A negative rating lands at 10:14 a.m. The conversation is closed, the assigned agent has moved to another queue, and the customer’s comment says only: “Still not fixed.”
What happens by 10:30?
Supportman alerts your team in Slack the moment a negative rating lands, so you can respond while the customer is still listening.
A useful negative-CSAT process answers that question in advance. It turns the rating into a trigger with an owner, a triage decision, a customer response, and a finish line.
1. Define the negative-rating trigger
Route each negative rating into a queue or Slack channel that somebody is responsible for during staffed hours. The alert should contain:
- The rating and any comment
- A direct conversation link
- The customer and account
- The assigned agent and current conversation status
- The time the rating arrived
- A way to claim ownership
Then set internal service levels. There is no universal target: a 24/7 incident desk and a weekday support team should not publish the same promise. The following are practical starting points to test against your coverage and case volume:
| Rating context | Ownership target | Customer-contact target |
|---|---|---|
| Security, safety, material financial harm, or widespread outage | Use the existing incident path immediately | Within 15 staffed minutes, if contacting the customer is safe and appropriate |
| Customer is blocked and the issue is still open | Within 30 staffed minutes | Within 1 staffed hour |
| Issue is resolved but the experience failed | During the same shift | Within 4 staffed hours |
| No comment or usable context | During the same shift | After review; contact is a decision, not an automatic step |
Define “staffed hour” and the after-hours handoff in the runbook. Otherwise a precise target hides an unowned overnight queue.
Speed helps only when it leads to useful action. Research on digital complaint handling found that faster responses increased satisfaction across complaint objectives, but an instant generic apology still leaves the work undone. Read the response-time study.
2. Triage before contacting the customer
The reviewer should read the full conversation, not just the rating card. In five minutes, establish:
- Current state: Is the issue open, partly resolved, or resolved?
- Impact: What can the customer not do, and is the impact getting worse?
- Failure: Was the problem the answer, delay, tone, handoff, policy, product, or outcome?
- Promises: Is an update, refund, callback, or other commitment already overdue?
- Risk: Does this belong in a security, safety, legal, billing, or incident workflow?
- Authority: Can the recovery owner fix it, or is an approval or specialist required?
Write the triage result in one sentence: “Customer still cannot export March transactions; the prior reply answered a different question; no sensitive-data risk; data-support specialist can resolve.” That is enough to route the case without conducting a full retrospective in the alert thread.
Do not treat the assigned agent as the cause by default. The rating may reflect a product limitation, policy, queue delay, or broken handoff. It is evidence to review, not a performance verdict.
3. Assign the right response lane
Use a small set of response lanes so every alert does not become a debate:
| Lane | Use when | Owner |
|---|---|---|
| Direct recovery | The issue and remedy are within normal support authority | Assigned agent or DSAT reviewer |
| Specialist recovery | Billing, data, or another specialist must act | One named support owner coordinates; the customer is not passed around |
| Manager recovery | A repeated failure, broken commitment, sensitive relationship, or conduct concern needs added authority | Support lead or manager |
| Review only | The rating is anonymous, abusive, unrelated, too old to act on, or contact could create privacy or safety risk | Reviewer records the reason for no contact |
Claiming an alert means owning coordination through closure. It does not mean personally completing every specialist action.
4. Send a specific first response
The first response has four jobs:
- Show that you read the conversation.
- Name the specific impact.
- State who owns the next action.
- Give a credible update time.
I reviewed the conversation and can see that we sent instructions for a full export when you needed only the filtered March transactions. I’m sorry we cost you another round of troubleshooting. I’m coordinating this with our data-support specialist and will update you by 11:30 a.m., even if I am still waiting on the completed file.
Do not ask the customer to repeat information already in the thread. Do not lead with company constraints, speculate about the cause, or promise a resolution time you do not control.
If the customer’s description is accurate but the requested outcome is unavailable, acknowledge the impact without pretending otherwise. If their conclusion is inaccurate, correct it after demonstrating that you understood the underlying problem.
5. Work the case to closure
A recovery case stays open until one of these conditions is recorded:
- The requested action is complete and the customer has been told.
- An alternative or remedy has been delivered and explained.
- The company cannot provide the requested outcome, and the final answer plus available options have been communicated.
- The customer does not respond after the team’s documented follow-up sequence.
- Further contact is inappropriate, with the reason recorded.
Before snoozing the case, record:
- The named recovery owner
- The action currently in progress
- The next customer-update time
- The dependency and its internal owner
Update the customer when promised even if there is no resolution yet. Silence after a deadline is a second service failure.
Choose a remedy that fits the failure: correct the answer, complete the missing action, remove repeated effort, expedite an escalation, provide an authorized refund or credit, or offer the closest workable alternative. For a fuller treatment of interactional, procedural, and distributive fairness, see The Service Recovery Paradox Is Not a Strategy.
6. Record the outcome
Keep the per-rating record short enough that reviewers actually complete it:
- Primary failure category
- Open or resolved when rated
- Recovery owner and response lane
- Time to ownership and first customer contact
- Remedy or final answer
- Recovery status
- Follow-up action, if any
Do the customer recovery first; run deeper trend analysis separately. DSAT Is an Operations Signal, Not Just an Agent Score covers cause taxonomies, controllability, recurrence, and routing systemic fixes.
If agent behaviour contributed, coach from the conversation and the QA rubric—not from the rating alone.
A worked negative-CSAT case
Imagine a customer leaves a negative rating after asking for an invoice with their new legal entity name. The first reply links to instructions for changing future invoices, but the customer needs a corrected invoice for a filing deadline tomorrow.
- Alert: The rating, “This article doesn’t fix the invoice,” appears in the DSAT channel with the closed conversation attached.
- Triage: The reviewer marks the issue unresolved and time-sensitive, confirms that billing operations can regenerate the document, and identifies no security or payment dispute.
- Assign: A support lead claims the alert and uses the specialist-recovery lane rather than transferring the customer to billing.
- Respond: The lead acknowledges that the article only affects future invoices, states that they are coordinating the corrected document, and promises an update within one hour.
- Close: The lead sends the corrected invoice, confirms the legal name, and records the resolution. The customer does not need to answer for the case to count as operationally complete.
- Learn: The case is tagged as an answer mismatch and documentation gap for the next pattern review. It is not treated as proof of an agent trend.
Negative-CSAT response template
Hi [name]—I reviewed your conversation and can see that [specific failure] meant [customer impact]. I’m sorry we created that extra work. I’m owning the follow-up from here. [Action or owner] is [next step], and I’ll update you by [time], even if the work is still in progress. You do not need to repeat [information already provided].
Delete any clause that does not apply. A short, accurate response is better than a complete-looking script.
Frequently asked questions
Should every negative CSAT rating get a reply?
Every rating should be reviewed, but not every rating requires contact. Anonymous feedback, abusive messages, unrelated ratings, stale cases, and privacy or safety concerns may justify review without a reply. Record the reason so “no contact” remains a decision rather than a missed alert.
How fast should we respond to negative CSAT?
Set targets by impact and staffed coverage. A reasonable starting point is ownership within 30 staffed minutes and customer contact within one staffed hour for an unresolved blocker; resolved experience failures can usually wait for same-shift review. Test those targets against your volume instead of promising a speed your team cannot sustain.
Who should own the response?
The assigned agent can recover routine cases when they have the authority and context. Use a specialist or manager lane when the case involves sensitive risk, a broken promise, repeated failure, or an action outside normal support authority. In every lane, one named person should coordinate the customer-facing follow-through.
Should we ask the customer to change their rating?
Usually no. Resolve the issue without making the recovery feel conditional or turning the conversation into a request for a better score. If your survey system lets customers update a rating, they can do so without pressure.
Catch the rating while the customer is still listening
Supportman sends selected negative Intercom ratings into Slack with the customer comment and conversation link attached, so the team can claim, triage, and act on each alert.
Turn every negative rating into an owned response workflow →