CSAT Follow-Up Action Plan: Copyable Template

Start with the exact experience the question covered. A rating after a support conversation may describe the explanation, the waiting time or the unresolved product problem. It does not automatically identify which of those needs changing. Read the accompanying words and keep the customer's observation separate from your explanation of it.

Record satisfied, neutral and unsatisfied counts using the same classification as the calculator. Include the number of valid ratings in the denominator, and record invitations separately if available. Someone who did not answer the rating question is not a neutral response. A score from respondents does not describe the experience of everyone invited.

Write down the collection window, audience and invitation timing. Before interpreting a change, check whether a different group answered or a new survey placement changed who saw the question. When the setup changed, label the result as a new baseline or a limited comparison. Do not call it a proven improvement from an action your team took.

Complete one record for a defined experience and review window. An investigation can begin with a small, clearly described sample, but it should not turn that sample into a claim about all customers. Identify which comments support the proposed problem and which disagree with it. Preserve contradictory evidence so the owner can test a real question.

Choose a next action that produces an observable result: check a confusing instruction, inspect the failed task with sample data or compare the queue timestamps with the respondent's description. 'Improve CSAT' is a goal, not an assigned action. Record the person who will bring back the finding and the review checkpoint they can maintain.

Keep service recovery separate from product prioritization. A customer may need immediate help even if a broader product change requires more evidence. Conversely, a polite support reply does not resolve a repeated workflow problem. Link the two work streams without using a public roadmap status as a substitute for either decision.

Use this reply only when a follow-up is appropriate in your existing customer conversation. Read the original comment first and remove questions it already answers. Start with an acknowledgement of the experience, then ask for the one observation that will help choose the next check. A low rating alone does not tell you what went wrong.

Ask the question without supplying a preferred explanation. 'Which part took longer than you expected?' leaves room for a useful answer; 'Was our slow export the problem?' introduces a cause the customer may never have reported. If they cannot recall the detail, preserve that limit and investigate the evidence already available.

A response is optional. Do not make help conditional on changing a rating or answering a research question. Keep account details and private evidence in the support channel. A public request can describe a suitable product problem without copying the customer's identity or full survey response.

This fictional example concerns an imaginary reporting app. Ten people rate support after asking about report delivery: six satisfied, one neutral and three unsatisfied. The CSAT result is 60%. Two unsatisfied comments say the agent was helpful but the report still required a manual download. The third describes a long wait. These comments support different investigation questions.

The reviewer keeps the waiting-time issue with the support owner and investigates the delivery task separately. A follow-up asks when the report was needed and who had to download it. Two respondents describe a weekly handoff, but the reviewer has not checked whether they represent two accounts or two people on one team. The plan records that uncertainty rather than inventing customer reach.

The product owner checks the handoff with sample data and reviews the existing request evidence before proposing a change. At the next checkpoint, the team can report what it learned even if it has not selected a solution. A delivery request remains under review until an actual decision changes its status.

For the next complete survey window, eight of ten respondents are satisfied, or 80%. That is a 20 percentage-point increase from 60%. With only ten ratings in each window, changing one rating between satisfied and another category changes the score by ten points. The new group may also differ. The team reports the counts and setup; it does not claim the investigation caused the increase.