Company Event Feedback Form: Template and 20 Questions That Show Whether It Worked

Short answer: they work on feelings and collaboration, and they do not reliably work on performance. That is not a criticism — it is the reason your feedback form has to ask different questions than it probably does today.

The meta-analysis by Klein and colleagues (2009, Small Group Research), covering 60 correlations, found a moderate positive overall effect of team building with a mean true-score correlation of r = .31 — and the effect was strongest for affective and process outcomes, weakest for cognitive and performance outcomes. Ten years earlier, Salas and colleagues (1999) found no significant overall effect on performance at all, with a nonsignificant tendency to decrease on objective measures. Their one notable exception: interventions emphasising role clarification were more likely to improve performance, while goal setting, problem solving and interpersonal relations were not.

This is the single most useful finding in the whole field, and no German page on this topic mentions it. Alliger and colleagues (1997, Personnel Psychology) meta-analysed 34 studies with 115 correlations and found that the average correlation between participant reactions and immediate learning was r = .08 — effectively zero. Liking something tells you almost nothing about whether it did anything.

But they also split reactions into two kinds — affective ("did you enjoy it?") and utility ("was this useful, will you apply it?") — and found that utility-type reactions related more strongly to learning and to on-the-job transfer than affective ones did. That converts a vague piece of advice into a hard rule you can apply to every line of your form: rewrite each liking question as a utility question. Not "wie hat dir der Tag gefallen", but "was machst du nächste Woche konkret anders".

Take blocks 1 to 4 for the first measurement (max. twelve questions in total), and blocks 5 and 6 for the second one four to six weeks later. Block 1 is the one nobody asks and the one worth the most: it is behavioural rather than attitudinal, which makes it far harder to inflate out of politeness.

A form handed out at the end of the evening does not measure the event. It measures the last twenty minutes plus the single most intense moment. That is the peak-end rule, first described by Kahneman and colleagues (1993) and confirmed meta-analytically by Alaybek and colleagues (2022) in Organizational Behavior and Human Decision Processes: retrospective evaluations of an extended experience are driven almost entirely by the emotional peak and the ending, while the actual duration barely registers.

Practically: a strong closing dinner produces glowing scores for a mediocre day, and you will never see it in the data. The fix is not to abandon the early measurement — you need it for organisation and for the details that fade — but to add a second one after four to six weeks that asks only about what remained. The gap between the two is the actual result. If the first is enthusiastic and the second is flat, you bought a nice evening rather than a change, which is a perfectly legitimate thing to buy as long as you know that is what happened.

No competing page on this topic covers this, and it is the part that stalls the project. A named, software-based survey of employees is generally co-determined: section 94(1) BetrVG covers Personalfragebögen, and section 87(1) no. 6 covers technical systems suited to monitoring behaviour or performance — the suitability triggers the right, not your intention. The Federal Labour Court held on 11 December 2018 (1 ABR 13/17) that co-determination can fall away where the survey is genuinely anonymous and voluntary.

The practical rule that follows is a minimum group size: results are only released once at least five people in a group have answered, and smaller groups roll up into the next level. Without it, filtering by department and location and tenure de-anonymises a six-person team in three clicks, and the anonymity you promised was never real. Treat the works council as an ally rather than an obstacle — a jointly announced survey demonstrably gets a higher response rate, because participation follows trust in the process. This is orientation, not legal advice: involve your works council and data protection officer before you send.

A Christmas party is the one company event where attributed feedback actively backfires. The honest answers involve colleagues' behaviour, alcohol and sometimes conduct — and a named form makes "ich habe mich unwohl gefühlt" effectively unwritable. Run it anonymously, and keep a separate, clearly signposted named channel for conduct issues so that a report reaches the right process instead of arriving through a satisfaction survey that was never designed to carry it. And send one question to the people who did not come at all: what kept you away. That is the most useful data a Christmas party produces and almost nobody collects it.

If you want the two-measurement design without building it by hand, this is the short version of the setup.

For the generic version of this form across all event types, see the event feedback form template. For seminars and training, use the training feedback form, which covers learning transfer.