At the top US-based company I work with, our CSAT numbers were solid for most of last year. Consistently above 85%, agents doing careful work, response times within SLA. Then the retention data landed and told a different story: customers who had opened multiple support tickets in a quarter were churning at roughly 20% higher rates than customers who had barely needed support at all. Not because the interactions were bad. Because the experience of needing help repeatedly was quietly exhausting them.
We were measuring the quality of individual conversations and calling it good. What we were not measuring was how hard it was to get to a resolution in the first place. That is a different question entirely, and it has a different metric.
What CSAT Actually Tells You (And Where It Stops)
Customer Satisfaction Score is a rating of the last interaction. Did the agent communicate clearly? Was the information correct? Did the customer feel heard? A well-run CSAT survey catches real problems at the conversation level: rude tone, inaccurate answers, missed follow-through. If your CSAT is consistently below 70%, you have agent quality issues worth addressing urgently.
But CSAT has a ceiling it cannot see through. A customer can rate a conversation 5 out of 5 and still feel like reaching resolution cost them too much. They had to call back three times about the same billing issue. They searched the knowledge base, found an article that was two product versions out of date, then opened a ticket anyway. They were transferred between two agents before reaching someone who could actually act. Each of those interactions may have been individually fine. The arc of the experience was grinding.
CSAT tells you whether the last conversation went well. It tells you almost nothing about whether the path to that conversation was reasonable. That distinction matters enormously for retention, because customers do not evaluate their experience one ticket at a time. They evaluate the cumulative effort it took to get sorted.
Customer Effort Score: The Metric That Predicts What CSAT Misses
Customer Effort Score (CES) asks one question after a support interaction: "How easy was it to resolve your issue today?" Usually on a scale of 1 to 7, where 7 means very easy. That is the entire survey. The simplicity is deliberate.
The research behind CES comes from the Corporate Executive Board (now Gartner). Their finding, published originally in the Harvard Business Review and later expanded into the book The Effortless Experience, was counterintuitive: reducing customer effort predicted loyalty better than satisfying customers or even delighting them. Customers who described getting help as high-effort were significantly more likely to reduce spending and actively share negative word of mouth. Customers who said it was easy returned even when the interactions were unremarkable.
The implication for support teams is pointed. Investing in agent warmth and friendliness training, without addressing the process friction around those agents, leaves your churn risk largely undetected. Warm agents inside a friction-heavy process still generate low CES scores and elevated churn. You see it in the CSAT data as teams doing great work. You see it in the retention data six weeks later.
Where the Gap Shows Up on the Support Desk
Once you start tracking CES alongside CSAT, the patterns that were invisible become obvious. A few I see regularly:
- Multi-contact tickets. A customer opens a ticket, gets a partial answer, replies to follow up, gets transferred, replies again. Each contact is handled correctly. The customer's effort score is 2 out of 7. CSAT, measured only at close, shows a 4. The repeat contacts show up in volume data as "normal churn," because nobody connected them to effort.
- Channel switching. Customer starts on live chat, gets directed to email for account verification, then calls in because the email took 24 hours. Each agent handles their segment professionally. The total experience involved three channel transitions the customer did not expect or want.
- Stale self-service. A customer searches the help center, finds an article, follows the steps, the steps are wrong because the product changed six months ago. They open a ticket. The agent solves it in one reply. CSAT is fine. CES on the overall resolution path is not, because the self-service sent them the long way around.
- AI tier-1 friction. An automated response requires the customer to rephrase their question twice before the bot understands and routes correctly. Fast and accurate once it routes. High-effort to get there. The latency of reaching a real answer is what CES captures that CSAT does not.
- Tracks: was the last conversation good?
- Catches agent quality and tone issues
- Score can be healthy while customers churn
- Silent on process friction and repeat contacts
- Tracks: was the experience easy to navigate?
- Surfaces friction in routing, self-service, handoffs
- Predicts churn before it shows up in retention data
- Directs improvement to the right layer of the stack
Adding CES to Your Stack Without Rebuilding Anything
If you are already sending a post-ticket CSAT survey, adding CES is one question appended to the same send. You do not need a new tool or a new survey platform.
In Zendesk: the CSAT trigger fires automatically after ticket close. In the satisfaction email template, add a second question below the rating: "How easy was it to get your issue resolved today?" with a 1 to 7 link scale. Both responses feed into a custom ticket field you can then filter in Explore. The whole setup takes under an hour if your trigger logic is already clean.
In Help Scout: the happiness rating lives in the conversation sidebar, and you can add a follow-up question directly in the satisfaction survey settings. CES data comes back in the Happiness report, filterable by mailbox and date range. Tag your tickets by type first so you can slice effort scores by category later.
The analysis segments that surface the most useful gaps:
- CES by channel (email, chat, phone, self-service). Self-service almost always underperforms on effort relative to direct contact, because customers who reach self-service often failed somewhere else first.
- CES by ticket type (billing, onboarding, technical, access). Billing issues tend to run high-effort because they often involve account verification steps. Identifying that pattern allows you to simplify those verification flows specifically.
- CES trend over 90 days. A single point-in-time CES score is moderately useful. A trend line tells you whether process improvements are landing or whether new friction is accumulating.
Reading CSAT and CES Together
The two metrics tell different parts of the story, and the combination is more diagnostic than either one alone.
CSAT high, CES high: Agents are delivering quality interactions and the experience is easy to navigate. This is the target state.
CSAT high, CES low: Individual conversations are good but the process around them is creating friction. Fix the process, not the agents. Common causes: poor knowledge-base coverage, unclear escalation routing, too many verification steps, self-service that routes customers to outdated articles.
CSAT low, CES high: Getting help is easy but the help itself is not landing well. Fix the coaching. Common causes: accuracy gaps, tone issues, agents closing tickets before confirming the customer's underlying concern is resolved.
CSAT low, CES low: Both the interaction quality and the experience architecture need attention. Start with agent coaching since that directly affects both signals, then map the friction points in the process.
The reason I track both is that CSAT alone tells me our agents are doing good work, which is true. CES tells me where the process is quietly failing the customers those agents are trying to serve. You cannot fix what you do not measure, and healthy CSAT is a genuinely unreliable proxy for healthy retention.
The customers most worth retaining are often the ones who needed the most support early in the relationship. If your support experience is friction-heavy, they are the first to leave and the hardest to win back. One question added to a survey you already send is a low-cost way to see that before it shows up as churn.
Want to talk through your support metrics stack or how your team is building toward better retention? Reach out here.
Comments