How to build an NPS and survey to CRM sync

What matters here

• Send the verbatim, not just the score. The number says there is a problem and the sentence says what it is.
• Route a detractor within the hour. A complaint answered the same day is a save; the same complaint in a monthly report is a churn post-mortem.
• Match to the person as well as the account. Who said it changes what it means, especially if it was the champion.
• Enforce fatigue rules. Surveying somebody three times a quarter reduces your response rate and annoys the people who reply.
• Track whether detractor follow-up actually happened. The routing is worthless if nobody acts on it.

Who this is for

You run customer success or marketing operations. Surveys go out, scores get reported quarterly, and the individual responses that would have changed something are read by nobody.

How it works in practice

What happens between somebody answering a survey and anybody acting on it.

1. 1

   Fatigue is checked before the invitation goes

   Recently surveyed, recently responded, or in an open escalation. Asking then is worse than not asking.

2. 2

   The response is matched to a person and an account

   Both, because a detractor who is the champion is a different situation from one who is an occasional user.

3. 3

   The verbatim travels with the score

   A paragraph explaining why is worth more than the number it accompanies.

4. 4

   Detractors route within the hour

   To the account owner with the response attached, not into a monthly summary.

5. 5

   Promoters route somewhere useful too

   A reference request or a review, which is the return most survey programmes never collect.

6. 6

   Follow-up is tracked to an outcome

   Whether somebody actually responded to the detractor, because routing without follow-up is theatre.

Paste these into Claude Code or Codex with the Tray Headless plugin installed. Each stage runs on its own. The systems named in them are the worked example rather than a requirement, and every prompt says so.

Once per project, run /tray-workflows:set-workspace to pick the workspace these build in. Point it at a sandbox first.

1. 1

   Set up, then decide who should not be asked

   Fatigue rules protect the response rate, which protects everything else.

   Headless skills build-workflow

   Copy

   Use build-workflow. The systems in play are Qualtrics and Salesforce,
   or whatever we run in those seats.
   
   I need the survey and response model with the contact identifier used,
   and from Salesforce the contact, account, open cases and any escalation.
   
   Then build the suppression list before anything else: anybody surveyed in
   the last quarter, anybody who responded in the last two, anybody on an
   account with an open severity one, and anybody who asked not to be
   contacted.
   
   Over-surveying is how a programme quietly dies. The response rate falls,
   the sample skews to the annoyed, and the score stops meaning anything.

2. 2

   Match to the person and the account

   Who said it changes what it means.

   Headless skills build-workflow

   Copy

   Use build-workflow. Match every response to both a contact and an account.
   
   Key on the contact id carried in the invitation, not on the email
   address. People change email and an email-keyed match silently breaks the
   history for exactly the long-tenured contacts you most want to track.
   
   Where a response arrives anonymous or unmatched, keep it, count it in the
   aggregate, and flag it as unattributed rather than discarding it.
   
   Write to the contact: score, date, verbatim. Write to the account: latest
   score, trend across the last three, and the count of detractors among its
   contacts. That last field is the one worth looking at, because one
   detractor is a person and four is an account.

3. 3

   Route detractors within the hour, with the words

   The half life of a useful response to a complaint is about a day.

   Copy

   Route any detractor immediately, not in a digest.
   
   Send the account owner and the customer success manager: the score, the
   full verbatim, who said it and their role, the account context, and any
   open support tickets.
   
   Include the verbatim in the message itself. Not a link to a dashboard.
   The paragraph is the reason to act and putting it behind a click means it
   does not get read.
   
   Create a follow-up task with a due date. Then track whether it was
   actually completed, because routing without follow-up is a workflow that
   makes everybody feel efficient and changes nothing.

   If the detractor is the named champion on a renewing account, that should page somebody instead of create a task.

4. 4

   Do something with promoters too

   The return most survey programmes never collect.

   Headless skills tray-gotchas

   Copy

   Use tray-gotchas, then handle promoters deliberately:
   
     Promoters on accounts with no reference on file go to a reference
     request, throttled so the same person is never asked twice a year
     Promoters after a support interaction go to a review request
     A promoter who is a new champion on an account whose previous champion
     left is worth telling the account team about
   
   Handle the awkward cases: a promoter on an account that is churning
   anyway, which usually means you surveyed the wrong person, and a
   detractor whose complaint is about pricing, which is a commercial
   conversation instead of a support one and should route accordingly.

5. 5

   Report the things that change behaviour

   A quarterly average changes nothing on its own.

   Copy

   Report: score and trend by segment, product area and tenure; response
   rate and whether it is falling, which is the health of the programme
   itself; detractor follow-up completion rate; and churn among detractors
   against churn among everybody else.
   
   That last comparison is what tells you whether the survey predicts
   anything. If detractors churn at the same rate as everybody else, the
   question is wrong or the sample is.
   
   Do not report the company-wide average as the headline. It moves slowly,
   it hides every segment that is struggling, and nobody has ever acted on
   it.

6. 6

   Validate, then hand the rules over

   Because who gets asked and how often is a relationship decision.

   Copy

   Run the per-step schema checks and the whole-workflow audit before this
   touches production.
   
   Then open the same workflow in Tray Build so customer success operations
   can adjust the fatigue windows, the detractor routing and the promoter
   paths in the visual canvas.

What it connects to

Responses arrive in one tool and have to reach the people who own the relationship.

Named systems are the ones most teams run, not the only ones that work. Each is an authentication in your Tray workspace, referenced by name, so the workflow never holds a credential. Where we have a connector page, the name links to it.

Questions people ask

Further reading

Background on the same subject, for the case rather than the build.

• How to integrate NPS into Salesforce CRM

Related guides