A quarterly review lands in your inbox with one line in it: how many people attended? The sponsor wants the headcount, the finance partner wants it in the same spreadsheet as last quarter, and nobody wants a slide on behaviour change. You have two options. You can argue that attendance proves nothing and watch the renewal get “reviewed”, or you can send the number and quietly attach a second one that makes the first one honest.
Send the number. Then add exactly one more. That is the whole move, and it is harder than it sounds, because the second number has to be one a finance reviewer will accept, not one the L&D team finds meaningful.
Why attendance is the number they ask for
Attendance wins for reasons that have nothing to do with laziness. It is a count of something that definitely happened, it is collected inside the room by the trainer, it needs no cooperation from line managers, and it costs nothing to produce. Procurement can compare it across vendors. Finance can tie it to an invoice. Level 1 reaction scores sit close by for the same reason: roughly 80 per cent of training events include a Level 1 measure, an estimate Training Industry’s overview attributes to a variety of sources, because the data is cheap and available on the day.
The result is an odd split. In ATD Research’s 2025 survey, 71 per cent of organisations said they use Kirkpatrick as their evaluation framework. In ATD’s 2016 research, only 35 per cent of 199 talent development professionals said their organisations evaluate the business results of learning. Most programmes are bought and renewed on the cheapest number in the ladder, and the expensive numbers are the ones that get skipped.
Refusing to report attendance does not fix this. It removes the one figure the sponsor trusts, gives them nothing equally cheap to put in its place, and turns a reporting request into an argument about method. At renewal time the argument loses. The sponsor does not need to be convinced that level four is hard; they need a number they can put in a board pack. Take that need seriously rather than resenting it.
What makes a leading indicator survive a finance review
The number you add should not be a satisfaction score, a self-assessment or a promise to measure impact later. It should be a leading indicator: a measure of the work that is supposed to change the outcome, recorded while the work is happening and before the outcome is known.
OSHA’s guidance on leading indicators, publication 3970, defines them as proactive, preventive and predictive measures that say something about whether the activity is being done, and contrasts them with lagging indicators, which count events that already occurred. The same guidance makes a point that matters more than the definition: build leading indicators from data you are already collecting. It even uses attendance at refresher training as an example, turned into a percentage with a goal.
Two design rules follow, and they are the difference between a number finance accepts and one it waves away.
First, express it as a rate, not a count. OSHA’s SMART checklist asks whether the indicator is “presented as a number, rate, or percentage that allows you to track and evaluate clear trends over time”. A count of 40 attendees is not auditable beyond the invoice. “40 of 44 scheduled supervisors attended, 91 per cent” is, because the denominator is named. A leading indicator needs the same discipline: a numerator, a denominator, and a fixed window, all written down before anyone sees a result.
Second, take the denominator from a system the business already runs and already trusts. The shift log, the call record, the CRM, the ticket queue, the inspection register. If you have to invent a new form, a new survey or a new meeting to feed the indicator, the cost advantage is gone and the measure will be quietly dropped after the first cohort. OSHA’s phrase is that “just one or two indicators can make a positive impact”. One is the right number here.
The method, in the order the work actually happens
Do these in sequence. The order matters, because steps three and four are what make the number defensible, and they have to happen before the programme starts.
- Give them the attendance figure exactly as asked, in the same format, on the same sheet, on time. Do not annotate it and do not attach a caveats paragraph. The point of this step is to keep the renewal, not to win a debate.
- Find one behaviour the programme is meant to change that already leaves a timestamped record in an operational system. One behaviour, not a competency map. If the programme teaches supervisors to run a pre-shift briefing, the shift log already holds the briefing timestamp.
- Write the indicator on one line as a rate: numerator, denominator, window, source system, and the person who owns that system. “Briefings logged before shift start, over shifts scheduled, per week, from the shift log, owned by the production analyst.”
- Put that line in front of finance or the system owner before kickoff and get the denominator agreed in writing. This is the whole test of whether the number survives review. If they will not confirm the denominator, the indicator will be dismissed later no matter how well it moves.
- Fix the measure in writing before the first cohort, fix the expected direction, and name who supplies it each period. No choosing the metric after you have seen the result.
- Print it as a second row on the same one-page report as attendance. Two rows, no new deck. Publish the flat weeks and the down weeks as well as the good ones.
| Attendance (what they asked for) | The one leading indicator (what you add) | |
|---|---|---|
| Form | A count of heads | A rate: numerator over a named denominator |
| Timing | Lagging, captured in the room | Leading, visible in the work before the result |
| Source | The sign-in sheet | A system the business already runs |
| Owner | The trainer | A line manager or system owner |
| What it proves | The programme was delivered | The target behaviour is showing up on the job |
Three programme types make the choice concrete:
| Programme on | Leading indicator (rate) | Denominator | Source system |
|---|---|---|---|
| Pre-shift briefings | Briefings logged before shift start / shifts scheduled | Shifts scheduled | Shift log |
| Consultative selling | Calls with a logged next step inside 24 hours / calls made | Calls made | Call record or CRM |
| Walkway inspections | Daily walkway checks completed / workdays | Workdays | Inspection register |
Each of these is a rate with a denominator somebody other than the trainer already produces. That is what makes it auditable.
Where a fuller evaluation is actually funded, the honest four-level version is a separate job, covered in four levels of evaluation, minus the theatre. This article is for the more common case, where it is not. Fixing a metric before a proposal is signed is the same discipline we bring to consulting engagements, and a baseline, if one is needed, belongs with the assessment work. The behaviour you settle on has to be one the programme genuinely teaches, or the indicator measures nothing.
A worked example
Illustrative example. A plant ran a two-day programme for 104 frontline supervisors on running a ten-minute pre-shift briefing. The sponsor asked for one number: attendance. The report gave it, 96 of 104, in the same spreadsheet as the previous quarter.
Underneath it sat a second row. The team picked one behaviour, a briefing logged before the shift began, from the existing shift log. The indicator was written as “briefings logged before shift start over shifts scheduled, per week”, with the roster as the denominator and the production analyst as the owner. Before kickoff, the analyst confirmed the roster was the authoritative count of scheduled shifts. Four weeks of baseline were pulled from the same log.
At baseline, 61 per cent of scheduled shifts had a briefing logged before start. By week eight of the programme the figure was 84 per cent. Attendance stayed on row one at 96 of 104, unchanged, exactly as requested. No new survey, no observer form, no extra meeting. The second row cost one query on a system that already existed.
Those figures are illustrative, not a client result. The transferable part is the shape: one rate, one agreed denominator, one existing system, printed next to the number the sponsor asked for.
The honest limit
A leading indicator is a proxy, and a proxy is easy to fool. Safety measurement writers put it plainly: a leading indicator that is easy to score is easy to game. A supervisor can log a briefing that never happened. A rep can type a next step into the CRM to protect the percentage. Once the number becomes a target, the number gets managed. Guard it with occasional spot checks against reality, and never tie it to an individual bonus.
The deeper limit is that the term is loose. A 2023 synthesis of the leading-indicator literature, reported by a safety consultancy, found its definition, application and function “mostly ambiguous and inconsistent”, and concluded that lagging indicators remain the most appropriate evidence base from which leading indicators are built. In other words, do not sell this second row as impact. It says the behaviour is appearing more often. It does not say the training caused it, and it does not say a business result moved.
It also only works when the behaviour already leaves a record. If the thing you want to count has no timestamp anywhere, you have a measurement project, not a reporting tweak, and it belongs in a priced training solutions conversation rather than smuggled into a spreadsheet. And if the sponsor will genuinely never look at the second row, one indicator still beats the argument you would lose, but call it what it is: a hedge, not an evaluation.
What to do on Monday
Take one programme renewing this quarter where the sponsor asks only for attendance. Open the operational system the business already runs, the shift log, the CRM, the ticket queue, and find one behaviour the programme targets that leaves a timestamped record.
Write it on one line as a rate: numerator, denominator, window, source, owner. Take the denominator to finance or the system owner and get it agreed in writing before the next cohort. Add it as a single row beneath attendance on the same one-page report. Then publish the flat weeks too.
Next quarter, when the sponsor asks for the headcount again, send both numbers in the same reply. The first keeps the renewal. The second is the one that eventually changes how the programme is bought.
Sources
https://www.devlinpeck.com/content/kirkpatrick-model-evaluation https://trainingindustry.com/wiki/measurement-and-analytics/the-kirkpatrick-model https://www.osha.gov/sites/default/files/publications/OSHA_LEADING_INDICATORS.pdf https://www.soter.com/blog/leading-vs-lagging-safety-indicators https://isienvironmental.com/blog/osha-leading-indicators-vs-lagging-indicators