To measure satisfaction in Service Cloud, send a short survey when a case closes and store each response against that case. Then roll the scores up by agent, queue and account. Salesforce has its own survey tool, and many third-party survey products also write results back. The hard part is not sending surveys. It is deciding which metric to ask, how often to ask it, and who acts on a bad score.
Should we measure CSAT, NPS or CES?
Use CSAT for single interactions, CES for how hard the process felt, and NPS for the overall relationship. Most support teams start with CSAT on closed cases and add the others only when someone will act on them.
CSAT asks how satisfied the customer was with a specific interaction, usually on a short rating scale. It is the natural fit for case-level feedback because the customer is rating something recent and concrete. Customer Effort Score asks how easy it was to get the issue handled. It tends to expose process friction, such as being transferred twice or repeating account details.
Net Promoter Score asks how likely someone is to recommend you. That reflects the product, pricing and account relationship, not one agent's call. Sending NPS after every case blames support for things it does not control. Send it on a schedule to account contacts instead, and let customer success own the follow-up.
| Metric | Question it asks | When to send | How to act on it |
|---|---|---|---|
| CSAT (customer satisfaction) | How satisfied were you with the help you received? | Shortly after a case closes, for sampled cases | Coach agents, review low-scoring cases, fix the case reasons that score worst |
| CES (customer effort) | How easy was it to get your issue resolved? | After resolution, or after a self-service or bot session | Remove handoffs, repeated questions and broken knowledge articles |
| NPS (net promoter) | How likely are you to recommend us to a colleague? | On a set cadence to key contacts on each account, not per case | Route detractors to the account owner; feed the result into account health |
| Open-text comment | What is the main reason for your score? | Paired with any of the three | Theme the comments by reason, product and channel; review them in team meetings |
We do not publish target scores here. Benchmarks depend on scale, industry and how you sample, so set your own baseline from a few cycles of real responses.
What does Salesforce include for surveys, and what costs extra?
Salesforce Surveys comes with a small free allowance of responses, and larger volumes need a Feedback Management license. Check what your org has before you design anything.
Salesforce help documentation describes basic platform surveys with 300 responses by default. Above that, a Survey Response Pack adds blocks of 1,000 responses. Feedback Management Starter includes 100,000 responses, and Feedback Management Growth has no response cap. Starter and Growth add features such as merge fields, data mapping into records and sentiment insights on text answers. Growth adds customer lifecycle tracking.
Edition eligibility, industry-cloud bundles and packaging shift between releases. Ask your Salesforce account team which product is provisioned, and whether a Health Cloud or Financial Services Cloud license changes the limits.
Third-party survey tools are the other route. Several established survey vendors offer managed packages or connectors that write responses to Salesforce records. They suit teams that already run surveys outside Salesforce, or need survey design features the native tool lacks. Judge them on how cleanly each response lands on the right case and contact, not on the survey builder alone.
How do we trigger a survey when a case closes?
Use a record-triggered flow on Case that fires when the status changes to closed, then calls the Salesforce Surveys action to send an invitation. Salesforce also offers simpler setup options, but Flow gives you control over who gets asked.
Trailhead walks through this pattern. The flow runs when a case moves to closed and sets the survey subject to the triggering case. The invitation email goes to the case contact. Support Settings also has a Customer Feedback Survey field for a basic close-case survey, and paid licenses add survey invitation rules.
Some prerequisites catch teams out:
- Customers outside your company need an active Experience Cloud site assigned for survey invitations.
- The site guest user needs read access to surveys and invitations, and create access to responses.
- Invitation emails cannot be sent from contact, person account or lead records owned by a queue, according to Salesforce's survey considerations.
- Surveys stop working if the user who created them is deactivated, so build them under a stable admin account.
- Invitation email templates have limits on which merge fields they accept.
Put the exclusion logic in the flow rather than in someone's head. Skip cases closed as spam, duplicates or internal requests. Skip contacts with no email address or with a recent invitation already sent.
How do responses link to the case, contact, account and agent?
Link each response to the case through the survey subject, then copy the case owner, contact and account onto a record you can report on. The native objects do not do all of that by themselves.
Salesforce Surveys stores data in its own objects, including survey invitations, survey subjects, responses and question responses. The survey subject ties an invitation to a record such as the case. Salesforce notes that responses are not related to a contact or lead automatically. Anonymous guest responses are not tied to any record at all.
A practical pattern is a small flow that runs when a response is completed. It reads the score and writes it onto the case, or onto a custom feedback record. That record holds the case, contact, account, the agent who owned the case at close, the queue, channel and case reason. Capturing the agent at close matters, because ownership often changes after the survey goes out.
How often should we survey the same customer?
Less often than you think. Set a sampling rule and a quiet period per contact so busy customers are not asked after every case.
- Cap invitations per contact, for example one in any rolling period your team agrees on.
- Sample high-volume queues instead of surveying every closed case.
- Never survey on reopened cases until they close again.
- Pause NPS invitations to an account while a CSAT survey is outstanding.
- Record every invitation, sent or suppressed, so response rates are honest.
Over-surveying can lower response rates and tilt results toward customers with strong feelings. A flow can check the contact's last invitation date before sending. Store that date on the contact so the check stays cheap.
What should the survey itself ask?
One rating question and one open-text question. Anything longer belongs in a separate research survey, not a post-case check.
Keep the rating scale the same across channels so phone, chat, email and messaging scores compare fairly. Word the question about this case, not the company. Make the open-text box optional and ask for the main reason, which gives cleaner themes than a generic comments field. If you need case reason or product, read it from the case record instead of asking the customer again.
How do we close the loop on low scores?
Treat a low score as work. Create a follow-up task or case for a named owner, alert a supervisor, and track whether the customer was contacted.
The same response-completed flow can branch on the score. A low CSAT might create a task for the team lead on the original case. A detractor NPS might create a task for the account owner instead. Salesforce release notes also describe custom notifications for survey responses, which can alert users in the app. Check what your license supports.
Avoid reopening the original case automatically. It distorts resolution metrics and can trigger another survey. A linked follow-up task or a child case keeps the history clean.
How should we report on CSAT in Salesforce?
Report from the record that holds the score plus case context, grouped by queue, agent, case reason and channel. Show response counts beside every average.
Useful views include CSAT trend by queue, distribution of scores by agent, worst-scoring case reasons and channels, and open follow-ups on low scores. Small samples swing wildly, so set a minimum response count before ranking agents. Use agent views for coaching conversations, not public leaderboards. For building views executives actually open, see our dashboards article linked below.
Can AI summarize open-text survey feedback?
Yes, with limits. AI can group comments into themes and flag sentiment, but someone should review the themes before they drive decisions.
Feedback Management Starter and Growth list sentiment insights for text responses, according to Salesforce's license comparison. Other teams pass comments to a generative AI tool or a data platform for theme summaries. Either way, check a sample by hand. Sarcasm, product names and mixed feedback are easy to misread. For drafting replies, case summaries and agents, see our separate piece on AI in service teams.
What privacy and consent rules apply to surveys?
Surveys are customer communications, so they follow your email consent, data retention and regional privacy rules. Decide those before launch.
Respect opt-outs on the contact before sending. Say in the invitation how responses will be used and whether they are anonymous. If you promise anonymity, do not link the response to the contact. Restrict who can see free-text comments, which sometimes contain personal or health details. Regulated firms should also route survey emails through their archiving and supervision process.
A wealth advisory client of ours ran its feedback surveys from Account Engagement rather than Service Cloud. Three days after a planning meeting is marked complete, a flow-triggered journey sends the survey. Native survey forms mapped to client records replaced Google Forms. BCC-to-advisor settings fed the firm's archiving system to keep messages supervised.
How do survey scores feed account health?
Roll case-level CSAT and relationship NPS up to the account as separate fields, then use them as two of several health inputs. Neither should decide health alone.
Store a recent average and a response count on the account, refreshed on a schedule. Show the last detractor comment on the account page so a CSM sees it before a renewal call. Weight survey inputs lightly when an account has few responses. Our guide to Salesforce for customer success teams explains how to build the rest of the health score.
