The Measurement Problem

Every support tool shows you 30+ metrics. Response time, handle time, first contact resolution, CSAT, NPS, ticket volume by channel, by hour, by agent, by category, by phase of the moon.

Small teams don't need this. You need to answer one question: is our support good and getting better?

Four numbers answer that. Everything else is noise until you have 20+ agents.

The Four Numbers

1. First Contact Resolution Rate (FCR)

What percentage of tickets are resolved without the customer having to follow up?

Target: 75%+

How to track: count tickets where the customer didn't contact you about the same issue within 7 days. Or add a "was this resolved?" yes/no to your responses.

If FCR is low: your auto-responses might not be answering the question fully, or your agents are sending "I'll get back to you" instead of solving on the spot.

2. Median Response Time

How long until the customer gets a response? Use median, not average — averages get skewed by one 48-hour outlier.

Target: Under 1 hour during business hours. Under 5 minutes for auto-resolved messages.

How to track: most support tools calculate this automatically. If you're using email, track the time between the customer's message and your first reply.

If it's too high: add more auto-responses for common questions, or set up notifications so tickets don't sit unnoticed.

3. Automation Rate

What percentage of messages are resolved without human involvement?

Target: 50-70% for a well-configured system. Under 40% means you need more routing rules. Over 80%, spot-check accuracy — you might be auto-resolving things you shouldn't.

How to track: auto-resolved messages / total messages.

If it's low: look at your top un-automated intents and create rules for them. Each new rule bumps the rate.

4. Customer Satisfaction (CSAT or thumbs up/down)

Do customers think the support was helpful?

Target: 85%+ positive. Anything under 75% needs immediate attention.

How to track: add a thumbs up/down to auto-responses. For human responses, send a brief "was this helpful?" follow-up. Keep it simple — one click, not a survey.

If it's low: read the negative responses. They'll tell you exactly what's wrong — bad auto-responses, slow follow-up, or unresolved issues.

The Weekly Review (15 Minutes)

Every Monday, check these four numbers:

1. FCR this week vs last week — going up or down?
2. Median response time — any spikes?
3. Automation rate — changing? Why?
4. CSAT — any negative responses to review?

If all four are stable or improving, you're done. 5 minutes.

If one is declining, dig into why. Check the specific tickets that drove the change. This takes another 10 minutes.

That's your entire QA process. 15 minutes a week.

What to Ignore (For Now)

Net Promoter Score (NPS). NPS measures overall brand sentiment, not support quality. It's useful at scale but overkill for small teams and not directly actionable for support.

Handle time per agent. If you have 2 people doing support, you already know who's fast and who's thorough. You don't need a metric for it.

Channel distribution. "40% email, 30% chat, 30% social." Nice to know, but what are you going to do about it? Just answer wherever they contact you.

Ticket reopens. This is just FCR measured differently. Track one, not both.

Response time by hour of day. Unless you're scheduling shifts, hourly response time data is trivia, not actionable.

When to Add More Metrics

Scale triggers for more measurement:

• 5+ support agents: Add per-agent metrics (FCR per agent, CSAT per agent). This helps with coaching and workload balancing.
• Contractual SLAs: Add SLA compliance tracking. Only if customers have SLA agreements with penalty clauses.
• Multiple channels with different strategies: Add channel-specific metrics. Only if you're intentionally treating channels differently (e.g., faster SLA for chat vs email).
• Enterprise sales: Add ticket-to-customer-value correlation. Know which high-value accounts have support issues.

Until you hit these triggers, four numbers and 15 minutes is everything you need.