Skip to content

Part 2 of 7 · Feedback follow-up router series ~5 min read

How a response arrives and matches

Almost all of the value in a satisfaction response comes from knowing what it is about, and that is a property of how the survey was sent rather than anything you can recover afterwards.

Key takeaways

  • The survey link carries an order token, not just a customer identifier.
  • That means a response arrives already attached to a specific transaction.
  • Anonymous responses are kept and counted but cannot be followed up, and that is stated.
  • Ask once per order, not on a schedule, and never twice about the same thing.
  • The timing of the ask changes the score more than most people expect.

Token the order, not the person

How a survey is sent with an order token attachedA vertical chain of five steps entered by a box labelled An order completes, meaning delivered or the job done. Step one waits the right interval, taken from ask rules held per order type. Step two asks whether this order has been asked about before, ever; if so it exits to Do not ask again, because it is one ask per order. Step three asks whether this person has been asked recently at all, against a frequency cap; if so it exits to Skip this one, because survey fatigue is real. Step four sends the survey with a signed single-use order token. Step five receives a response that is already attached to a specific order. A note says one ask per order and a cap per person are both needed, because either alone is not enough.AWS ACCOUNTAn order completesdelivered, or the job doneWait the right intervalfrom the ask rulesAsk rulesper order typeAsked about this order?everDo not ask againone ask per orderyesAsked recently at all?frequency capSkip this onesurvey fatigue is realyesSend with an order tokensigned, single-useA response, already attachedto a specific orderOne ask per order and a cap per person. Both are needed; either alone is not enough.
Fig 1. How a survey gets sent so that its response is useful. The token carrying the order is what makes every downstream step possible without asking the customer what they mean.
  • App integration
  • Machine learning
  • Networking
  • Security & identity
  • Management

Why the order and not the customer

A response attached to a customer tells you somebody is unhappy. A response attached to an order tells you which experience they are unhappy about, which is the difference between a reply that asks what went wrong and a reply that already knows.

For a customer with one order the distinction is academic. For a repeat customer it is the whole thing: “sorry to hear that, can you tell me more?” reads very differently from “I can see the 14th arrived two days late — that is on us.”

Two caps, both needed

One ask per order prevents asking twice about the same experience, which is annoying and produces contradictory data. A frequency cap per person prevents a customer who orders weekly being surveyed weekly, which is how a business trains its best customers to ignore its emails.

Neither cap on its own is sufficient, and the frequency cap is the one usually missing. Sixty days is a reasonable default: a customer is asked about at most one order every two months regardless of how often they buy.

Timing the ask

Order typeAsk afterWhy
Delivered goods2 days after deliveryLong enough to have opened it
A service visitSame day, that eveningWhile it is fresh and specific
A long projectAt completion, then neverMid-project surveys measure mood
A subscriptionNever per orderAsk twice a year instead

Asking too early produces a score about the ordering experience; asking too late produces a score about how they feel in general. Neither is what the survey is for, and the interval is worth setting per order type rather than globally.

Anonymous responses

How an anonymous survey response is handledA horizontal row of five boxes. A response: with no token. From a widget: or a shared link. Cannot follow up: there is nobody to reply to. Still counted: in the themes. Reported separately: so the response rate is honest. A note says counting anonymous responses in the same average as tokenised ones flatters or distorts, usually the latter.THE RESPONSE WITH NOBODY ATTACHEDA responseno tokenFrom a widgetor a shared linkCannot follow upnobody to reply toStill countedin the themesReported separatelyso the rate is honestCounting anonymous responses in the same average as tokenised ones flatters or distorts, usually the latter.
Fig 2. How anonymous responses are handled. They are real feedback and they cannot be followed up, and reporting them separately is what keeps both numbers meaningful.
  • Security & identity
  • Analytics
  • Front-end & mobile

Anonymous responses skew low, consistently, because somebody moved enough to seek out a feedback widget is more often unhappy than the average recipient of a survey email. Blending them into one score produces a number that moves for reasons nothing to do with the business.

Next: deciding urgency.

All posts