The State of Customer Support Operations 2026
Satisfaction has not improved since 2013, complaints are at record levels, and the median support team resolves fewer than one in five tickets on first contact. In 2026 the gap between teams is operational, not technological. These are the customer support benchmarks for 2026, every figure traceable to a public source.
Key findings
The satisfaction gap
Is the bar actually rising, or does it just feel that way?
Every support leader has been told that customer expectations are climbing. It is the opening line of most vendor decks and most board presentations. The evidence says something more uncomfortable, and more useful: expectations are not the thing that changed. Delivery has been flat for thirteen years, and the money spent trying to move it has not moved it.
The index that has not moved since 2013
The American Customer Satisfaction Index measures roughly 200,000 customers a year across more than 400 companies in over 40 industries. It was founded at the University of Michigan in 1994 and is recognized by the US federal government as a benchmark for customer satisfaction.
In the first quarter of 2026 it fell 0.3% to 76.7 on a 100-point scale. That is the same score it recorded in 2013.
Thirteen years, no movement. And it is not for want of spending. The Index’s founder, Claes Fornell of the University of Michigan, puts the investment at “well over $100 billion annually with no detectable returns,” adding that “if anything, the returns have been negative.”
Source: American Customer Satisfaction Index, National ACSI Q1 2026. The three figures use different units and are shown side by side rather than plotted on one scale.
Customers are less satisfied, complaining more, and staying. Fornell calls the result pent-up defection: “When customers come back for more, even if they are less than satisfied, pent-up defection, the inverse of pent-up demand, intensifies.” The bill arrives later, as lost future revenue plus the cost of replacing the customers who eventually go.
For a support leader the practical reading is this: retention is currently hiding the quality of your service, and it will not hide it indefinitely.
A second index, different method, same answer
Forrester’s Customer Experience Index analyzed more than 275,000 customers’ perceptions of 469 brands across 12 industries and 13 countries in 2025.
In the US it was worse, and for the second year running. In most US industries quality declined across all three dimensions Forrester measures: effectiveness, ease and emotion. Forrester attributes the multi-year decline to “weaker employee experience, waning customer obsession, disappointing tech implementations, and economic volatility,” and notes that the gap between the experience brands intend to deliver and the one customers actually get is widening.
Two indices, one academic and one commercial, built on different samples with different methods, describing the same stall.
What this looks like in your industry
ACSI publishes annual scores by industry. Across the sectors most relevant to mid-market support operations, the pattern is not decline. It is stasis.
Three of these did not move at all. The only meaningful decline is nonhospital care, down 5%. The only meaningful gain is hospitals, up 3%, driven by inpatient care rising from 77 to 81.
So what do customers actually expect?
Zendesk surveyed 6,182 consumers and 5,115 business respondents across 22 countries. The fieldwork was done in June 2025 and published as a 2026 report, which is worth knowing when reading any figure from it.
And from the other side of the desk: 85% of CX leaders say customers abandon brands that cannot resolve an issue on first contact.
Hold that number. Chapter 3 shows what first contact resolution actually runs at, and the distance between the two is the single largest gap in this report.
The bar is not moving. Delivery is not moving. Over $100 billion a year is being spent on the gap without closing it, and rising retention is currently disguising the cost. The rest of this report measures the four things a support leader can actually change: where contacts enter, how fast they resolve, what the team costs, and whether AI has moved any of it.
Where contacts land
Where is the work coming from, and is the channel shift real?
Chapter 1 covered the conditions you do not control. This chapter starts on the ones you do, beginning with the front door.
The channel projection is real, but it describes entry, not resolution
Gartner surveyed 265 customer service and support executives and found that self-service and live chat are expected to overtake phone and email as the most valuable customer service technologies by 2027. Nearly three in four customer service organizations planned to implement agent assist solutions by the end of 2025.
That projection is widely quoted and it is accurate. It is also routinely misread. It describes where contacts arrive. It says nothing about where they get solved.
Three in four start in self-service. One in seven finishes there.
Neither of those is a technology problem. Both are content and taxonomy problems, and both sit inside the support team’s control.
When the work actually arrives
Observational data from more than 50,000 help desk tickets across 30+ companies, collected over 14 months to February 2026:
Source: Fixify, 2026 IT Help Desk Benchmark Report. 50,000+ tickets across 30+ companies over 14 months to 28 February 2026.
The shape matters for staffing. A queue that is 82% concentrated in business hours and peaks hard on Tuesday mornings is a scheduling problem before it is a headcount problem.
53% of practitioners say their single biggest challenge is managing growing ticket volumes without significantly increasing headcount. That is the constraint every other chapter in this report runs into.
Self-service is where almost everyone starts and almost nobody finishes: 73% in, 14% out, and nine in ten of those journeys needing another channel anyway. The 2027 channel projection is about the front door. Before rebuilding the front door, fix the two things customers actually complain about, content that does not match their issue and a system that does not appear to understand it.
Customer support benchmarks: speed, resolution and quality
What is normal, what is good, and where do you sit?
This is the chapter to bookmark. Two independent benchmark sets are used, they are built very differently, and where they disagree this report explains why rather than picking the friendlier number.
Sources: Freshworks, US companies, ticketing; Zendesk CX Trends 2026. That gap is the one this report exists to name.
| US companies, ticketing | Trendsetter | Performer (median) | Aspirant |
|---|---|---|---|
| First response time | 3m 47s | 1h 12m | 8h 10m |
| Resolution time | 43m 23s | 7h 58m | 41h 3m |
| Average response time | 7m 1s | 1h 36m | 8h 53m |
| Response SLA compliance | 39% | 15% | 4% |
| Resolution SLA compliance | 98% | 88% | 65% |
| First contact resolution | 32% | 15% | 4% |
The time figures are kept as a table because they are a three-tier matrix rather than a single series. Source: Freshworks.
Your chat number is flattering you
Run the same class of company on conversational support and the picture transforms.
| Growing companies, 250 to 5,000 employees | Trendsetter | Performer | Aspirant |
|---|---|---|---|
| CONVERSATIONS | |||
| First response time | 10s | 36s | 2m 52s |
| Resolution time | 2m 32s | 9m 9s | 29m 34s |
| CSAT | 98.75% | 95.15% | 85.79% |
| First contact resolution | 97.74% | 81.88% | 62.16% |
| TICKETING | |||
| First response time | 2m 31s | 53m 16s | 6h 27m |
| Resolution time | 25m 5s | 5h 1m | 31h 50m |
| Resolution SLA compliance | 97% | 91% | 65% |
| CSAT | 97.50% | 88.00% | 61.10% |
| First contact resolution | 35% | 15% | 6% |
Source: Freshworks.
Same companies. Same period. First contact resolution of 81.88% on chat and 15% on tickets.
SQM puts the contact center industry average CSAT at 78% top box. Freshworks shows 95.15% at the conversational median. Both are measured honestly. They disagree because they are not measuring the same thing.
Channel is doing most of the work. Live chat and messaging handle high-volume, low-complexity contacts, resolve them in minutes and get scored well. Email and case work is where the genuinely hard problems go, and it scores far worse: the same class of company runs 88.00% CSAT on ticketing and 61.10% in the bottom tier.
Measurement method does the rest. SQM surveys customers after a contact, by phone, with a minimum of 400 surveys per center, so it reaches people who would never click a widget. In-product CSAT widgets are answered disproportionately by customers whose issue was resolved.
The practical warning. A team reporting the conversational median of 95.15% while its ticketing queue sits at the 61.10% bottom tier is not measuring its support operation. It is measuring its easiest channel. Report both, or report the harder one.
The SQM industry standards
For teams running voice, or wanting a post-contact benchmark rather than an in-product one:
| Metric | Industry average | Good | World class |
|---|---|---|---|
| First call resolution | 70% | 70 to 79% | 80%+, reached by 5% |
| CSAT, top box | 78% | 75 to 84% | 85%+, reached by 5% |
| Customer service QA | 85% | 90 to 99% | 100% |
| Average handle time | 7 minutes | varies by contact type | varies |
| Abandon rate | 6% | under 5% | 3% or less |
| Average speed of answer | 34.4 seconds | no CSAT penalty within 120s |
A four-column standards grid is a genuine lookup table and is kept as one. Source: SQM Group, 500+ North American contact centers, minimum 400 surveys each. Metric names are reproduced as the source publishes them.
Two further findings worth acting on. Agent-assisted calls show 16% lower first call resolution and 7% lower call resolution than calls an agent handles alone, which is an argument for better first-line enablement rather than faster escalation. And SQM’s own ratio guidance is 15 support employees to 1 customer service agent.
First contact resolution by contact type
Aggregate first call resolution across SQM’s benchmark base was 69%, ranging from 43% to 88% depending on the center.
Technical support resolves at 60% and complaints at 48%. If your mix is weighted toward either, a blended target borrowed from a general benchmark will be wrong in a way that punishes your team for the work they were given.
The IT and technical support picture
Source: Fixify, 2026 IT Help Desk Benchmark Report.
That last set is the most encouraging finding in this report. Most tickets do not arrive angry, and most of the ones that do end better than they started. Sentiment is recoverable, and speed is what recovers it.
By industry
| Industry, conversations | First response | Resolution | CSAT | FCR |
|---|---|---|---|---|
| Software and Internet | 32s | 11m 8s | 96.98% | 79.10% |
| Retail and eCommerce | 1m 5s | 12m 7s | 95.92% | 82.43% |
| Business Services | 41s | 9m 20s | 96.47% | 85.51% |
Ticketing medians for the same industries: first response 46m 51s, 1h 29m and 1h 8m; resolution 7h 14m, 8h and 7h 10m. Retail and eCommerce ticketing CSAT runs 94.1% at the top tier, 82.6% at the median and 52.4% at the bottom, with reopen rates of 3%, 8% and 16%. Source: Freshworks.
A note on company size
Freshworks bands companies at 0 to 250, 250 to 5,000 and 5,000+ employees. A 100 to 500 person company straddles the first boundary, so both relevant bands are shown here rather than blended.
Smaller teams are not worse. On conversational support the small-company median beats the growing-company median on CSAT and first contact resolution, and loses only on raw speed. Scale buys response time. It does not buy quality.
A competitive 2026 operation answers tickets inside an hour, resolves inside eight, and resolves more than 15% of them on first contact. It reports the ticketing number alongside the chat number, because the chat number alone describes the easiest work the team does.
The staffing arithmetic
What does a support seat cost in 2026, and why do the people keep leaving?
Chapter 3 ended on a constraint: 53% of practitioners say their biggest challenge is absorbing more volume without more headcount. This chapter is about the headcount.
There is no credible public source for contact center attrition rate or cost per ticket. The widely circulated figures, “30 to 45% annual attrition” and “$6 to $40 per ticket”, trace back to vendor articles citing one another with no survey, no sample and no method behind them. None of them appear here. What follows is US government data, which is less convenient and considerably harder to argue with.
The job, in federal statistics
The US Bureau of Labor Statistics tracks customer service representatives as occupation 43-4051. As of May 2025:
The pay spread is twice as wide at the top as at the bottom. If you are paying below $15.27, you are in the bottom decile of a national market for a role with no formal entry barrier, competing against every other employer who can train someone on the job.
Source: US Bureau of Labor Statistics. BLS attributes such openings to workers leaving the occupation or exiting the labor force entirely. A shrinking occupation generating roughly 2.9 million hires over ten years.
That is the attrition picture, and it comes from the federal government rather than from a vendor article. You do not need a survey to tell you that support roles turn over constantly. The occupation is contracting and still needs to refill itself roughly ten times over in a decade.
For a support leader the operational reading is blunt: budget for continuous hiring and continuous onboarding as a permanent cost line, not as a bad year. The job market has priced this in even where individual employers have not.
Staffing ratios, from real data
The IT ratio comes from observed staffing rather than a survey, and the six-fold spread between the bottom and the top of the range is the interesting part. There is no single correct ratio, which means a ratio alone is a weak defence of a headcount request. Pair it with the queue shape from chapter 2 and the resolution times from chapter 3.
What the outsourcing data says, carefully
Deloitte’s 2025 Global Business Services Survey drew responses from leaders in more than 30 countries, fielded in the second half of 2024. The sample size is not published.
Deloitte’s own framing is worth repeating because it cuts against the usual pitch: it reports that cost is “increasingly becoming a deteriorating value proposition,” with leaders shifting the rationale toward experience and transformation.
The occupation is projected to shrink by 141,800 jobs while generating 289,500 openings a year. Treat hiring and onboarding as a standing cost, pay attention to where you sit in a pay range that doubles from bottom decile to top, and do not defend a headcount number with a staffing ratio that varies six-fold across real companies.
AI: wide adoption, narrow results
Where has AI actually moved a number, and where has it not?
Every chapter so far has run into the same constraint: more volume, flat headcount, flat satisfaction. AI is where the entire sector is looking for the way out. The adoption data and the outcome data tell noticeably different stories.
The adoption ladder
Each rung comes from a different named source. The shape is the finding.
Nine in ten under pressure. One in ten mature. Agentic AI use has risen fast, from 39% to 66% in a year. Maturity has not kept pace with it.
The gap between leaders and everyone else is widening
McKinsey’s State of Customer Care survey segmented organizations by maturity for the first time. The separation is stark:
42% of leaders have reversed rising inbound volumes through smarter self-service and digital deflection. McKinsey’s diagnosis is not technological: “Agentic AI stalls less because models underperform and more because organizations lack the orchestration, governance, and operating discipline required to scale.” For wider context, only 39% of organizations across all functions report EBIT impact at enterprise level from AI.
What AI has measurably moved
The honest answer is speed, and the strongest evidence for it is observational rather than self-reported.
Deflection is not resolution, and the difference is where most claims fall apart
Freshworks reports AI agent deflection rates of 48% at companies under 250 employees, 55% at 251 to 5,000 and 56% above that, with the Americas at 47%. Again, its own product. Gartner reports that 14% of customer service issues are fully resolved in self-service.
These are not contradictory, because they do not measure the same event. Deflection counts a contact that did not reach a human agent. Resolution counts a customer whose problem was actually solved. A customer who gives up, opens a second ticket, or calls instead has been deflected and not resolved.
When a vendor quotes a deflection rate, the question to ask is what happened to the deflected contact. Chapter 2 already answered it in the general case: nearly 9 in 10 journeys that start in self-service need another channel anyway.
What customers actually do, which is not what anyone assumes
Gartner surveyed 3,566 B2B and B2C customers in February and March 2026.
Three findings here, and the one everybody quotes is the least interesting.
1. The 49% versus 7% gap is an availability problem, not a trust problem. Half of customers were willing. One in fourteen encountered one they used. That is the cheapest gap in this entire report to close.
2. One bad experience is close to terminal. Only 27% will try again, no matter how much the bot improves afterwards. This is an argument for launching narrow and working, rather than broad and approximate.
3. Customers are roughly three times more likely to ask a third-party AI than to use your chatbot. The answer engine is already inside your support journey. It is answering questions about your product from whatever it can find, and most support organizations have no view of it at all.
Source: J.D. Power, 2026 U.S. Digital Banking and Credit Card Mobile App Satisfaction Studies, 18,028 customers fielded January to March 2026. In their words: “limited escalation options often trap customers in frustrating self-service loops that erode trust instead of connecting them to the help they need.”
J.D. Power surveyed 18,028 retail bank and credit card customers using a different method in a different sector, and arrived at the same shape.
Gartner’s 87% human-access requirement and J.D. Power’s collapse on fraud and disputes are the same finding from two entirely independent datasets. AI handles the routine well and the exceptional badly, and the exceptional is where trust is made or lost.
The perception gap inside support teams
Support teams believe customers trust AI considerably more than customers say they do. And on transparency: 95% of consumers expect an explanation for AI-made decisions, while only 37% of companies offer any reasoning at all. 79% say plain-language reasoning is important.
What is holding deployments back
That last pair deserves reading together. More than nine in ten expect returns inside a year, and about half say their recent purchases delivered.
The workforce effects are already arriving: 97% of service leaders report AI affects workforce planning, 40% of teams report agents spending more time training and optimizing AI systems, and 58% of leaders intend to upskill agents into knowledge management specialists.
That last one connects directly back to chapter 2. The most common complaints about self-service were content that did not match the issue and a system that did not appear to understand it. Both are knowledge management problems, and knowledge management is what determines whether an AI deployment works.
Nine in ten support leaders are under pressure to deploy AI and one in ten has a mature deployment. AI has measurably moved speed and has not yet moved satisfaction. The cheapest wins available are an escalation path customers can find, because 87% require one, and a chatbot that is actually offered, because half would use one and one in fourteen met one.
Benchmark yourself
Nine dimensions. Every benchmark comes from a chapter above. Score one point per row where you match or beat it.
| Dimension | The 2026 benchmark | You |
|---|---|---|
| First response, ticketing | Median US company 1h 12m. Top tier 3m 47s | |
| Resolution time, ticketing | Median US company 7h 58m. Top tier 43m | |
| First contact resolution, ticketing | Median US company 15%. Top tier 32% | |
| CSAT, and which channel you quote | Post-contact average 78%. Ticketing median 88.00%, bottom tier 61.10% | |
| Self-service resolution | Only 14% of issues fully resolve there | |
| Abandon rate | Average 6%, good under 5%, world class 3% or less | |
| QA score | Industry average 85%, good is 90 to 99% | |
| Pay position | Median $21.53/hr. Below $15.27 is the national bottom decile | |
| AI maturity and escalation | 10% report mature deployment. 87% of customers require a reachable human |
A scorecard is a genuine lookup grid, so it stays a table. Benchmarks drawn from sources 13, 16, 11, 17, 20 and 24, as cited in the chapters above.
How to read your score
Three things worth doing whatever you scored
1. Report your ticketing numbers next to your chat numbers. If leadership only ever sees the chat figure, they are seeing the easiest work the team does.
2. Check what a third-party AI says about your product. Customers are about three times more likely to ask one than to use your chatbot.
3. Make the human escalation path findable. 87% of customers say it is essential, and the industry with the best data shows satisfaction collapsing precisely where escalation is missing.
All six chapters, every benchmark table, the nine row scorecard you can run against your own team, the methodology note and all 28 sources. One field, and no follow up you did not ask for.
Frequently asked questions
What is a good first contact resolution rate in 2026?+
On ticketing, the median US company resolves 15% of tickets on first contact and the top 20th percentile reaches 32%. On voice, SQM puts the industry average first call resolution at 70%, with 70 to 79% considered good and 80% or above world class, achieved by only about 5% of contact centers. The two differ because they measure different channels and use different survey methods, so compare against the one that matches how you measure.
What is a good average CSAT score for a support team?+
SQM’s post-contact industry average is 78% top box, with 75 to 84% good and 85% or higher world class. In-product widget CSAT runs much higher: the ticketing median is 88.00% and the conversational median 95.15%. A chat-widget CSAT is not comparable to a post-contact survey score.
How long should customers wait for a first response?+
The median US company responds to a ticket in 1 hour 12 minutes and the top tier in 3 minutes 47 seconds. On conversational channels the median is 36 seconds. In IT help desks the median first response is 5 minutes.
How much of customer service can self-service realistically handle?+
73% of customers use self-service at some point but only 14% of issues fully resolve there, and nearly 9 in 10 journeys that start in self-service need another channel. Vendor deflection rates of 47% to 56% measure something different: a contact that did not reach an agent, not a problem that was solved.
Is customer satisfaction actually getting worse?+
It is not falling so much as stuck. The national index is 76.7, the same as 2013, while complaints rose 16% in one quarter. Forrester found 25% of US brands declined in 2025 against 7% that improved.
How many support teams have actually succeeded with AI?+
91% report executive pressure to implement AI and 82% invested in the last year, but only 10% describe their deployment as mature and 22% have a fully integrated AI framework. 62% say their metrics improved.
Do customers want AI handling their support?+
50% say interactions are easier when companies use GenAI, but 87% say it is essential to be able to reach a human, and only 27% would try a chatbot again after one bad experience. 49% would have used a chatbot if offered, while only 7% encountered one.
What does a customer service representative cost in the US?+
Median pay is $44,770 a year, $21.53 an hour as of May 2025, with the bottom 10% under $15.27 and the top 10% over $30.57.
Why is support turnover so high?+
The occupation is projected to shrink 5%, or 141,800 jobs, between 2025 and 2035, while still generating about 289,500 openings a year. A contracting occupation refilling itself roughly ten times over in a decade is the clearest public evidence of how much churn is structural.
Methodology and sources
Every statistic in this report was collected from the source named below and verified before publication. Data years are stated where they differ from publication years. Figures are reported as published. No estimates or interpolations were made. Where a survey covers one geography or one channel, the text says so. Charts are drawn from the cited figures only.
Where sources disagree, both figures are published and the reason for the difference is explained rather than resolved by preference. The clearest case is customer satisfaction scoring, where post-contact survey methods and in-product widgets produce systematically different numbers, discussed in chapter 3.
First, several sources are vendor-published product data. Where that is the case it is labeled in the text, and vendor measurements of a vendor’s own AI product are never presented as industry rates.
Second, no credible public source exists for contact center attrition rate or cost per ticket. Widely circulated figures for both trace to vendor articles citing one another without surveys or samples. They are excluded, and US Bureau of Labor Statistics employment and wage data is used instead.
A note on terminology. This report uses “contact center” throughout. Where a source’s own title or quoted text uses “call center”, it is reproduced exactly as published.
The sources
Gartner restricts automated access to its newsroom. Every Gartner figure in this report was verified against independent trade coverage that names the survey and its sample size, cited above at sources 10 and 12 and, for the remainder, in reporting by CX Dive, CX Today, Communications Today, IT-Online and insurance-canada.ca. Gartner is the source and is credited as such. The trade coverage is how each figure was checked.
About Touch Support
Touch Support has built embedded support teams since 2003. Founded in Chicago, it provides dedicated CX, IT, Finance and Accounting, and Software Development teams that work inside a client’s own systems, tools and workflows rather than from a separate queue. Teams go live in 7 days and run 24/7/365, with an average CSAT of 92.7% across supported accounts.
This report was compiled because the benchmark data support leaders need is scattered across nearly thirty sources, most of it published once a year and rarely in one place. It is free to read and free to quote.
Citation. You may quote this report with attribution (“Touch Support, The State of Customer Support Operations 2026”) and a link to this page.
