Sentiment Surveys That Produce Something

Asking people is the only way to reach the friction telemetry cannot see. Most surveys ask badly and then conclude that surveys do not work.

The boundary in “Sentiment Surveys That Produce Something” matters because operational data can easily be read as a performance score it was never designed to be. If a team considers the feature summary, it should define a legitimate purpose, explain the collection and restrict access before rollout, using the information to improve work design rather than infer intent from activity.

For an independent benchmark, compare this approach with ICO employment information guidance; the useful test is whether the evidence remains proportionate, accessible and understandable to the people whose work is being measured.

What goes wrong

Too long, so the response rate falls and skews toward the annoyed.

Too abstract: "how satisfied are you with IT" produces a number with no action attached.

Too infrequent: annual surveys describe a year nobody remembers.

And no visible consequence, which teaches people that responding is pointless.

The question that works

"What wasted your time this week?"

Open text, one question, specific period.

It produces a list of actual frictions, in people's own words, ranked by how often they recur across responses.

Nothing else in this field produces as much usable material per minute of anybody's time.

Adding one number

If a score is required, use one: how easy was it to do your job this week, on a short scale.

Tracked over time it shows direction.

On its own it shows nothing, which is why it needs the open question beside it.

Frequency and sampling

Short and often beats long and annual.

A single question to a rotating tenth of the organisation each month produces a continuous signal at negligible cost.

Everybody every quarter also works.

What does not work is everybody annually, which is the most common arrangement.

Anonymity

Necessary for honest answers about friction, because much of it is caused by somebody identifiable.

Which means no individual attribution, and a minimum group size for any team-level breakdown.

Its own note covers the small-numbers problem, which applies here exactly as it does to telemetry.

Closing the loop

The single thing that determines whether the second survey gets responses.

Publish: here is what you told us, here are the three things we fixed, here is what we are not fixing and why.

That last part matters more than it sounds. A stated reason for not fixing something is better received than silence, and silence is what most organisations provide.

Reading free text

Read it, do not just count keywords.

A hundred responses takes an hour and produces the themes directly.

Automated sentiment classification on this volume loses more than it saves, and the useful findings are usually in the specifics rather than in the sentiment.

What to check

How long is your current survey, and what is the response rate?

Does it contain an open question about wasted time?

When did you last publish what was fixed as a result?

And has anybody read the free text this year?

The point

One open question — what wasted your time this week — produces more usable material per minute than any survey instrument.

Then publish what you fixed.

Underlying all of this

Everything in this collection reduces to four habits: find the friction cheaply before buying anything, fix what needs no budget first, report the worst tenth rather than the average, and keep the data about systems rather than about people. None requires a better platform, and a programme doing all four changes more than one twice its size.

The recurring pattern

The recurring pattern across every section here is the same: the measurable is mistaken for the important. Device health stands in for experience, ticket categories for causes, a composite score for a finding. Each substitution is convenient, each produces confident decisions on thin ground, and each is corrected by going and looking at the thing itself.