Don't announce the improvement. A 2-point change on a 42% sample is well within non-response-bias noise.

What's wrong:

Non-response bias. Respondents are typically either very engaged or very disengaged. Both ends self-select; the middle doesn't. A 42% sample can swing ±5 points from year to year purely from which half of the disengaged middle responded.
Selection effects. Response rates differ by team, function, and tenure. If one team's rate jumped and another's fell, you're comparing two different populations Y/Y.
2-point change on a 0–100 scale is inside the confidence interval of any realistic survey even at a higher response rate.

What to say to the CHRO:

'The score is up 2 points, but with a 42% response rate I can't tell you whether that's a real improvement or a sampling artifact. Before we announce, I'd want to: (1) check response-rate change by team and demographic, (2) run the score weighted back to the full-population composition, (3) compare the directional signal across sub-questions. If the signal survives all three checks, we announce. If it doesn't, we don't — and we focus the communication on what we learned from the open comments instead.'

Response-rate improvement plan:

Shorter survey (response rate falls roughly linearly with length past 20 questions).
Leader visibility (known drivers of response rate).
Explicit action on last year's feedback — the biggest driver of next year's response rate is whether people believe it matters.

Red flag: accepting the score at face value because it's the direction the CHRO wants.

Engagement survey response rate is 42%. What do you do?

Draft your answer

Stuck? Peek at a hint

More people analytics questions