Zendesk CSAT Improvement Playbook: Lift Scores Without Gaming the Survey
CSAT is the most common quality metric in support. It is also the most commonly gamed. Teams tweak survey timing, suppress surveys after bad interactions, or celebrate high scores without asking what drives them.
This playbook takes a different approach. It starts with what actually moves CSAT — speed, resolution quality, and follow-through — and gives you a step-by-step process to diagnose issues and make improvements that stick. No survey tricks.
Why CSAT matters (and why it is not enough alone)
CSAT measures how the customer felt about a specific interaction. That makes it useful as a quality signal — but limited as a standalone metric:
- It is biased. Only a fraction of customers respond to CSAT surveys. Satisfied and very unsatisfied customers are over-represented; the middle is under-represented. See CSAT response rates: sample size, bias, and when to ignore the score for the nuance.
- It is lagging. By the time CSAT drops, the underlying problem (slow FRT, bad resolutions, unresponsive follow-up) has been happening for days or weeks.
- It does not explain “why.” A low CSAT score tells you customers are unhappy. It does not tell you whether the issue is speed, accuracy, tone, or something outside support’s control.
Use CSAT alongside reopen rate, first reply time, resolution time, and first contact resolution to get the complete picture. For the full KPI framework, see support metrics dashboard.
Step 1: Establish your baseline
Before trying to improve, understand where you are:
Response rate
Pull your CSAT survey response rate from Zendesk. If fewer than 15–20% of tickets receive a CSAT response, your sample is too small to draw conclusions from week to week. Focus on increasing response rates before optimizing the score itself.
Industry benchmarks for ecommerce sit around 76% CSAT (3.8 out of 5) on average, with best-in-class teams pushing above 85%. But your baseline is what matters — compare yourself to last month before comparing to the industry.
Score distribution
Look at the distribution, not just the average. In Zendesk, CSAT is typically “Good” or “Bad” (binary). If you collect scaled feedback (1–5 or 1–10), look at the distribution curve. A bimodal distribution (lots of 5s and lots of 1s) tells a different story than a normally distributed one (mostly 3s).
Segment by group, channel, and tag
CSAT averages hide the variation. Segment by:
- Agent group — Which teams have the highest and lowest CSAT?
- Channel — Chat and messaging often have different CSAT profiles than email.
- Tag or topic — Some ticket types (billing disputes, account cancellations) inherently produce lower CSAT regardless of agent quality. Knowing this prevents you from blaming agents for systemic issues.
Use the Zendesk CSAT report to pull these breakdowns.
Step 2: Find the root causes of low CSAT
Low CSAT usually traces to one of four root causes. Your Zendesk data can help identify which one dominates.
Root cause 1: Slow first reply
Customers who wait hours for a first response are already frustrated before the conversation starts. Check whether tickets with low CSAT correlate with high first reply time.
How to check: In Explore, cross-reference CSAT with FRT. If the negative CSAT cluster has median FRT significantly above your team average, speed is a driver.
What to fix: See the real cost of slow first reply time and how to report on first reply time for diagnosis and tactics.
Root cause 2: Incomplete resolutions
Customers rate the interaction “Bad” not because the agent was slow, but because the problem was not actually solved. The ticket was closed prematurely, the answer was wrong, or the customer had to follow up to get a real resolution.
How to check: Cross-reference CSAT with reopen rate. If tickets that were reopened (or have high touches per ticket) disproportionately receive negative CSAT, resolution quality is the issue.
What to fix: - Review tickets with negative CSAT and high reopen rate for patterns - Invest in agent training on the topics that produce the most reopens - Use the reopened tickets report to find repeat issues by tag and group
Root cause 3: Poor follow-through
The agent replies once, then the ticket goes quiet. The customer follows up days later. The ticket is eventually resolved, but the customer felt abandoned.
How to check: Look at reply time (not just first reply — subsequent replies too) and requester wait time. If customers wait a long time between responses after the first reply, follow-through is the issue.
What to fix: - Set internal targets for next reply time, not just first reply - Use the Zendesk reply time report to track follow-up speed - Create views or triggers that highlight tickets with pending customer replies older than X hours
Root cause 4: Issues outside support’s control
Sometimes customers rate the support interaction “Bad” because they are unhappy with the product, the policy, or the outcome — not the agent. A customer who is told “no” to a refund may rate CSAT low even if the agent handled the conversation perfectly.
How to check: Segment CSAT by tag and topic. If certain categories (cancellation, refund denied, feature request rejected) consistently have low CSAT, the issue is the outcome — not the service.
What to fix: - Separate these categories in your CSAT reporting so they do not drag down overall scores unfairly - Work with product and policy teams to address the underlying issues - Train agents on empathy and expectation-setting for difficult conversations - Do not suppress surveys on these tickets — that is gaming the metric
Step 3: Make targeted improvements
Based on your root cause analysis, prioritize fixes:
| Root cause | First action | Second action |
|---|---|---|
| Slow first reply | Fix routing and staffing for peak hours | Set FRT targets by priority |
| Incomplete resolutions | Review negative-CSAT tickets for quality patterns | Create training around top reopen topics |
| Poor follow-through | Set next-reply-time targets | Build views for aging pending tickets |
| Outcome dissatisfaction | Segment CSAT by topic for accurate measurement | Work with product/policy on root issues |
Do one thing at a time. A focused improvement in the dominant root cause will move CSAT more than scattered improvements across all four.
Step 4: Track improvement without gaming
What gaming looks like
- Sending CSAT surveys only on tickets you think will get good ratings
- Delaying surveys until frustration fades
- Agents asking customers to rate them positively before closing
- Suppressing surveys on complex or escalated tickets
- Celebrating score improvements without understanding what changed
What genuine improvement looks like
- CSAT improves and reopen rate stays flat or drops
- CSAT improves and first reply time improves
- CSAT improves and survey response rate stays stable (not declining because you are sending fewer surveys)
- CSAT improves in the segments that were previously weakest
Track these together in your support metrics dashboard so you can see whether score movement reflects real improvement or measurement distortion.
Review cadence
- Weekly: Check overall CSAT trend and response rate. Flag any week-over-week drop greater than 3 points for investigation.
- Monthly: Review CSAT by segment (group, channel, tag). Identify which root cause is currently dominant.
- Quarterly: Assess whether targeted improvements are producing sustained gains. Adjust focus if a different root cause has become the bottleneck.
Common CSAT improvement mistakes
- Optimizing the survey instead of the service. Changing when or how you ask the question does not improve the experience. Fix the root causes.
- Treating all negative CSAT equally. A “Bad” rating on a billing dispute is different from a “Bad” rating on a simple how-to question. Segment and diagnose — do not average.
- Ignoring response rate. If your response rate drops from 20% to 10%, your CSAT score may look better simply because casual respondents stopped answering. A shrinking sample is not an improvement.
- Setting CSAT targets in isolation. “Get CSAT to 90%” without context incentivizes gaming. Set CSAT targets alongside FRT, reopen rate, and resolution time targets so improvement is holistic.
- Blaming agents for systemic issues. If CSAT is low on a specific ticket type, the problem is usually process, policy, or product — not individual agent performance. Use segmented data to separate the two.
FAQ
What is a good CSAT score for a support team? It depends on your baseline and industry. For ecommerce, 78–85% is considered good, and above 85% is best-in-class. For B2B SaaS, targets are often higher because ticket complexity is lower per interaction. Start by improving from your own baseline rather than chasing an industry number.
How many CSAT responses do we need for reliable data? At minimum, 30–50 responses per segment per period to draw meaningful conclusions. If you have 200 tickets per week and a 15% response rate, that is 30 responses — barely enough for a weekly team-level view and not enough for agent-level analysis.
Should we use CSAT or CES or NPS? Each measures something different. CSAT measures interaction-level satisfaction. CES measures how easy it was for the customer. NPS measures loyalty and willingness to recommend. For support ops, CSAT is the most actionable at the ticket level. Use CES or NPS as supplementary signals. See customer effort score vs CSAT for a deeper comparison.
How do we increase CSAT survey response rates? Send the survey close to resolution (within hours, not days). Keep it short (one question + optional comment). Send it via the same channel the customer used. Do not gate it behind login or extra clicks. Higher response rates give you better data, even if the score initially looks worse.