The State of Customer Satisfaction (CSAT) 2026

At a glance

Executive summary

A customer satisfaction survey, or CSAT survey, is a short questionnaire used to measure how satisfied a customer felt with a specific interaction, product, or journey step. The most common customer satisfaction score uses the top-two-box CSAT formula: (Satisfied + Very Satisfied) / Total × 100. In practice, that usually means counting 4s and 5s on a five-point scale and converting that share into a percentage. CSAT is strongest when sent close to the moment being evaluated. [Gainsight; SurveyMonkey]

For operators publishing or benchmarking a CSAT survey, the hard truth in 2026 is that there is no universal "good CSAT score". Scores vary sharply by industry, touchpoint, timing, wording, scale design, and channel. The most defensible benchmark remains "your own score over time, compared with your industry's best available benchmark." Still, the public evidence is strong enough to answer the questions buyers search most often: what is CSAT, how is CSAT calculated, what is a good CSAT score, CSAT vs NPS, and how to improve CSAT.

On the ACSI's 100-point national index, U.S. overall satisfaction fell to 76.7 in Q1 2026, down from 77.0 in 2025 and back to the same level seen in 2013. The strongest 2026 scores were regional and community banks (83), financial advisors (82), and specialty retailers (80). The weakest were AI platforms (73), gas stations (75), and several travel categories clustered at 76. [ACSI 2026]

What a customer satisfaction survey measures

Early customer satisfaction research did not begin with today's SaaS dashboards. In Robert A. Westbrook's 1980 work, satisfaction was operationalised as a product/service evaluation using a "Delighted–Terrible" scale. Richard Oliver's 1981 framework treated satisfaction as the psychological state that results from comparing expectations with experienced performance — the classic expectancy-disconfirmation view that still underpins modern CSAT thinking. [Westbrook 1980; Oliver 1981]

That history matters because it explains why "what is CSAT" is not just a software-glossary question. CSAT is not a single official scale. It is an umbrella term for transactional satisfaction measurement: a customer gives an evaluative judgment after a defined experience. Fornell's national barometer work in Sweden and then the ACSI in the United States moved the field from ad hoc single items toward structured, multi-item modelling systems designed to connect satisfaction with loyalty, retention, and financial performance. [Fornell 1992; Fornell et al. 1996]

This is also why question phrasing changes scores. Changing the object of evaluation — "today's checkout", "your recent support interaction", "our company overall", "value for money", "how likely are you to recommend us?" — changes the underlying construct being measured. ACSI itself explicitly warns that neither oversized questionnaires nor a single question for everything is appropriate, because customer data are noisy and measurement must be calibrated to the business objective. [ACSI methodology]

How CSAT is calculated

The standard modern CSAT formula on a five-point scale is:

CSAT = (Number of satisfied responses / Total responses) × 100

In most systems, "satisfied responses" means ratings of 4 or 5. Gainsight states the formula as (# of satisfied customer responses / total # of customer responses) × 100, and Intercom describes the same top-two-box method for SaaS customer success teams. [Gainsight; Intercom]

Three common variants

For publication-quality benchmarking, the safest editorial rule is simple: never compare a top-two-box CSAT percentage to a mean-score benchmark without stating the methodological difference. Much of the confusion in "average CSAT score" articles comes from mixing these methods.

When each variant makes sense

Top-two-box works best when the business question is binary in spirit: "Did we satisfy the customer or not?" It is especially useful for support, transactions, deliveries, and post-resolution measurement. Mean score is more useful when you care about gradation within the positive band — for example, whether customers are merely satisfied or strongly satisfied — or when modelling the relative effects of multiple satisfaction dimensions. The ACSI's own methodology is closer to the latter tradition than to a pure one-question top-two-box design.

ACSI national satisfaction trend, 2010–2026

The American Customer Satisfaction Index remains the most coherent public benchmark set for U.S. customer satisfaction because it uses a common methodology across industries and publishes both national and industry-level results. In Q1 2026, the national ACSI fell 0.3% to 76.7 after a flat 2025, and ACSI notes that the current national score is effectively back where it was in 2013. [ACSI Q1 2026]

The main editorial takeaway is straightforward: the national satisfaction environment is not collapsing, but it is not meaningfully improving either. For any CSAT programme, "holding flat" in a stagnant benchmark environment is not the same as leading your category.

Top ACSI industries in 2026

ACSI does not release every sector on the same day. As of mid-May 2026, some industries already had 2026 results while others still showed 2025 as their most recent update. The chart below covers the highest-scoring industries and segments whose 2026 results were publicly available.

This makes one thing obvious: high-70s to low-80s is the real leadership band in current U.S. satisfaction benchmarks, not mid-60s optimism masquerading as success.

Weakest ACSI industries in the 2026 set

Among the 2026 datasets already public, the weakest scores were concentrated in newer AI services and lower-scoring travel and utilitarian categories. This is the weakest band of sectors with 2026 scores available, not necessarily the worst ACSI industries overall — categories like social media, wireless, ISPs, and video streaming still showed 2025 as their latest update at publication.

Company leaders by category

The strongest named companies visible across the public benchmark set as of publication:

What is a good CSAT score?

The question people actually ask is not "what is the formula?" — it is "what is a good CSAT score?" The most defensible answer is: it depends on your scale and your industry, but on a 0–100 style CSAT percentage, 80+ is excellent, 75–79 is good, 70–74 is fair, and below 70 is weak. That cut-off is an informed synthesis from current ACSI distributions, not a single official law of CX physics. It is grounded in the fact that the U.S. national ACSI sits at 76.7, current 2026 leaders cluster at 80–83, and obvious laggards often sit in the low 70s. Gainsight notes that ACSI itself treats 75 as the cut-off between good and bad in airline satisfaction.

Practical cutoffs by industry

For software and SaaS, the public benchmark picture is weaker because there is no ACSI-equivalent, open, cross-vendor SaaS CSAT index. Retently's guidance suggests anything over 50% can be positive, ideally 60% or 70%+, while case-study evidence from SaaS leaders routinely lands in the mid-80s to low-90s. That makes "70 is fine" a risky target if you sell a sticky subscription product.

What is the average CSAT score?

If you mean the broad U.S. benchmark environment, the closest public answer is the national ACSI level — 76.7 in Q1 2026. If you mean percentage-style vendor CSAT across all business types, there is no single universally accepted public average because methodologies differ. That is why many vendor articles disagree. The closest reliable public average for the U.S. economy is the ACSI national level, but your specific CSAT survey should be benchmarked against a matching peer set, not a random internet average.

CSAT vs NPS vs CES

The most important answer to "CSAT vs NPS which is better?" is that neither is universally better. They answer different questions.

What the literature says

Frederick Reichheld's original Net Promoter argument was bold: he claimed recommendation intent was the "one number" most predictive of growth. That idea became enormously influential because it made CX measurement simple and executive-friendly. [Reichheld 2003]

But the academic response was more sceptical. Keiningham and colleagues, in their longitudinal work, found that NPS does not consistently outperform other attitudinal or satisfaction measures in predicting growth. In other words, NPS is useful, but the evidence does not justify treating it as a magical universal replacement for CSAT or other measures. [Keiningham et al. 2007]

For service contexts, Dixon, Freeman, and Toman's Stop Trying to Delight Your Customers made the counterargument that reducing customer effort matters more for loyalty than attempting "wow" moments. That insight is why CES is especially powerful in support, onboarding, claims, billing, and B2B service operations. [Dixon, Freeman & Toman 2010; Gartner CES guidance]

Which metric predicts revenue or retention best?

The practical rule: use CSAT for moments, NPS for relationships, and CES for friction. In mature VoC programmes, the right answer is not choosing one metric forever — it is choosing the right metric for the right decision.

15 CSAT survey questions by use case

A CSAT survey should be short, specific, and tied to a clearly identifiable moment. SurveyMonkey recommends two to three questions at most: the rating item plus one or two contextual follow-ups.

Source design logic: SurveyMonkey, Qualtrics, Gainsight, and Pendo — ask about a specific experience, avoid abstraction, and keep the scope narrow enough that the respondent can answer from memory rather than general sentiment.

Single-question vs multi-question CSAT

Most operational teams default to a single-question CSAT because speed matters. That is reasonable. But ACSI explicitly argues that both extremes — one question for everything or 200-question surveys — are methodologically flawed. The ideal design depends on use case: one rating item for real-time operational monitoring, then a tightly designed diagnostic layer when you need to understand drivers. Mittal and Kamakura's work strengthened the case for taking satisfaction seriously because it linked satisfaction to repurchase behaviour.

Timing, channel, length and framing

The best public guidance on timing is directionally consistent. SurveyMonkey recommends sending transactional surveys as close as possible to the interaction, including immediately after support chats or ticket closure, while post-event surveys should go within 24–48 hours. Delighted makes the same point: a post-support survey sent 24 hours after an interaction should outperform one sent 7 days later because recognition and recall remain fresh. [SurveyMonkey; Delighted]

Survey length matters enough to change participation materially. SurveyMonkey reports that surveys beginning with a simple multiple-choice question had an average 89% completion rate, versus 83% when the first question was open-ended. Qualtrics similarly warns that once a survey contains more than three open-text boxes, completion rates start to decline and respondents write less. [SurveyMonkey; Qualtrics]

That is one of the clearest evidence-based design rules in all CSAT practice: lead with the rating question, place the comment box after it, and keep open text sparse.

Response rates by channel

The strongest current public channel data comes from Refiner's 2025 in-app survey benchmark and Delighted's customer-survey response benchmarks. Refiner analysed 1,382 in-app surveys with more than 50 million views and found an average 27.52% response rate and 24.84% completion rate. Within that dataset, in-app CSAT specifically averaged 26.29% response and 22.73% completion. Mobile app in-app surveys averaged 36.14% response, versus 26.48% in web apps. [Refiner 2025; Delighted; Pendo]

Pendo's long-standing in-app survey guidance says organisations often see 2× to 10× higher response rates when they move surveys from email into the product itself, because customers are asked while already inside the experience rather than having to act on an email later.

The operational lesson is clear: if you can responsibly ask in context, in-app beats email. For broader relationship measurement, email still matters — especially for users who did not encounter the in-app touchpoint — which is why Pendo supports backup email sends.

Survey fatigue is real

Public, apples-to-apples time-series data for CSAT response rates from 2015 to 2026 is surprisingly thin. Many articles claim long-run declines, but few publish consistent primary-source series by channel and survey type. What is public supports a narrower conclusion: digital survey fatigue is real, email participation is structurally harder than in-app, and mobile optimisation now matters more than ever. [Qualtrics; SurveyMonkey; UK Office for Statistics Regulation]

AI sentiment, chatbot CSAT, and voice surveys

In 2026, the important shift is not "AI replaces CSAT." It is AI augments CSAT. Forrester's State of AI 2025 says that more than 70% of surveyed firms have generative or predictive AI in production, while Gartner reported that 85% of customer-service leaders would explore or pilot customer-facing conversational GenAI in 2025. [Forrester 2025; Gartner 2024]

But customer-facing AI still carries a trust deficit. ACSI's 2026 AI Platforms study gave the category an overall score of 73, putting it below airlines, social media, and mortgage lenders in satisfaction terms. The biggest concerns were reduced human interaction, data security, and trust — exactly the issues that pure sentiment automation can miss if you stop asking direct satisfaction questions. [ACSI AI Platforms 2026]

In other words: AI sentiment analysis should not replace the CSAT survey — it should prioritise, classify, and enrich it. Use AI to summarise open text, identify themes, and trigger alerts. Use CSAT to keep the measurement anchored to an explicit customer judgment.

For voice-based CSAT, Google Cloud's contact-centre guidance recommends announcing the post-call survey before the interaction ends because callers may otherwise hang up before the survey starts, reducing response rates. Voice CSAT can work well, but only if the handoff is designed carefully.

How to improve CSAT

Improving CSAT is rarely about writing a cleverer survey. It is almost always about fixing the experience that the survey reveals. The best public evidence supports a simple playbook:

1. Put the survey at the truth moment

Send the survey when the experience is still cognitively available. For support, that often means immediately after chat ends or ticket closure. For events, SurveyMonkey recommends 24–48 hours. Delighted warns that a post-support survey sent after 24 hours is more likely to perform well than one sent after 7 days.

2. Prefer in-app over email when context allows

If the interaction happened in your product, measure it in your product. Refiner's 2025 benchmark put average in-app response at 27.52%, with in-app CSAT at 26.29%, while Delighted's email benchmarks sat in the 6–16% range. Pendo's own guidance says in-app delivery can drive 2× to 10× the response rate of email. Build your free CSAT survey on SpaceForms if you want to test this in real time with voice, chat, and in-app-style flows rather than relying on email alone.

3. Keep it short and close-ended first

If you want diagnostic colour, ask for it after the rating question. SurveyMonkey's data shows a 6-point completion penalty when a survey starts with open text instead of a multiple-choice item. Qualtrics warns that more than three open-text boxes drags completion further. A good default template:

4. Optimise for mobile

Qualtrics notes that surveys are more likely than ever to be taken on mobile and recommends minimising scrolling, limiting open text, and breaking up large questions. Refiner's benchmark adds the practical payoff: mobile app in-app surveys reached 36.14% response versus 26.48% in web apps — a 9.7pp gap.

5. Close the loop fast

Collecting feedback without follow-up is measurement theatre. A CCMA case study showed a reduction in negative-feedback recovery time from 7 days to 2 days after redesigning how cases were surfaced and managed. That case used NPS language, but the operational lesson applies equally to CSAT: fast follow-up turns passive listening into retention work.

6. Fix the operational driver, not just the score

The strongest ACSI category gains in 2026 came from concrete service improvements, not survey tweaks. Airlines improved mobile-app reliability, reservation ease, boarding, flight information, baggage handling, and seat-adjacent experience measures as the category rebounded to 76. Banks improved branch/ATM access and in-branch transaction speed while staying at 80 overall. Lodging rose to 77 as reservation ease, check-in / check-out, app quality, and room cleanliness improved.

Case studies with named companies

FAQs

Methodology and limitations

This report prioritised primary or primary-adjacent sources: ACSI industry pages and press releases, J.D. Power press releases, Press Ganey reports, AirHelp methodology, official documentation from SurveyMonkey, Qualtrics, Pendo, Gartner, Forrester, Bain, and public customer case studies. Where the public web did not provide a clean benchmark, the report says so instead of inventing one.

Principal sources used

Cite this report

@techreport{lundberg2026csat,
  title       = {The State of Customer Satisfaction (CSAT) 2026: Benchmarks, Trends & Best Practices},
  author      = {Lundberg, Eric},
  institution = {SpaceForms Research},
  year        = {2026},
  version     = {1.0},
  url         = {https://spaceforms.io/reports/state-of-csat-2026}
}

Related reports