Return on objectives for events when the goal was never a revenue number

The launch edition of a new category zone came in at a loss and the show director is entirely happy about it. She was asked to establish a position in a category the show had never covered, before a competitor did, and she has.

Then the portfolio review asks for the ROI, gets a negative number, and writes the zone down as underperforming. Nobody in the room believes it, including the person who wrote it, and the zone survives on the strength of an argument made verbally. Next year, with a different director in the room, it will not.

The failure is that the objective was never written down in a form that could be scored. The event ROI calculation that D17 sets out is a perfectly good instrument for an edition whose purpose was money. Point it at an edition whose purpose was market position and it returns a confident wrong answer, which is the gap return on objectives for events exists to fill.

Score against the levels, and set them before the show

The framework worth borrowing comes from training evaluation, where the same problem was solved decades ago.

Donald Kirkpatrick published his four-level training evaluation model in 1959: reaction, learning, behaviour and results. Jack Phillips later added return on investment as a fifth level, which is the structure the ROI Institute still teaches.

For an edition of an exhibition I would run six levels, numbered zero to five, with one added underneath Kirkpatrick's first. Level zero is target audience. Level one is satisfaction, which is what reaction means once the room is a hall. Then learning, behaviour, impact, and return on investment. The extra level at the bottom is my own addition and it earns its place, because a training course knows who is sitting in the room and a show does not.

The numbering is the useful part. ROI is level five, which means it sits on top of five other things that have to be true first, and an edition can succeed completely at levels zero through four while never being scored at level five at all. Level zero, target audience, is the one organisers skip most and it is the one that decides whether anything above it means anything.

The rule that makes this work is timing. Objectives and their thresholds are fixed before the show opens, in writing, with a named owner per level. Objectives set afterwards are descriptions of what happened, and everybody in the room knows it.

Running the six levels on the category zone

Take the launch zone from the opening scene. Six feature areas were built, 38 exhibitors took space in them, and the purpose was to establish the show as the place the category meets.

Level zero, target audience. The objective was 40 qualified buyers per feature zone, where qualified means the registration profile names the category as a purchasing responsibility and the seniority is manager or above. Delivered was 31 per zone. Across six zones, 240 targeted and 186 delivered, which is 77.5 per cent of target. Missed. This is the level that leans hardest on attendee analytics, because it is a claim about who was in the building rather than how many.

Level one, satisfaction. The objective was 70 per cent of zone exhibitors rating the zone four or five out of five. Delivered was 24 of 38, which is 63.2 per cent. Missed, narrowly.

Level two, learning. The objective was 60 per cent of surveyed zone visitors able to name two suppliers in the category unprompted after the show, against a pre-show baseline of 18 per cent. Delivered was 41 per cent. Missed on the target, and a rise of 23 points on the baseline.

Level three, behaviour. The objective was 25 per cent of the qualified buyers requesting a follow-up meeting with a zone exhibitor. Delivered was 41 of 186, which is 22.0 per cent. Missed, narrowly.

Level four, impact. The objective was 12 of the 38 zone exhibitors rebooking for the next edition, which is 31.6 per cent. Delivered was 15, which is 39.5 per cent. Beaten.

Level five, ROI. Deliberately not set, because the zone was funded as a market position investment with a three-year horizon and a one-year ratio would have scored it against a goal nobody had.

Read that scorecard and the picture is coherent. Fewer qualified buyers turned up than hoped, the ones who did behaved roughly as expected, awareness moved a long way from a low base, and the exhibitors who paid for it want to come back at a higher rate than the target. That is a first edition that worked, with a specific and fixable problem at level zero.

Why is the level zero miss the finding?

The temptation with a scorecard like that is to average it, or to count three misses against one beat and call it a failure. Both are wrong, because the levels are not independent.

Level zero feeds everything above it. 186 qualified buyers instead of 240 mechanically caps level three, since 25 per cent of 186 is 47 meetings against 25 per cent of 240 being 60. The behaviour result of 22.0 per cent on a smaller pool is close to the rate that was targeted on a larger one, so the conversion worked and the audience did not arrive.

That reframes the action list entirely. If you read the scorecard as four separate results you commission four workstreams. If you read the dependency you commission one, aimed at getting qualified category buyers into the building, and you expect three of the four numbers to move on their own next year.

The general rule is to work from the bottom of the stack upwards when diagnosing, and from the top downwards when setting targets. Set what impact you need, work out what behaviour produces it, what learning produces that, and what audience has to be present at all. Most objective sets fail because they were written as a list rather than as a chain.

What thresholds do you set with no history?

The structure is the easy part. Picking the numbers is the hard part, and for a first edition you have nothing to pick from.

Three sources are usable and all three should be named in the objective document.

Your own comparable areas. If an established zone at the same show delivers 52 qualified buyers per feature area, a new one at 40 is a stated discount for being new, and the discount is a judgement somebody signed.

The exhibitors' own expectations. Ask the 38 zone exhibitors before the show what they need to see, and take the median. This is the most useful of the three and the least used, and it has the side effect of making the post-show conversation vastly easier because they told you the threshold.

A published benchmark, used carefully. Freeman's End-of-Year Trends Recap on 2025, released in January 2026, reports that 20 per cent of attendees do not believe their objectives are being met at events, and that only 40 per cent say they experienced what it calls a peak moment while 85 per cent of those who did are likely to return. Numbers like those set a realistic expectation of what a satisfaction threshold should be. They are not your show and should never be presented as though they were. If your level one instrument is a recommendation score, how far net promoter score for events can be pushed is D27's question.

Where none of the three gives you anything, set the threshold anyway and label it as an estimate. An arbitrary threshold recorded before the show is enormously more useful than a sophisticated one derived after it, because the point of the exercise is to make the comparison honest rather than to be right in advance.

Criteria have to be measurable with data you will actually have

An objective set that requires an instrument you have not built is not an objective set.

Level zero needs a qualification field on the registration form, present before registration opens. Level one and two need a survey with the right questions and a response population you can weight. Level three needs meeting requests captured in a system rather than remembered by a stand team. Level four needs the rebooking file joined to the zone exhibitor list.

Walk the six levels and, against each, write the table and column the answer will come from. Any level where you cannot name a source is a level you cannot score, and the choice is to build the instrument now or to drop the objective. Carrying an objective you cannot measure is how a scorecard turns into a narrative, and the post event survey design that supports levels one and two has enough in it to be its own subject, which is D22's.

The measurement plan is also where the honest constraints appear. Unprompted brand recall needs a pre-show baseline, which means fielding a survey before the show that nobody will want to pay for. If that will not happen, change the level two objective to something the post-show survey can carry, and record that you traded rigour for feasibility.

Where this stops

Return on objectives is only as good as the objectives, and objectives are set by people with an interest in the result.

There is no mechanism in the framework that stops somebody setting soft thresholds. The only real control is that the thresholds are published to a group wider than the team being scored, before the show, and that last year's thresholds are visible next to this year's. A threshold that falls every year in line with performance is visible in about four seconds when the series is printed together.

The second limit is that most of the levels rest on survey data, which means they rest on who answered. A zone satisfaction figure of 63.2 per cent computed from 38 exhibitors is 24 people, and if the four angriest did not respond it is a different number. At small populations the level one and two results should be reported as counts rather than percentages, because 24 of 38 invites the right amount of scepticism and 63.2 per cent does not.

The third is the one to be most honest about. Levels zero through four can all be met by an edition that loses money, and at some point somebody has to decide how many editions of that a business will fund. The framework deliberately does not answer that. It makes the trade explicit, which is worth a great deal, and it does not make the decision.

Take the objective document for your next edition, or write one if it does not exist, and put a level number and a data source against every line in it. Any objective without both is a sentence, and it will not survive the review it was written for.