Best practices
A playbook for implementing Sentriment well — and getting business value from it.
Sentriment works out of the box, but a few decisions early on multiply what you get out of it. This is the playbook we'd give a new team — how to implement well, and how to turn it into business results.
1 · Get identity right first
#Identity is the multiplier. Anonymous feedback gives you themes; identified feedback gives you health scores, revenue at risk, and segment insight. The single highest-leverage thing you can do on day one:
Always send an email
Send email with every user — via the widget's data-email, the API's email field, or an email trait on identify. Email is the join key that unifies a person across your app, Intercom, Zendesk, and CSV imports. Skip it and the same customer fragments into three or four separate “users”, each with its own half-formed health score. This is the mistake that's expensive to fix later and free to avoid now.
Use a stable userId too if you have one — it's the strongest identifier and never changes. Full mechanics in Identifying users.
2 · Metadata vs traits — send both, deliberately
#The distinction quietly shapes every chart and segment you'll build later:
- Traits describe the person and persist:
plan,company,mrr,signup_date,churned. - Metadata describes a single item:
page,app_version,experiment,widget.
Every metadata key becomes a filter and a chart breakdown in Reports; every trait becomes a segment and feeds Impact. Be generous — send app_version and you can answer “did the last release spike bug reports?” without any extra work.
3 · Wire up revenue impact
#To make feedback a revenue conversation, send two traits from the start: a value trait (mrr) and an outcome trait you'll set when a user churns (churned). Because traits merge, you can flag churn after the fact and old feedback correlates retroactively — so start sending mrr today even if you can't mark churn yet. Turn on the model in Impact.
4 · Team workflows that stick
#Docs are only useful if they fit how teams actually work. What we see land:
- Product lives in Clusters → Triage weekly: sort by revenue at risk, file the top themes to Linear/GitHub with quotes attached, mark fixed themes resolved (they reopen automatically if the fix fails).
- Customer Success works the Users at-risk list and gets a Slack alert the moment an account turns critical — outreach before the churn email, not after.
- Leadership gets a saved Report of negative feedback by area, and a tracked question per bet (“is onboarding landing?”) whose verdict they watch.
- Engineering subscribes to spike and regression webhooks to watch deploys in real time.
Set tracked questions before a launch, not after
The value of a tracked question is watching the verdict move. Create “what do users think of the new editor?” the day before you ship, and you'll see sentiment turn in real time instead of reconstructing it later.
5 · Business use cases
#Where teams get concrete returns:
Cut churn
At-risk health scores + alerts let CS intervene before renewal. Impact tells you which fixes protect the most MRR.
Prioritise the roadmap
Rank themes by revenue at risk, not raw volume — and back every roadmap call with “this protects $X and N accounts”.
Harvest testimonials
Filter to praise, copy the quotes. Your happiest users are marketing gold you already own.
Deflect support
Spot the top recurring issues and fix the docs or product so the tickets stop arriving.
Validate launches
A tracked question per launch turns “I think it went well” into a dated sentiment curve.
Report to the board
Saved reports + revenue-at-risk numbers make a feedback slide that speaks in dollars.
A good order to roll out
1. Get feedback flowing (widget or a connector). 2. Add identity with email + traits. 3. Turn on Impact and Health. 4. Wire alerts to the right channels. Each step compounds the last.