Support Metrics: CSAT, FRT, Resolution Time, Deflection Rate

The Four Metrics At A Glance

• CSAT (Customer Satisfaction): lagging. Healthy band is 85–92% positive on a "satisfied/very satisfied" 5-point scale.
• First Response Time (FRT): leading. Chat: 1–5 minutes. Email priority: 1 hour. Email standard: 4 hours. Phone: answer before 30-second abandonment threshold.
• Resolution Time: leading-ish. P1: 4 hours. P2: 1 business day. P3: 3 business days. P4: 1 week.
• Deflection Rate: structural. Healthy KB deflection sits at 30–50%. Below 10% means the knowledge base is broken or invisible.
• Escalation Rate (bonus): 8–15% is normal for L1. Over 25% means triage or training is the problem, not the specialist.

It's Monday morning. You open Slack and your manager has left you a message from Friday afternoon: "Let's chat this week about your CSAT — needs improvement."

You stare at the dashboard. Your weekly score is 84%. Last week it was 91%. The week before, 88%. There's no comment, no tagged ticket, no theme. Just a number that dropped seven points and a one-line note that's now sitting in your head all weekend.

This is the problem with how most support specialists are managed against metrics. CSAT is treated like a thermometer that tells you how you're doing. It isn't. It's a thermometer that tells you how you did, on a small sample, after the customer already had the experience. By the time the score lands, the bad ticket is closed, the customer has moved on, and you have no way to walk it back.

The four metrics on your dashboard aren't equally useful. Some are leading: they tell you a ticket is going wrong while you can still fix it. Some are lagging, telling you a ticket already went wrong, last week, in aggregate. A specialist who watches the leading ones every morning will have a CSAT score that mostly takes care of itself.

Here's the working model.

Why Metric Type Matters More Than the Metric

Before we get into specific numbers, you need to internalize one distinction.

Lagging indicators measure outcomes after the fact. CSAT is the canonical example. A customer fills out a survey two hours after their ticket closed. The score lands in your dashboard the next day. By the time you see it, the experience is over. You can't influence the score, only the next ticket.

Leading indicators measure things you can still change. First response time is leading: you control whether the next ticket in your queue gets a reply in three minutes or thirty. Escalation rate is leading on a weekly horizon. If you spot it climbing on Tuesday, you can pull a teammate's tickets and review them by Thursday.

The mistake new specialists make is staring at the lagging number and trying to will it upward. You can't. The number is already cooked. The only real lever is to influence the leading metrics today so the lagging metric reads better on Friday.

This sounds like a small reframe. It isn't. It changes the shape of your week. Instead of dreading the CSAT review, you build a Monday-morning ritual around the leading numbers and let the lagging one report whatever it reports.

CSAT vs CES vs NPS: What Each Survey Actually Asks

Three customer survey types get blended together in conversation. They measure different things and shouldn't be averaged or compared.

CSAT (Customer Satisfaction Score) asks: "How satisfied were you with this interaction?" Usually a 1–5 scale or thumbs up/down. CSAT is transactional. It scores a single ticket. Use it to evaluate individual support interactions.

CES (Customer Effort Score) asks: "How much effort did you have to put in to get your issue resolved?" Usually a 1–7 scale where lower is better. CES is the most predictive of churn risk in support contexts. A customer who got a "yes" but had to fight for it is more likely to leave than one who got a "no" cleanly. If your tool only supports one survey, CES is often the better signal.

NPS (Net Promoter Score) asks: "How likely are you to recommend us to a colleague?" Scale 0–10. NPS measures relationship-level sentiment, not interaction quality. It's a CEO-level metric. A specialist shouldn't be judged on NPS, because it reflects the whole product, pricing, and company experience, not your last 50 tickets.

Why this matters: a 5-point CSAT scale is statistically noisy at the weekly level. If you handle 60 tickets a week and 12 customers respond, one angry one-star can swing your average by 0.4 points. That looks dramatic on a dashboard. It isn't a real signal.

The fix: read the comments, not the average. One-star and two-star ratings almost always include text. Five-star ratings rarely do. Your weekly review should start with reading every sub-3 comment from the last seven days. Three of them, four of them, whatever the count is. Read the words. Patterns will jump out faster than spreadsheets ever will.

Want a deeper read on how triage upstream of these surveys shapes the score downstream? See Ticket Triage: Prioritizing the Queue.

First Response Time: The Most Controllable Number on Your Dashboard

First response time is the gap between a ticket being created and the first human (not autoresponder) reply landing. It's the most directly controllable number on your dashboard and the leading indicator most correlated with CSAT in study after study.

Benchmarks by channel:

Track median, not average. One ticket that took 14 hours to reply (because it landed at 11 PM) will pull your average into the red even when 49 of 50 tickets were under 30 minutes. Median is honest. P90 (the 90th percentile, the slowest 10% of tickets) tells you about your tail. Your scorecard should have both.

The cheap FRT trick to avoid: sending a generic "thanks, we're looking into this" autoresponder doesn't count. Customers know. Tools that score it as "first response" are lying to you and to your manager. If you're going to acknowledge fast, acknowledge with at least one substantive sentence: what you saw in the ticket, what you'll check first. That's a real first response. The script library in Support Scripts That Actually Work covers acknowledgments that earn the FRT credit honestly.

Resolution Time and the Re-Open Trap

Resolution time is the gap between ticket creation and ticket close. Standard targets by priority:

• P1 (system down, blocked customer): 4 hours
• P2 (significant impairment, no workaround): 1 business day
• P3 (functional issue, workaround exists): 3 business days
• P4 (cosmetic, low impact, feature request): 1 week

These are starting points. Your team's SLAs may be tighter or looser. The numbers matter less than the question they force: do you actually triage tickets into priority buckets when you open them, or do you treat them all as one undifferentiated queue?

The re-open trap. A specialist with a great resolution time number can still be doing poor work if their tickets re-open. Closing fast only counts if the issue stayed closed. Track first-contact resolution rate (FCR) alongside resolution time:

FCR = (tickets resolved in single interaction) / (total tickets)

A healthy band is 70–80% for L1 support on a mature product. Below 60% means you're band-aiding, not fixing. Above 90% may mean you're cherry-picking easy tickets, or the queue has been over-triaged before it reaches you.

The other trap: tickets closed because the customer stopped replying. These look like resolutions in the dashboard. They aren't. If your tooling can flag "auto-closed after no response," exclude those from your FCR numerator. They're a separate category that deserves its own review, usually a sign that your follow-up cadence is off, not that the issue was solved.

Deflection Rate: The Metric Specialists Pretend Doesn't Apply to Them

Deflection rate measures the percentage of users who solve their own problem before opening a ticket. Most specialists ignore it because it doesn't show up in their personal queue. That's a mistake. Deflection is the most leveraged metric on the entire support team. Every point of deflection is a ticket you didn't have to handle, freeing time for the ones that actually need a human.

The formula:

KB deflection rate = (KB sessions − tickets opened from KB) / KB sessions

If 10,000 people read a help article last month and 1,500 of them ended up opening a ticket anyway, your deflection rate on that article is 85%. If the same 10,000 reads produced 6,000 tickets, your rate is 40% and the article is failing.

Healthy bands:

• KB deflection: 30–50% of total potential ticket volume
• AI/bot deflection: 15–35% (varies enormously by product complexity)
• Community/forum deflection: 5–15%

Below 10% means one of three things:

1. The KB articles exist but are unfindable (search is broken, or they're not linked from product UI)
2. The articles exist but are wrong, outdated, or written for the wrong audience
3. The product itself has trust-eroding bugs that no article will deflect, and customers want a human

Your role here isn't to write the entire knowledge base. Your role is to flag the gaps. The five tickets you closed today: were any of them solvable with a 200-word KB article that doesn't exist yet? Tag them. Send the list to the documentation owner on Friday. That's how deflection grows: 2–3 percentage points per quarter, ticket by ticket, gap by gap.

Escalation Rate: Quality Signal, Not Failure Signal

Escalation rate is the percentage of tickets you push to L2, an engineer, or a specialist queue.

Escalation rate = (tickets escalated) / (tickets received)

Healthy bands by tier:

• L1 generalist support: 8–15%
• L2 specialist support: 3–8%
• Anything over 25%: triage is broken upstream, training is incomplete, or product complexity has outpaced documentation

The mistake is treating escalation as a failure. It isn't. A specialist who escalates a billing edge case at the 20-minute mark is doing better work than one who flails on it for two hours and lands a 2-star CSAT. The right escalation, fast, is a quality move. See Support Escalation: When to Push It Up for the criteria that separate good escalations from bailouts.

What you should watch is the trend and the mix. If your escalation rate is climbing for three weeks straight, something changed. New product release, new ticket category, or your own confidence is wobbling on a topic. If 70% of your escalations are the same root cause, that's a training gap or a documentation gap, not a personal performance problem.

The Personal Metric Scorecard

Pull this every Friday afternoon. Fifteen minutes. Same time every week.

The point of the scorecard isn't to grade yourself. It's to spot drift. A 4-week rolling number that quietly trends down by 1.5 percentage points is invisible week to week and obvious in aggregate.

The 15-Minute Friday Self-Review

Block this on your calendar. Fridays, 4:00–4:15 PM. Recurring.

1. Pull your last 5 closed tickets. Open each one fully.
2. Score yourself on each: Was the response time honest? Was the resolution clean? Did you escalate at the right moment, too early, too late, not at all?
3. Read every sub-3 CSAT comment from the week. All of them, not the top one.
4. Identify one pattern. Just one. "I'm slow on the second response, not the first." "I keep over-explaining on billing tickets." "I'm escalating fraud questions I should be handling."
5. Write the pattern down. Bring it to your 1:1 next week.

Specialists who do this for 12 weeks straight see CSAT drift upward 4–8 percentage points without ever directly trying to "improve CSAT." That's the leading-vs-lagging point in action.

A Threshold Table for Managers

If you're a manager reading this, the numbers tell you what to do, not just what to feel.

Common Pitfalls (And How to Avoid Them)

Chasing the score, not the driver. "Improve your CSAT" is unactionable. The actionable version is "your second-response time on billing tickets is 90 minutes; bring it under 30 and watch what happens." Always push past the headline number to the driver underneath. The full inventory of these mistakes lives in Support Specialist Common Pitfalls.

Ignoring deflection because it's not in your queue. Every untagged KB-gap ticket is a future ticket you'll handle again. Tag them.

Having no leading metric. If the first time you find out your week went sideways is Friday afternoon when CSAT lands, you're flying blind. FRT median by Tuesday morning. Escalation count by Wednesday. CSAT comments daily. Those are your real dashboard.

Averaging surveys. Read individual sub-3 ratings. Don't compute the mean and call it a week.

Gaming the autoresponder. Tools that count "we received your ticket" as first response are tools that lie to you. Calibrate against substantive replies only.

What "Good" Looks Like in 90 Days

A specialist who works this system for one quarter typically lands here:

• FRT median holds inside SLA every week, p90 within 2× SLA
• Escalation rate sits in the 8–15% band with intentional pattern (not random bailouts)
• FCR climbs from baseline to 70%+
• CSAT comment themes shift from "took too long" toward "didn't solve my exact issue" — a sign you're now competing on resolution depth, not response speed
• The Friday scorecard takes 12 minutes instead of 25 because the patterns are familiar

The CSAT score itself? It mostly takes care of itself. That's the whole point. You stop watching the thermometer and start managing the room temperature.

The specialists who burn out are the ones who tried to grip the lagging number directly. The specialists who advance are the ones who built a leading-metric ritual, ran it weekly, and let the dashboard report whatever the work earned.

CSAT is yesterday's weather forecast. FRT, escalation rate, and the comments under your sub-3 ratings are this morning's sky. Read the sky. The forecast will follow.

For the role definition this playbook supports, see the Customer Support Specialist job description.