CSAT in Zendesk: How Satisfaction Surveys Work

This guide is the plain-English tour of how the Zendesk CSAT survey actually works: when and how it's sent, the rating scales Zendesk offers (the classic Good/Bad and the newer expanded scales), how the score is calculated, where to report on it, and how to customize the survey so people actually respond. We'll finish with best practices, the mistakes that quietly tank your response rate, and where AI fits. The mechanics here are verified against Zendesk's own documentation.

What CSAT is — and why it matters

CSAT (Customer Satisfaction) measures how satisfied a customer was with a specific support interaction. In Zendesk, it works at the ticket level: after a ticket is resolved, the requester is asked to rate the experience, and those individual ratings aggregate into your CSAT score — a percentage from 0 to 100.

It's worth being precise about what CSAT does and doesn't tell you, because it's easy to confuse with other metrics:

• CSAT is transactional. It rates this ticket, this interaction — not your brand overall. That's different from NPS (Net Promoter Score), which asks about likelihood to recommend the company as a whole.
• CSAT is an outcome metric, not an effort metric. It tells you whether the customer was happy, but not why. The "why" comes from the reason and comment fields on a bad rating (more on those below).
• CSAT pairs with your speed metrics. A great SLA attainment number with a sliding CSAT score is a warning sign: you're fast but not actually helping. CSAT is the quality counterweight to your time-based KPIs, which is why it earns a spot in the metrics every support team should track.

In short: CSAT is the one number that comes straight from the customer's mouth rather than from your queue.

How Zendesk sends the survey

The single most important thing to understand is the trigger: the CSAT survey is only sent for tickets that reach the Solved status. Open, Pending, and On-hold tickets are never surveyed — Zendesk waits until the agent (or an automation) marks the ticket Solved, because that's the moment there's actually an experience to rate.

From there, a Zendesk automation (a time-based business rule) does the sending. The defaults differ by channel:

• Email. The survey is delivered as an email roughly 24 hours after the ticket is set to Solved. The 24-hour delay is deliberate — it gives the customer a beat to confirm the fix really worked before they're asked to rate it, and it avoids reopening-and-re-solving churn from immediately triggering a second survey.
• Messaging (web, mobile, and social). The survey is presented in the conversation itself, immediately after the ticket is solved, as a quick in-thread rating rather than an email a day later.

A few mechanics that trip people up:

• You must activate at least one channel. CSAT won't send anything until you turn on (activate) a channel in the Admin Center; activating a channel is what switches on the underlying business rule.
• The link stays open for 28 days by default. Customers have a generous window to respond, and they can rate the ticket whether it's still Solved or has since auto-Closed.
• It's one survey per solved ticket under the default rule, so a ticket that reopens and is re-solved doesn't spam the customer with repeats.

Because the timing lives in an automation, it's fully editable — you can shorten or lengthen the delay, or change conditions. We'll come back to that under customization.

Automation vs. trigger: the default 24-hour delay vs. sending immediately

This is the single most useful piece of plumbing to understand, because it's where teams accidentally double-survey their customers.

By default, the email survey is fired by a system automation — Zendesk ships one called "Request customer satisfaction rating (system automation)" — which is a time-based rule. That's exactly why the email arrives roughly 24 hours after the ticket is set to Solved: automations evaluate on a clock, so the survey waits a beat before going out. For most teams the 24-hour default is the right call and you can leave this automation alone.

If you'd rather the survey go out the instant a ticket is solved, you don't edit the automation — you build a trigger instead. Triggers fire immediately on a ticket event (here, the status changing to Solved) rather than on a timer. The Zendesk workflow is to create a trigger with two conditions — status changed to Solved and satisfaction is Unoffered — plus the actions to offer the satisfaction rating and notify the requester.

The catch — and the part people miss — is that you must deactivate the default "Request customer satisfaction rating (system automation)" first. If you leave it active and add the immediate trigger, the customer gets surveyed twice: once instantly from your trigger and again 24 hours later from the automation. Turn the system automation off so exactly one survey goes out per solved ticket.

The same trigger-and-automation conditions are also how you exclude tickets, users, or groups from being surveyed at all. A common pattern is to add a condition (or tag) so that, for example, users tagged partner never receive a survey — or to skip trivial, auto-resolved tickets. Anything you can express as a ticket condition or tag can gate who does and doesn't get asked to rate.

The rating model: Good/Bad and expanded scales

Zendesk's classic survey is deliberately blunt. It asks a single question — "How would you rate the support you received?" — and offers two answers: Good or Bad. That binary is the foundation of the whole system, and it's why CSAT in Zendesk has historically been a positive-vs-negative measure rather than a 1-to-5 average.

That said, Zendesk now supports more than two buttons. There are two layers to know about:

Numeric rating scales. Beyond the plain Good/Bad, you can configure a survey on a 1–2, 1–3, or 1–5 scale. The key detail — and the thing most people get wrong — is that every scale still collapses to Good or Bad under the hood for scoring purposes. Zendesk maps them like this:

So a 1–5 survey gives the customer more nuance to express, but Zendesk's headline CSAT score is still computed from the good/bad buckets above — a 3 out of 5 lands in the "Bad" bucket.

Customizable CSAT. Zendesk has also rolled out a newer Customizable CSAT experience (introduced through an early-access program) that goes further: you can choose 2, 3, or 5 rating options, display them as numbers, emoji, or plain text, add your own labels to each option, and attach richer follow-up questions. This expanded experience is gated to higher plans — generally Support Professional / Suite Growth and above — so what you see in your account depends on your tier. If you only have the classic Good/Bad toggle, that's why.

On any version, a negative rating can prompt a follow-up: the customer is shown a drop-down of preset reasons (for example, "took too long," "didn't solve my problem") and an optional free-text comment. Those reasons and verbatims are where the real diagnostic value lives — the rating tells you that someone was unhappy; the reason tells you what to fix.

How CSAT is calculated

The formula is simple, and it follows directly from the Good/Bad model:

CSAT score = (number of Good ratings ÷ total number of ratings) × 100

A few things this formula makes clear:

• Only responses count. The denominator is total ratings received, not surveys sent. Tickets that were surveyed but never rated don't drag the score down — they simply aren't in the math. (That's also why a low response rate makes the score noisy — more on that below.)
• Only solved tickets are in scope. Since surveys only go out on solved tickets, your CSAT score only ever reflects resolved work.
• "Good" is the whole positive bucket. On a 1–5 scale, a 4 and a 5 both count identically as one Good rating; there's no weighting for a 5 over a 4 in the standard score.

A worked example: if 200 customers responded to your survey last month and 170 gave a Good rating, your CSAT is (170 ÷ 200) × 100 = 85%. For context, a result in the mid-to-high 80s is commonly cited as a rough benchmark for a healthy support team — but "good" varies widely by industry, audience, and how you survey, so treat any external number as a loose reference rather than a target. What matters far more is your own trend over time.

Reporting on CSAT in Explore

Individual ratings are useful, but the value is in the trend. Zendesk surfaces CSAT in two places:

• The Support dashboard gives quick, agent-level satisfaction numbers for day-to-day visibility.
• Zendesk Explore is where the real analysis happens. Explore ships a dedicated Satisfaction dataset and prebuilt dashboard, so you can break CSAT down by agent, group, channel, ticket tag, time period, or custom field — and combine it with your other metrics in custom reports.

This is where CSAT stops being a vanity number and becomes operational. Filtering CSAT by tag shows you which topics generate unhappy customers; slicing by agent (handled fairly — small samples are noisy) supports coaching; trending it against your response and resolution times reveals whether speed and quality are moving together or pulling apart.

Customizing the survey

Because CSAT is built from standard Zendesk pieces — a survey configuration plus an automation — most of it is editable in the Admin Center:

• Timing. The 24-hour email delay lives in the CSAT automation. Lengthen it for products where the customer needs time to confirm the fix; shorten it for fast, simple interactions where the memory fades quickly.
• Channels. Turn email and/or messaging on independently, depending on where your customers actually are.
• The question and scale. On the customizable experience you can reword the headline question, switch the scale (2/3/5 options), and choose numbers, emoji, or text.
• Reasons for a bad rating. Define the drop-down list of reasons a customer picks from after a negative rating. Keep this list short and mutually exclusive so the data stays clean — these reasons are the most actionable thing CSAT produces.

A note on visuals: the Admin Center CSAT settings and the end-user survey screen are behind a login, so we're describing them precisely rather than showing fabricated screenshots. When you open Admin Center, the customer-satisfaction settings live under the Tickets/objects area, and the customer-facing survey renders either as an email landing page (with the rating buttons, optional comment, and reason drop-down) or as an in-thread message flow on messaging channels.

Plan availability

CSAT is included on Zendesk's paid Support and Suite plans — it's not an add-on you buy separately — but the flavor you get depends on your tier. The classic Good/Bad survey is broadly available; the newer Customizable CSAT (multi-option scales, emoji, custom labels, richer follow-ups) requires higher tiers, generally Support Professional / Suite Growth and up. Plan packaging changes over time, so confirm what your specific plan includes in your own account before you build a workflow around it.

Best practices

A few habits separate teams that get useful CSAT data from teams that get noise:

• Keep it short. The classic one-tap Good/Bad survey exists for a reason — every extra required field costs you responses. Ask for the rating first; make the comment optional.
• Mind the timing. Survey while the interaction is fresh but after the customer has had a chance to verify the resolution. The 24-hour email default is a sensible starting point; tune it to your product.
• Always collect a reason on bad ratings. A Bad with no reason is a dead end. A Bad with "took too long" pointed at a specific tag is a fixable process problem.
• Act on the verbatims. CSAT only improves anything if someone reads the comments and changes something. Route low ratings to a follow-up workflow or a manager so unhappy customers get a second touch.
• Pair it with your speed metrics. Watch CSAT alongside first reply time and resolution time so you can see quality and speed together, not in isolation.

Common mistakes

• Reading a tiny sample as gospel. If only a handful of customers responded, one Bad rating swings the percentage wildly. Look at response volume before you react to a CSAT move, and don't grade an agent on five ratings.
• Letting the response rate stay low without trying to fix it. Long, multi-question surveys, bad timing, and surveying through a channel customers don't use all suppress responses. A low response rate doesn't just shrink the sample — it biases it, since the people who bother to respond often skew to the extremes.
• Surveying too often. Blasting a survey after every interaction — including trivial auto-resolved tickets or repeated re-solves of the same ticket — causes survey fatigue and trains customers to ignore them. Survey meaningful resolutions, not every status change.
• Treating the score as the goal. CSAT is a thermometer, not a thermostat. Chasing the number (or worse, gaming it by not surveying hard tickets) defeats the point; improving the experience is what moves it honestly.
• Ignoring the "Bad" reasons entirely. The single highest-value output of CSAT is the structured reason on negative ratings — and it's the most commonly ignored.

Where AI fits

CSAT is also one of the clearest mirrors for how well your AI is actually performing. If an AI agent layer like Macha is resolving routine tickets on top of your Zendesk, those resolutions get surveyed exactly like a human's — so CSAT becomes an honest, customer-graded scorecard for automation quality, not just a marketing claim. Macha sits on top of Zendesk (it's not a help desk and not a Zendesk replacement), so the tickets it handles flow into the same Solved status and the same CSAT survey your agents' tickets do.

There's a second angle, too: AI is good at the part of CSAT teams usually skip — reading the verbatims. Hundreds of free-text comments and bad-rating reasons can be summarized, clustered into themes, and tagged automatically, turning a pile of complaints into a ranked list of what to fix. In Macha's model you pay per AI action (any automated step — summarize a batch of comments, tag a ticket, draft a reply, resolve — costing 0.5–9 credits depending on the model you choose), not per closed ticket, because most of this work is analysis and automation rather than a single "resolution." If your CSAT is telling you something but you don't have time to read every comment, that's exactly the gap. You can try it free — 7-day free trial, no credit card required.

Frequently asked questions

How does the Zendesk CSAT survey get sent? A Zendesk automation sends it when a ticket reaches the Solved status. For email, the survey goes out roughly 24 hours after the ticket is solved; for messaging channels, it appears in the conversation immediately after solving. At least one channel has to be activated for any survey to send, and the email rating link is valid for 28 days by default.

What rating scale does Zendesk CSAT use? The classic survey is a binary Good / Bad ("How would you rate the support you received?"). Zendesk also supports 1–2, 1–3, and 1–5 numeric scales (all of which map to Good/Bad for scoring), and a newer Customizable CSAT experience on higher plans that allows 2/3/5 options shown as numbers, emoji, or text with custom labels and follow-up questions.

How is the CSAT score calculated in Zendesk? CSAT = (number of Good ratings ÷ total number of ratings) × 100. The denominator is responses actually received, not surveys sent, and only solved tickets are ever surveyed. On a 1–5 scale, 4s and 5s count as Good while 1–3 count as Bad.

Which tickets get surveyed? Only tickets that reach the Solved status. Open, Pending, and On-hold tickets are never surveyed — there's no experience to rate until the ticket is resolved.

Where do I see CSAT results? Quick agent-level numbers appear on the Support dashboard, and full analysis lives in Zendesk Explore, which has a dedicated Satisfaction dataset and prebuilt dashboard for slicing CSAT by agent, group, channel, tag, and time.

Is CSAT available on all Zendesk plans? CSAT is included on paid Support and Suite plans (not a separate add-on), but the Customizable CSAT experience — expanded scales, emoji, custom labels — requires higher tiers, generally Support Professional / Suite Growth and up. Confirm specifics in your own account, since plan packaging changes.

The bottom line

Zendesk CSAT is the customer's verdict on your support, captured in one tap. The mechanics are straightforward once you see them: an automation fires a survey on Solved tickets (about a day later by email, instantly in messaging), the customer answers a Good/Bad question — possibly on an expanded 1–5 or customized scale that still rolls up to good vs. bad — and Zendesk computes (Good ÷ total responses) × 100. The score itself is just the headline; the real value is in pairing it with your speed metrics, trending it in Explore, and actually acting on the reasons behind every Bad. Keep the survey short, time it well, read the verbatims, and CSAT stops being a number on a dashboard and starts telling you exactly where to get better.

CSAT mechanics verified against Zendesk's official documentation, June 2026. Zendesk updates its product periodically — confirm specifics (especially plan availability and the customizable-CSAT rollout) in your own account before relying on them.