap and honest proxy for shortlisting. A buyer with a real requirement walks the floor on day one, narrows to a handful of suppliers, and goes back to those with the specific questions they could not ask in an aisle. A buyer collecting brochures does not.

CEIR's 2015 study on Exhibitor ROI and Performance Metric Practices found that among exhibitors who qualify leads, whether before, during or after the exhibition, 39 per cent of leads turn out to be qualified against their own criteria. The multi-day flag is not a substitute for that qualification, and it is available for every badge including the ones where the staffer skipped the qualifying questions, which on a busy stand is most of them. Where an exhibitor has both, the interesting cross-tab is grade against multi-day, because a stand whose A grades and whose returners barely overlap has a qualification process measuring something other than intent. How the grades themselves get defined belongs to E6.

Reading it across the whole floor

Compute the same figure for every exhibitor and you get a floor-wide distribution that is more useful than any single stand's number.

Two patterns show up reliably. Exhibitors selling long-cycle capital equipment run high multi-day shares, because their buyers bring a colleague back. Exhibitors selling consumables and accessories run low ones, because a single conversation is enough to decide. That is a category effect and it means the figure has to be read within category, which E26 covers properly.

The other pattern is positional. Stands adjacent to a feature area, a theatre or a catering point pick up incidental repeat scans that have nothing to do with intent. If your multi-day share correlates with distance to the nearest coffee stand, that is worth knowing before anyone builds a renewal argument on it.

I would publish the exhibitor's own share, the median across their category, and the eligible population count. Three numbers, one line of the report. Anything more and stand managers stop reading.

Where this stops

The measure needs a show long enough for returning to be possible, and it needs floor-wide presence data to build the honest denominator.

On a two day show the whole thing collapses to a binary that most exhibitors could estimate by eye, and the eligible population is often a small minority because a large share of visitors buy a one day ticket. Below about three days I would report the raw multi-day count and skip the percentage, because a percentage computed on 80 eligible visitors invites more confidence than it deserves.

The harder limit is that a return visit and a good lead are different things, and the correlation between them is a hypothesis you have not tested on your own floor. A visitor may come back because your stand had the only free seating in the hall. A serious buyer may come back on day two, find the right person absent, and never get scanned at all, which makes the strongest lead in the file look like a single-day walk-past. The measure sees badge reads, and badge reads are a partial record of what happened on the stand.

That is a reason to report it as a behaviour signal with the eligible denominator attached, and to keep it out of any composite score in your exhibitor analytics until an exhibitor has told you it tracks their pipeline.

Take your last four day show, join the exhibitor scan table to the entrance scan table on badge id, and compute two columns per exhibitor: distinct scan days per badge, and days present on site per badge. The gap between the two is the number you have been missing, and the query is short enough to write before lunch.