Help Scout satisfaction ratings as a churn signal
What Help Scout's Great, Okay and Not Good ratings can and cannot tell you about churn: how ratings work, the Happiness report, comments, workflows and the API.
By Florian
Help Scout puts a small question under every reply: was this Great, Okay or Not Good? The answers add up to a Happiness Score that many teams watch closely. It is tempting to treat a falling score, or a customer’s “Not Good”, as a sign they are about to leave. Sometimes it is. Often it is not.
This guide looks at Help Scout satisfaction ratings specifically as a churn signal: how they are collected, what the Happiness report really measures, where the useful information hides (the comments), and how to test whether ratings predict cancellations for your customers. Facts about Help Scout come from its documentation: email satisfaction ratings, chat satisfaction ratings, the Happiness report, workflow conditions and actions, and the Inbox API’s ratings and happiness reports endpoints (all accessed November 2026).
How ratings are collected
Per Help Scout’s email ratings guide:
- Ratings are turned on per inbox (Settings > Inboxes > your inbox > Satisfaction Ratings), and rating links are added to outgoing replies.
- The customer picks Great, Okay or Not Good, lands on a page with an optional comment box, and the rating is saved when they press Send.
- Customers have 10 days from a reply to rate it or change their rating.
Chat ratings from Beacon are collected separately and also feed the Happiness report (chat ratings).
Three consequences for churn analysis:
- A rating is about one reply, not about the product or the relationship. “Not Good” may mean “your answer didn’t solve my problem”, “I don’t like your pricing”, or “I wanted a refund and didn’t get one”.
- Only some customers rate. Those who do are not a random sample.
- A rating can change within 10 days, so a snapshot taken too early may differ from the final picture.
What the Happiness report measures
The Happiness report shows how many ratings you received in the period and the share of Great, Okay and Not Good. The Happiness Score is the percentage of Great minus the percentage of Not Good. Ratings are indexed by the date of the reply, not the date the customer rated.
Per-person ratings are in the Company report, under each user. On Standard plans, reports cover the last two years of data; Plus and Pro cover the lifetime of the account, according to the same page.
The Happiness Score is a support quality metric. It tells you whether replies, in aggregate, are landing well. It is not designed to tell you which customers will cancel, and it moves with things that have nothing to do with churn: a new teammate, a seasonal spike, one bad release.
Where the churn signal actually is
The useful information is in three places, in this order.
1. The comment
“Not Good” with no comment is a weak signal. “Not Good: third time I’m asking and still no fix, we’re evaluating alternatives” is a strong one. Read every comment on a Not Good rating. Most weeks there are few enough to read them all.
2. The topic of the conversation
A Not Good on a conversation tagged billing or cancellation means something different from one on a how-to question. Combine ratings with your tags: Not Good on a pricing conversation, or on a repeat bug report, is worth more attention than the score itself.
3. The pattern for one customer
One bad rating is one bad reply. Two Not Good ratings from the same customer within a month, or a Not Good from someone who had always rated Great, is a change in the relationship.
Acting on low ratings with workflows
Help Scout suggests using workflows to follow up on ratings, for example flagging poor ratings that include a comment. Its workflow conditions and actions page lists the available conditions and actions for building that. A practical setup:
- An automatic workflow on conversations rated Not Good.
- Actions: add a tag such as
rating-not-good, assign to a team lead, and move the conversation back to active so it reappears in the queue. - A view of
rating-not-goodconversations from the last 14 days, reviewed every few days.
The follow-up is a reply to what the customer said, from a person, without a discount attached. If the rating was about a cancellation that went badly, the follow-up is to fix it, not to win them back.
Testing whether ratings predict churn for you
Do not assume. Check it with your own data.
Get the ratings out
The Inbox API has a Get Satisfaction Rating endpoint and reporting endpoints including Happiness Ratings, which encodes ratings as 1 (Great), 2 (Okay) and 3 (Not Good). Per Help Scout, the reporting endpoints are available on Plus and Pro plans. Collect, for each rating: customer email, conversation, rating, comment and date.
Join with cancellations
From your billing system, add whether and when each customer cancelled. For each customer, take their ratings in a quarter and compare:
| Group in the quarter | Customers | Cancelled within 90 days |
|---|---|---|
| At least one Not Good with a comment | your count | your count |
| At least one Not Good, no comment | ||
| Only Great or Okay | ||
| No ratings at all |
Always include the “no ratings” row. It is usually the largest group, and leaving it out makes ratings look more predictive than they are.
Read the result carefully
- Small numbers. If the Not Good group has twelve customers, a difference of two cancellations is noise. Show counts next to every percentage.
- Correlation, not cause. A bad rating and a cancellation may both come from the same broken feature. The rating is a flag to look closer, not an explanation. See finding what predicts churn.
- Ratings on cancellation conversations. Exclude them, or analyse them separately. A Great on a smooth cancellation is a good outcome, and a Not Good on a cancellation measures your process, not future churn.
What to change if ratings do predict churn
If customers with commented Not Good ratings cancel clearly more often:
- Make the Not Good review a fixed weekly task with an owner.
- Track the comments by theme (bug, price, missing feature, support speed) in a custom field or tags.
- Send the top themes each month to whoever can fix them, with the customers’ own words.
- Re-run the comparison next quarter.
If they do not, keep using ratings for what they are good at, coaching and support quality, and look for churn signals elsewhere: repeat contacts, billing conversations, failed payments, usage.
Doing this with NoChurn
NoChurn’s Help Scout connector is read-only: it reads conversations and never changes ratings, tags or anything else in Help Scout. Conversations are matched to your customers by email; doubtful matches are suggested for you to confirm and never merged automatically. They feed a churn score from 0 to 100 and a churn drivers view that ranks attributes by correlation with churn, with sample sizes shown, and serve as dated, verbatim evidence for disputes on Stripe and Braintree. Rating collection, the Happiness report and follow-ups stay in Help Scout. See NoChurn + Help Scout.