Create CSAT survey — measure customer satisfaction

CSAT (Customer Satisfaction Score) and NPS (Net Promoter Score) measure different things — but are often confused. CSAT measures satisfaction with a single interaction ("How satisfied were you with the support call?"), NPS measures loyalty to the brand overall ("How likely are you to recommend us?"). For most operational questions CSAT is the right choice.

The practical difference: CSAT values are volatile and react quickly to changes — a bad service shift shows within days. NPS is slow — loyalty changes over months. Anyone planning weekly NPS measurements harvests noise instead of insight. CSAT per touchpoint, NPS once per quarter across all customers — that is the usual split.

A common mistake is to pack both metrics into one survey. This dilutes both signals because the context collides. Separate the measurements: CSAT directly after the touchpoint with two to three questions, NPS in its own survey, ideally three months removed from the last interaction. Each metric then gets the right context.

The optimal timing for a CSAT survey is tighter than most teams think: between one and four hours after the interaction ends. Right afterwards the customer remembers details but is too emotionally involved for a reflective rating. After 24 hours the detail memory has faded and the rating becomes blurry.

For support tickets a trigger one hour after status "closed" works well. For sales calls two to four hours after the meeting ends. For e-commerce orders 48 hours after delivery — the customer has unpacked the product but not yet started using it. These windows maximise response quality and response rate at the same time.

Avoid a common anti-pattern: the "monthly customer survey" to all customers. It hits at most 5 percent response rate because most customers currently have no reason to judge. Touchpoint-triggered CSAT regularly reaches 25 to 40 percent — with much more relevant answers. A webhook from the helpdesk on "ticket closed" starts the form at the right moment.

A survey without follow-up is data collection for its own sake. The most valuable answer in a CSAT survey is the low rating — and that is exactly what many companies ignore because no automatic workflow is defined. By the time a human looks at the list, days have passed and the customer has churned.

The pattern is conditional logic plus webhook: a rating of 1 or 2 immediately triggers a Slack alert to the customer success manager, notifies the ticket owner by email and creates a follow-up ticket in the CRM. A rating of 3 launches a standard "what bothered you?", 4-5 triggers an automatic thank-you plus a Google-review request.

Important: react quickly. Service-recovery studies show that responding to a bad rating within 24 hours repairs the relationship in 60 to 70 percent of cases — after a week the rate drops below 20 percent. The "detractor recovery workflow" is therefore more important than the bare score display on a dashboard.

CSAT surveys face a trade-off: anonymous answers are more honest, identifiable answers are actionable. Anonymous you know that service quality is sliding — but not which customer is about to churn. Identified you can react — but customers often phrase more cautiously because they fear a direct consequence.

The pragmatic middle path is two-stage anonymity. Stage 1: score and free-text are stored with customer ID (you can react). Stage 2: aggregation for reports shows only aggregates without IDs, access to individual cases is restricted to the customer-success team. This gives you honest answers and actionability at the same time.

Communication is decisive here. Right on the form it should say what happens with the data: "Your answer helps us serve you better — your customer-success manager reads it personally". This transparency lifts the response rate because customers see the purpose. Compared to "anonymous survey" it feels more honest — true anonymity is hard to guarantee with a touchpoint trigger anyway.

A single CSAT value of 4.2 says almost nothing. Is that good or bad? Compared to what? The real value of CSAT only emerges from the trend view over weeks and months. A value dropping from 4.5 to 4.2 is an alarm — even if 4.2 in absolute terms still sounds passable.

For meaningful trends you need three ingredients: a consistent measurement interval (daily, weekly), a minimum sample (about 30 answers per period for statistical significance) and a segmentation axis (per product, per channel, per team). Aggregates without segmentation hide local problems — a slump in the support team is masked by strong sales ratings.

Visualization as a rolling 7-day average smoothes daily noise without hiding the trend. Mark releases, staff changes or pricing updates as annotations on the timeline — that way you immediately see whether a CSAT dip correlates with an internal change. The activity-log function provides the timestamps, the dashboard connects them to the CSAT line.