Reference

What's actually in each thing

The four artefacts the product produces, and what each one holds. If you want to copy the method rather than the software, this is the part worth copying.

A feedback request

Sent to one named person, for one audience. Management get asked whether it's worth what they pay; the people using it get asked whether it's faster than before. Both get one rating and one open question, and an optional 'anything else'.

who it went towhich audiencethe questionsa one-click tokenwhen it was sent

A usage event

One action in the client's app. Deliberately thin: enough to count and compare, not enough to profile anyone. The actor reference is opaque and never resolved to a person.

clientmetric keyvaluewhen it happenedwhich appopaque actor reference

An incident

Something we got wrong, on the record. It pulls the score down while it's open — an unresolved one counts double — and it appears in the value report next to what we did about it.

kindseverityone-line summaryopenedresolvedresolution

A value report

The renewal artefact. Usage against what was expected, what each audience said in their own words, the incidents and their resolutions, and the assumptions behind every figure. Printable, and written to be read by the client.

usage vs expectedfeedback by audienceincidents and fixesstated assumptions

Decisions still open

Listing them beats pretending they're settled — these are operating choices, not missing features, and the honest answer today is that we haven't made them.

How often do you ask for feedback?
Undecided. The candidates are a fixed cadence, or triggering on events that matter — after an incident, before a renewal. Asking too often trains people to ignore it; asking once a year tells you nothing you can act on.
Who chases a client who doesn't reply?
Undecided, and it matters more than it sounds. An unanswered request is itself a signal, but only if someone notices it went unanswered.
Does the client ever see the score?
Our recommendation is no — they see the report. A number we control, shown to the person it judges, invites an argument about the number instead of the substance.
How long do you keep raw events?
Undecided. Per-event rows are what make the evidence credible; they're also the thing that grows without limit. At some point they roll up into stored snapshots.