Intercom CSAT Benchmarks: What’s a Good CSAT Score in 2026?

Your Intercom CSAT says 82%. Someone in a leadership meeting asks if that's good, and nobody in the room actually knows.

There's no official Intercom-wide benchmark, and anyone quoting one at you is guessing. Your score depends on your product's complexity, your conversation mix, whether Fin is answering half your queue, and which customers bother to tap a rating at all.

That doesn't mean you fly blind. Here's the operating framework I use:

CSAT	How I'd read it
90%+	Excellent
85–89%	Strong
80–84%	Healthy, with room to improve
70–79%	Worth investigating
Below 70%	Something needs attention now

These aren't Intercom's numbers. They're a working rule of thumb built from published help-desk benchmarks and the results high-performing Intercom teams share publicly.

For context: Zendesk pegs typical help-desk CSAT at roughly 64–80%. Documented Intercom customers regularly report 90% or better.

The score itself is the least interesting part. The real question is what's causing it, and whether your team sees those signals fast enough to do anything about them.

What is Intercom CSAT?

Intercom asks customers to rate a conversation after it ends. Those ratings feed the Surveyed CSAT report, which breaks out:

• Overall CSAT
• Per-teammate CSAT
• Fin AI Agent CSAT
• CSAT over time
• Individual ratings and customer remarks
• Survey request and response rates
• Topics that keep showing up in unhappy conversations

Intercom's own guidance: use overall CSAT as a baseline, then compare against your history, your internal targets, and external benchmarks.

The formula is positive responses ÷ total responses × 100. Get 92 positive ratings from 100 completed surveys and you're at 92%.

The math is the easy part.

So, what is a good Intercom CSAT score?

For most SaaS support teams, 85%+ is a strong target. Holding above 90% consistently is excellent.

But a headline number without volume behind it will lie to you. Two teams, same month:

• Team A: 94% CSAT, from 20 ratings on 1,000 conversations
• Team B: 89% CSAT, from 300 ratings on 1,000 conversations

Team A looks better on the dashboard. Team B is the one I'd trust.

Twenty ratings is a 2% sample of what actually happened in that queue. Intercom itself describes surveyed CSAT as incomplete and potentially biased, because the customers who respond are self-selecting, and the furious and the delighted are the most motivated to tap the button.

The teams that get real value out of CSAT read the score and the conversations behind it. Not one or the other.

Real-world Intercom CSAT benchmarks

The most useful calibration data is what named Intercom customers report publicly.

tado°: up to 90% CSAT

Smart-home company tado° runs Intercom and Fin through seasonal spikes where support volume can jump 400%. Intercom reports they hold up to 90% CSAT through peak season.

An earlier case study has them improving overall CSAT from 79% to 87%, with chat satisfaction hitting 90%. That's a useful real-world scale: 79% was the improvement project, 87% was strong, 90% was the ceiling worth bragging about.

Bailey Nelson: 95% CSAT

Australian eyewear retailer Bailey Nelson reports 95% CSAT running Intercom with automation and Fin, alongside a 96% cut in first-response time.

At scale, 95% is exceptional territory.

3Commas: 93% CSAT

3Commas reports 93% CSAT, with 95% of customers resolving their query through the help centre and automation before a human ever gets involved.

The pattern across all three: for a mature Intercom operation, 90%+ is achievable. Not typical. Achievable.

A practical Intercom CSAT benchmark

Here's how I'd read each range when reviewing a team's performance.

90–100%: Excellent

At this level, stop chasing the aggregate. Going from 92% to 93% changes nothing for the business.

Finding out that six of your ten negative ratings last month came from the same billing workflow? That changes something.

85–89%: Strong

A healthy place for most teams. There will still be bad conversations in the queue.

Your attention should go to recurring DSAT reasons, gaps between agents, gaps between conversation types, and how Fin performs against your humans on the same topics.

80–84%: Healthy, but investigate

Nothing is broken. But 15–20% of responding customers are telling you something, and it's probably clustered.

Pull the negative conversations and sort them: slow responses, incomplete resolutions, product limitations, billing, handoffs, policy decisions, wrong Fin answers. The cause matters far more than the number.

70–79%: Improvement needed

Negative experiences at this rate aren't isolated incidents. Read the bad conversations one by one and classify the root cause.

You may find the support team isn't the problem. Intercom makes this point directly: customers leave poor ratings over product limitations, company policy, and outcomes an agent can't control. Coaching your reps won't fix a broken refund policy.

Below 70%: Prioritise investigation

A sustained sub-70% score is a fire. But don't start by assuming every agent needs coaching.

Look for systemic causes first: unresolved conversations, a queue that's too deep, slow first responses, repeat contacts, broken escalation paths, product bugs, Fin marking things resolved that aren't.

The score tells you a problem exists. The individual conversations tell you what it actually is.

Human CSAT and AI CSAT shouldn't have the same benchmark

This matters more every quarter for teams running Fin.

Intercom reports teammate CSAT and Fin CSAT separately, and recommends comparing the two. Do that. But don't panic when Fin scores lower, because Intercom also notes customers rate AI and human support differently, which makes a head-to-head comparison misleading on its own.

The better benchmark for Fin is Fin last month. Track its trend, then read it alongside resolution rate, handoff rate, conversation type, and what human CSAT looks like after a handoff.

An AI agent at 82% CSAT that resolves thousands of routine questions on its own is doing enormous work. An AI agent at 91% that only touches the easy 5% of the queue isn't outperforming it.

Your CSAT response rate matters too

A 95% score looks great. Always ask: 95% of whom?

Surveyed CSAT only counts customers who chose to respond, and that can be a startlingly small slice of your actual queue - your CSAT score is missing most of your customers.

This is why Intercom's Surveyed CSAT report includes a request rate and a response rate, and why Intercom recommends watching the response rate before trusting the score to make decisions.

It's also why quality measurement is moving past surveys. Intercom's CX Score uses AI to evaluate experience across conversations rather than waiting for voluntary responses, and its AI Insights add a predicted CSAT on top. Supportman does the same with AI evaluation: every closed conversation gets an internal quality score and a predicted CSAT, including the ones where the customer never tapped a rating - see what 100% CSAT coverage actually looks like.

Benchmark yourself against yourself

Industry averages are calibration. Your own history is the real benchmark.

83% → 85% → 88% → 91% is evidence something you're doing works.

94% → 91% → 87% → 83% is a problem, even though that final 83% still clears the generic bar. A single average will hide exactly this kind of slide - why your average CSAT score can hide the problem.

A CSAT dashboard worth reviewing weekly shows: current CSAT, previous period, rating volume, positive vs. negative counts, team-level, agent-level, human vs. Fin, first-response time, and resolution time.

Then go read the actual conversations responsible for the movement.

Don't just measure DSAT. Route it.

Here's where teams lose most of CSAT's value.

A customer leaves a bad rating at 4:45pm Friday. It lands in Intercom's reporting. Someone opens the report Monday, maybe Tuesday. By then the customer has already disputed the charge with their bank.

The most valuable moment to see a bad rating is the moment it's submitted. Supportman sends every Intercom rating into Slack as it arrives, so you can run a dedicated DSAT channel that only carries negative ratings, filter for ratings with remarks, and jump straight back into the original conversation.

The difference in practice:

• Bad rating → buried in reporting → reviewed Friday
• Bad rating → Slack notification → someone acts while the customer is still online

Supportman also sends weekly team and agent reports with rating summaries, first-response times, and resolution times.

The goal isn't another dashboard. It's putting the feedback where your team already works.

What should your CSAT target be in 2026?

If you need one number: aim for 85%+ surveyed CSAT. Consistently above 90% puts you at the level the published Intercom customer stories operate at.

But don't let 90% become a vanity metric. A better goal set:

• Hold CSAT above 90%
• Review 100% of negative ratings
• Respond to recoverable DSAT while it's still recoverable
• Name the top three recurring DSAT causes each month
• Benchmark Fin and humans separately
• Watch the response rate, not just the score
• Measure quality on the conversations customers don't rate

CSAT isn't valuable because it produces a number. It's valuable because it shows you where customers are having a bad experience while you can still fix it.

This is exactly what we built Supportman for: your Intercom ratings land in Slack in real time, with weekly reports and AI quality scoring on the conversations nobody rates. If your team finds out about bad ratings on Monday morning, give it a look - setup takes about two minutes.

Intercom CSAT benchmark FAQ