Most seminar feedback forms are neither too short nor too long. They measure something other than what their authors believe — and in three cases they measure nothing you can evaluate at all. These three mistakes are the reason why numbers end up on the page that nobody can derive a decision from — and a decision gets derived anyway. That is the worse case: not the missing decision, but the wrong one.
Mistake 1: confusing satisfaction with learning
This is the biggest error of reasoning in the design of feedback forms, and it is already there in the first question of most of them: “How satisfied were you with the seminar?”
The research on training evaluation is unusually clear on this point. A meta-analysis by Sitzmann and colleagues (2008) shows that participant reactions are only moderately related to immediate knowledge gain — and to delayed knowledge practically not at all. Blume and colleagues (2010) go one step further: affective reactions, that is, whether people enjoyed it, are not related to transfer.
In plain terms: a seminar can get good marks all round and change nothing. And an exhausting day that caught people on a sore spot can be rated worse and have been the most effective one of the year.
That is no reason not to ask. Satisfaction is a separate, legitimate criterion — it says something about willingness to engage. It is simply not a measure of learning and must not be presented as one.
What to ask instead: about applicability and intention. “I know specifically what I am going to do differently in the next two weeks” is closer to what you want to know than any satisfaction question. And the question hardly anyone asks: “I can see obstacles that will stop me from putting this into practice.” It tells you what transfer will founder on, before it founders.
Mistake 2: questions whose answer has no direction
A question that appears in almost every form: “The balance between input and my own work felt right to me.” Answer: a 2 out of 5.
And now? You know it was not right. You do not know whether there was too much theory or too little practice. The two answers lead to exactly opposite changes to the design — and the question cannot tell them apart. The same goes for “The pace was appropriate” and “The level of detail was right”.
Questions like these need a two-sided scale with both ends named and the middle as the target:
| Aspect | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| Input and own work | far too much input | slightly too much input | just right | slightly too much own work | far too much own work |
| Pace | far too slow | slightly too slow | just right | slightly too fast | far too fast |
| Depth | far too superficial | slightly too superficial | just right | slightly too detailed | far too detailed |
Now the answer is usable: a 2 means “slightly too much input”, and you know what to change next time. The mean of such a scale also has a sensible meaning — it shows which way the group as a whole is tipping.
Mistake 3: two scale logics in the same form
This is where it gets dangerous, and this mistake only arises once you have fixed mistake 2.
The form now has two kinds of question: agreement questions, where 5 is the best result. And balance questions, where the middle is the best result.
If both kinds look the same — five boxes with numbers underneath — participants start ticking on autopilot after the third question and miss the switch. Later you take means across columns in which 5 means “very good” in one place and “far too fast” in another.
So change the format, not just the block. The balance questions get no number scale at all, but three named fields to tick:
| Aspect | ☐ | ☐ | ☐ |
|---|---|---|---|
| Input and own work | too much input | just right | too much own work |
| Pace | too slow | just right | too fast |
| Depth of detail | too superficial | just right | too detailed |
| Practicebeispiele | too few | just right | too many |
Three fields with words instead of five boxes with numbers. There is no number you could carry over from the previous question, and you cannot misread it without having read it. You lose the fine distinction between “slightly too much” and “far too much” — and that is a good trade. For the question of what to change next time, the direction is enough. A result in which seven out of twelve ticked “too much input” is clearer than a mean of 2.4 on a scale that half the room read the wrong way round.
Two more precautions against tiredness:
- The balance questions first. They demand more thought — so they belong where there is still attention. The agreement questions still come out right on autopilot, because they all point the same way.
- On one page. Whatever is on the back will not be filled in by part of the group at the end of the second day — and that part is not a random sample.
The direction trap: 1 is not self-explanatory
And now the point where forms are particularly vulnerable.
In several European school systems — Germany, Austria, Switzerland, the Czech Republic — 1 is the best grade and 5 or 6 the worst. On Amazon, on hotel review sites and in practically every app, five stars are the best and one is the worst. Both conventions are familiar to the same adults — and they run in exactly opposite directions.
A scale that only shows “1 – 2 – 3 – 4 – 5” is therefore not neutral but ambiguous. Part of your group reads it as a school grade, part as stars. The answers land in the same field and cancel each other out. This is not a theoretical risk: it hits precisely the people who fill things in quickly and routinely — that is, the majority.
Four rules that rule this out:
- Never bare numbers. Every scale point gets a label in words. Numbers may stand beside them, but never carry the meaning on their own.
- Keep the direction constant throughout the form. Negative always on the left, positive always on the right. On following pages too, even when space gets tight.
- State the direction once at the start. One line above the first block: “Left means disagreement, right means agreement.”
- Do not use school grades. Neither as a scale nor in the wording. “What mark would you give the seminar?” brings the school direction back into a form that is built around stars.
In multilingual groups there is one more point: the direction of school grades is not the same everywhere. In the Netherlands, in much of Scandinavia and in Spain the highest number is the best; in Germany and Austria it is the lowest. Labelled scales solve this problem too — bare numbers make it worse.
How many points, with or without a middle
Five points. Four is too coarse, ten suggests a precision nobody has — the difference between a 7 and an 8 means nothing comparable across two different people.
For balance questions the odd-numbered scale is essential: the middle is the optimum there, so it has to exist.
For agreement questions the middle is the familiar point of dispute. It allows genuine indecision — but on uncomfortable questions it becomes an escape hatch. The practical solution: an odd-numbered scale, and next to it a separate field, “cannot judge”. That moves the escape out of the middle, and the middle means again what it is supposed to mean.
A form that avoids these three mistakes
Then: agreement — five points, all labelled
- I know specifically what I am going to do differently in the next two weeks.
- What I learned fits situations that really occur in my working day.
- I can see obstacles that will stop me from putting this into practice.
- I would have dared to ask an uncomfortable question here.
- I had enough opportunity to contribute.
The fourth question is a usable proxy for how much safety was created in the room — and it is more uncomfortable, and therefore more honest, than “The facilitation was competent”.
First: balance — three fields to tick, no numbers
- Balance between input and own work
- Pace
- Depth of detail
- Anteil an Beispielen aus der Practice
Finally: two open questions
- What has changed for you compared with this morning?
- Which part would you pass on to a colleague? — shows what stuck, and incidentally supplies wording for your own seminar description.
Open fields belong at the end, not at the beginning: someone who has already ticked boxes is inside the form and is more likely to write.
When to fill it in
In the room, not by email afterwards. With an email sent later, it is mainly the enthusiastic and the annoyed who answer — the middle is missing, and that is exactly where the usable material is.
Plan five minutes for it, as a block with its own duration in the schedule. Not “if we still have time” — a block that has a duration gets moved; a leftover gets deleted. Why that is the single most effective trick for closing the day is in Feedback round in a workshop.
Anonymous — and honest about how far that goes. Two limits worth saying out loud instead of glossing over.
First, with small groups anonymity is a claim. With six people you recognise handwriting, turns of phrase, and who wrote about what.
Second, the forms usually travel on. In corporate training the client evaluates them personally — that is the normal case, not the exception. A promise that they stay with you would in most cases simply be untrue.
So say what actually happens: “These forms go to HR development, without names. Anything you want to tell me personally, tell me in the break.” That is honest, it costs nothing — and incidentally it creates the second channel, which brings the more usable feedback. Why this is not a side issue but the core of a structural conflict is in When trainer feedback becomes a report card.
And the recommendation question?
Many forms contain instead the single NPS question on a scale from 0 to 10. Why that is not a metric but noise for seminar groups under twenty people is in Is NPS any good for seminars?.
In summary
- Satisfaction does not measure learning — ask about intention and obstacles
- “It felt balanced” is not a usable question: use a two-sided scale with named ends
- Never give two scale logics the same look — balance questions as three named fields without numbers
- Never bare numbers — depending on where people went to school, 1 is either the best mark or the worst
- Five points, all labelled, direction constant — and everything on one page
- Two open questions right at the end, filled in in the room
