Skip to content

Net promoter score for events and what one point of movement means

Attendee analyticsUpdated 2026-08-238 min read

In short

Net promoter score for events is the promoter share minus the detractor share on a nought to ten recommend question, banded nine to ten, seven to eight and nought to six. On 900 responses one point equals nine people moving from detractor to passive, so read small movements as arithmetic rather than as sentiment.

The show NPS came in at 29 this year against 26 last year. It is on slide four with an upward arrow and a sentence about improved attendee sentiment.

Three points on the response base that show ran is about twenty-seven people, which is a thin foundation for a sentence about sentiment. Net promoter score for events asks one question and returns one number, so every design decision that could have been argued about has been pushed into two places: who you asked, and when. Both are invisible in the output.

The measure and where it came from

Reichheld published the argument in Harvard Business Review in December 2003, under the title The One Number You Need to Grow, drawing on two years of work testing survey questions against actual customer behaviour, purchasing patterns and referrals, and then against company growth rates. The claim was that willingness to recommend tracked growth better than the batteries of satisfaction questions companies were already running.

The mechanics are fixed by convention rather than by anything statistical. A nought to ten scale on likelihood to recommend. Nine and ten count as promoters, seven and eight as passives, nought to six as detractors. The score is the promoter percentage minus the detractor percentage, so it runs from minus 100 to plus 100 and the passives never enter the arithmetic at all.

That last property is the one people forget. A third of your respondents can be discarded from the calculation entirely, which is a lot of information to throw away for the convenience of a single number.

The banding is also asymmetric in a way that surprises organisers the first time they see the distribution. Seven of the eleven available scores count against you, two are neutral and only two count in your favour. An attendee who thought the show was decent and rated it a seven contributes nothing at all, while an attendee who thought it was fine and rated it a six subtracts. On a trade show audience, where a six often means the week was useful without being memorable, that boundary does a great deal of work that nobody chose.

What does one point of movement actually mean?

Take 900 responses: 369 promoters at 41.0 per cent, 423 passives at 47.0 per cent, and 108 detractors at 12.0 per cent. The score is 41 minus 12, which is 29.

Now move one person. A detractor who becomes a passive changes the detractor share by 1 in 900, or 0.11 percentage points, so the score rises by 0.11. A detractor who becomes a promoter changes both shares, so the score rises by 0.22.

Invert it. One point of NPS on this base costs you nine detractors converted to passives, or about four and a half converted all the way to promoters. The three point move on slide four is twenty-seven people out of nine hundred, and it is well inside the range that resampling the same audience would produce. How wide that range actually is comes out of the standard error of a difference of proportions, and it belongs with putting an NPS confidence interval on your show score in D29.

The useful habit is to convert every reported movement into people before it goes on a slide. A five point improvement on 900 responses means 45 people who would have scored you a six now scoring a seven. That is a specific, plausible operational target. Five points described as improved sentiment is not.

Two shows with the same score and different problems

Here is why the score alone is thin. Two shows in the same portfolio both report 29.

The first has 41 per cent promoters, 47 per cent passives and 12 per cent detractors. The second has 55 per cent promoters, 19 per cent passives and 26 per cent detractors. Both subtract to 29 and the two audiences have almost nothing in common.

On 900 responses the second show has 234 people actively willing to tell colleagues to stay away, against 108 at the first. It also has 495 enthusiasts against 369. It is a polarising event, which is a normal and sometimes deliberate position for a show to occupy, and the single number conceals it completely.

Report the three shares every time, next to the score, in the same line. Promoters 41, passives 47, detractors 12, score 29, base 900. It costs one row and it makes the number readable.

Ask visitors and exhibitors separately

They are different populations buying different things, and the temptation to combine them into one show score should be resisted.

Explori's documentation on how it builds event benchmarks splits scores by audience type, listing visitor, exhibitor and non-attendee as separate benchmark populations, alongside four standardised measures that include net promoter score, overall satisfaction and likelihood of return. The split exists because the underlying distributions differ, and because an organiser comparing against a benchmark needs to know which population produced it.

Inside your own show the reason is sharper. A combined score is a weighted average of two very different numbers, and the weights are the response counts, which change every edition for reasons that have nothing to do with the event. If exhibitor staff reply at a higher rate this year because you sent the survey from the account manager's address, the combined score moves and nobody will be able to say why. The exhibitor side has its own question design problem, which is that satisfaction answers arrive when you meant to ask about outcomes, and that sits with the exhibitor satisfaction survey in D28.

Two scores, two bases, two rows. Then, if a single portfolio number is genuinely required, compute it from verified attendance shares rather than from response counts, which is the post-stratification exercise in weighting survey results in D26.

Fixing the wording and the clock

Two things have to be frozen for a year-on-year comparison to mean anything.

The wording, exactly. Recommend the event to a colleague in your industry is a different question from recommend the event to a friend, and both are different from would you tell a colleague to attend next year. Changing three words changes the series and nothing in the output records that it happened.

If you want an externally comparable form, borrow one that is already standardised. The UFI and Explori Global Exhibitor Insights report of November 2017, built on 1,040 shows across more than 40 countries, lists the four questions its participating events all asked in a common format, and gives the recommend question as how likely respondents are to recommend the event to colleagues in the industry. Adopting a published wording costs you nothing and buys you the option of benchmarking later.

The clock, to a fixed number of days after close. A score collected at 24 hours and a score collected at day ten are not comparable, and if the send date drifts by a week between editions you have introduced a movement you cannot separate from a real one. Pick a number of days, write it in the methodology note, and hold it.

A third item belongs on the frozen list and gets overlooked because it feels administrative. The scale rendering. A nought to ten row of radio buttons, a slider, and a dropdown produce different distributions from the same respondents, and the ends of a slider are harder to hit than the ends of a button row. When your survey platform redesigns its question types, which it will, your NPS series can move without a single attendee changing their mind. Screenshot the question as rendered each year and keep the image with the methodology note, because in three years nobody will remember what the 2026 version looked like.

Should an organiser use NPS at all?

I would keep it, with reservations, and the reason is unglamorous. It is the one measure your exhibitors, your board and your benchmarking provider all already understand, and a number everybody can read badly is often more useful than a better number nobody reads.

What I would stop doing is treating it as the headline. The score belongs in the pack next to overall satisfaction, likelihood to return and objective achievement, all four with their bases, and none of them promoted to a single figure of merit. Where NPS earns its place is as a segmentation tool: pulling the detractors out by badge type, region and first visit status tells you where the problem sits, which is far more actionable than the aggregate. That segmentation runs off the same attendee analytics joins as the rest of the post-show pack.

Where this stops

The score measures stated intention to recommend at a moment a few days after the show. Whether recommendations actually happen, and whether they bring anyone through the door, is a separate question that this instrument does not touch.

There is a subtler limit in the banding. A seven is a passive and a six is a detractor, which means the entire measure hinges on where respondents place the boundary in their own heads, and that varies by culture in ways that are well known to anyone who has run the same survey across regions. A show whose audience shifts towards markets where respondents avoid extreme ratings will see its NPS fall without anything changing on the floor. Comparing your Asian edition against your German one on raw NPS is a mistake that will produce a confident, wrong conclusion.

The last limit is the one the arithmetic already showed. On a base under about a thousand, movements smaller than five points are mostly sampling noise, and reporting them as findings trains your board to react to nothing.

Take the last three editions of one show and rebuild the NPS series with the three shares and the response base shown for each. If a movement you previously reported turns out to be fewer than twenty people, strike the sentence that explained it and write down what the base would need to be for that movement to mean something.

Questions people ask about net promoter score for events

How is net promoter score calculated for an event?
Ask attendees how likely they are to recommend the event on a scale of nought to ten. Count nine and ten as promoters, seven and eight as passives, and nought to six as detractors. Subtract the detractor percentage from the promoter percentage. With 41 per cent promoters and 12 per cent detractors the score is 29, and the passives never appear in the arithmetic.
What does a one point change in event NPS represent?
It depends entirely on the response base. On 900 responses, one point is nine people moving from detractor to passive, or roughly four and a half people moving from detractor to promoter, because a full reclassification counts twice. On 300 responses the same point is three people. Always publish the response count next to the score.
Should visitors and exhibitors get one combined event NPS?
No. They are different populations buying different things, and blending them produces a number whose value depends on the ratio of respondents in each group, which changes every edition. Explori's benchmarking documentation splits its event scores by audience type for the same reason. Report two scores with two response counts.

Related reading

All post-show reporting articles