A simple CX program design: closing the loop on CSAT

What is the two-loop CSAT design?

The two-loop CSAT design is a customer service feedback program built from an inner loop and an outer loop. The inner loop responds to one customer straight away, and the outer loop looks at all the responses together on a schedule and changes something about how the service works.

Most teams that send a satisfaction survey already have the first half of a loop, which is the question and the score. The two loops are what get added so that a low score reaches a person who can act on it, and so that the same complaint stops arriving month after month. In the example below the customer support team owns the follow-up and the review, the survey platform (Qualtrics, in this case) does the sending, the routing and the reporting, and the customer decides whether they want to be contacted. The design follows the idea of insight-driven action loops described by the Qualtrics XM Institute.

How does it work, step by step?

Read the flowchart left to right, or press Step through to follow one step at a time. Every step and every branch is also written out underneath.

The two-loop CSAT design

1. A case is resolved and the survey is triggered. Within two business days the support team triggers the CSAT survey, and the trigger carries the agent, customer and interaction details with it. The team owns this step because it is the only part of the program that knows a case has closed.
2. The platform creates the contact and sends the invitation. A contact is created in the survey platform's directory (the XM Directory, in Qualtrics) with that interaction data, and an invitation email goes out with a personal survey link. Nobody on the platform side has to touch it.
3. The customer completes the survey. Nothing in this phase asks the customer to do more than answer.
4. The customer decides whether they want a follow-up. The permission question is theirs to answer. A no sends a CSAT notification email to the CS lead and the agent, and the inner loop ends there for that customer.
5. The platform checks the score. With permission given, the second decision asks whether the answer was extremely or somewhat dissatisfied. A no sends the same notification, so a good score still reaches the person who earned it.
6. A dissatisfied customer gets a person. An action-required ticket goes to the CS lead and the agent, the team follows up to resolve the problem, the customer adds context, and the ticket is closed with the root cause and the resolution recorded on it.
7. Every response reaches the dashboard. Closed tickets and notification-only responses both feed the CX dashboard, so the outer loop sees the ordinary cases as well as the worst ones.
8. The team reviews and acts. On a regular cadence the team reads emerging trends and the main drivers behind the scores, and takes systematic action to improve the experience.
9. Great service is celebrated and the customer comes back. Outstanding experiences are shared publicly, in team channels and reviews. The outcome the design works toward is a repeat customer, and their next interaction starts the loop again.

Why send the survey within two business days?

The window is short because the follow-up only works while the customer still remembers the interaction and the problem is still current. A survey that arrives weeks later gets vaguer answers, and a follow-up about a case the customer has already stopped thinking about can feel like an intrusion. Two business days is long enough for the team to be sure the case is closed, and short enough that the customer's answer is about this interaction and not about the company in general.

Where do the two decisions sit, and why ask permission?

The permission decision comes first and belongs to the customer. The dissatisfied decision comes second and sits on the platform. Both have to be yes for a ticket to open. A low score alone does not entitle the team to call someone, because a customer can be unhappy and still not want a conversation, and a ticket should only go where a conversation is welcome.

A customer who declines is still heard. Their score and comment arrive as a notification to the CS lead and the agent and go to the dashboard, so they count in the outer loop even though nobody contacts them.

What does the outer loop review, and how often?

The outer loop looks at the CSAT data on the dashboard for emerging trends and the main drivers behind the scores, then takes systematic action to improve the experience. It also reads the root causes and resolutions written on the completed follow-up tickets, since those are the closest thing to a list of what is going wrong. Systematic action means a change to how the service runs, such as a clearer answer in a help article or a different handover between teams, as opposed to a coaching conversation about one agent.

The source says the review happens on a regular cadence without naming one. As a starting point I would use this rhythm:

• Weekly New dissatisfied responses and open follow-up tickets.
• Monthly Trends and the drivers behind the scores.
• Quarterly Pick the larger fixes, and celebrate the outstanding cases as regularly as the bad ones are discussed.

What is the minimum version of this program?

The smallest useful version keeps both loops and drops most of the machinery. That is enough to learn whether the follow-up is welcome and whether the same causes keep coming back, and the rest can be added once those two questions have an answer.

The reasoning behind starting small is in Should you build the whole thing, or start with an MVP?

How do you adapt it when the tool is not Qualtrics?

The design needs three things from whatever tools you use, and none of them is specific to Qualtrics.

If one of the three is missing, the manual version usually works for a while, with a person copying the closed cases across once a day. The step that matters most to keep is the handover to a named owner, because that is the point where a response becomes an action.

What do you need to run it?

Tick off what you already have. Whatever is left is the gap between the survey you send today and a loop that closes.

Which measures show it is working?

Three things are worth watching, and I would track them as trends from your own starting point rather than against anyone else's numbers.

Where does this come from?

I have not published a client result for this exact program, so the evidence on this site is for the parts it is built from.

Knowing what Qualtrics can and cannot do at scale is also what made the custom research platform possible in under a week. For the survey design itself, six pitfalls to avoid when designing a customer experience survey covers what to ask and what to leave out.