Chatbot metrics that matter: containment, handoff rate, CSAT and revenue per chat
The four chatbot metrics worth tracking weekly are containment rate (bot-resolved ÷ total conversations), handoff rate (escalated ÷ total conversations), CSAT (satisfaction score on resolved chats), and revenue per chat (sales attributed to chat ÷ total conversations). Compute all four from your own analytics and CRM, not from assumptions.
- Containment and handoff rate are two sides of the same number — they should roughly sum to 100% of conversations.
- CSAT for bot-resolved chats should be tracked separately from human-resolved chats to catch quality gaps early.
- Revenue per chat connects automation directly to business outcomes, not just conversation volume.
- Build a simple weekly dashboard from four numbers rather than drowning in every metric a platform offers.
- Watch for the pitfall of counting every conversation the same — a one-message enquiry and a completed sale are not equally meaningful.
Why these four metrics, and not more
Chatbot platforms surface dozens of numbers: messages sent, active users, response time, and more. Most businesses only need four to know whether the bot is actually working: how much it resolves on its own, how much it hands off, whether people are satisfied with the outcome, and whether it drives revenue. Everything else is useful for debugging a specific problem, not for a weekly health check.
The four core metrics and their formulas
Containment rate = Conversations fully resolved by the bot ÷ Total conversations, over a given period. Compute it from your platform's Analytics, filtering for conversations that never reached a human handoff and did not end in the customer clearly unresolved (check a sample manually if your platform does not distinguish 'resolved' from 'abandoned').
Handoff rate = Conversations escalated to a human ÷ Total conversations. Containment and handoff rate should roughly add up to 100%, minus any conversations that were simply abandoned with no resolution either way. A rising handoff rate over time, with no change in your enquiry mix, usually points to a knowledge base gap or a flow that is not handling a common request well, worth investigating rather than just accepting.
| Segment | What to check for CSAT |
|---|---|
| Bot-resolved conversations | Send a short satisfaction prompt at the end of the flow; track the average. |
| Human-resolved (post-handoff) conversations | Track separately using your live-chat team's existing process. |
| Gap between the two | A wide, consistent gap signals the bot needs improvement before you expand containment further. |
Revenue per chat = Total revenue attributed to chat-originated leads or orders ÷ Total conversations in the period. Attribution requires connecting your CRM pipeline stages (or order records) back to the conversation that started them: a lead created via the bot that later closes as a sale should be traceable. This is the number that turns 'the bot handled 2,000 conversations' into 'the bot contributed ₹___ this month', which is what actually matters to a business owner deciding whether to invest further.
A simple weekly dashboard
| Week | Containment | Handoff | Bot CSAT | Revenue per chat |
|---|---|---|---|---|
| 1 | ___% | ___% | ___/5 | ₹___ |
| 2 | ___% | ___% | ___/5 | ₹___ |
| 3 | ___% | ___% | ___/5 | ₹___ |
Four numbers, four weeks, one screen: enough to spot a trend without needing a dedicated analyst. Do not treat every conversation as equally meaningful when reading these numbers. A single 'what are your hours' message and a multi-turn conversation that ends in a ₹15,000 order both count as 'one conversation' in a raw count, but they are not equally important. Revenue per chat helps correct for this: a rising containment rate alongside a falling revenue per chat could mean the bot is efficiently handling low-value questions while missing higher-value ones, worth investigating specifically.
Computing these on Convo360
Convo360's Analytics dashboard reports conversation volume, handoff rate and response metrics directly. CRM pipeline tracks leads through stages so you can connect closed deals back to their originating conversation for revenue-per-chat attribution. Flow Logs with per-step notes help you diagnose why a specific conversation escalated, turning a rising handoff rate into a specific, fixable flow gap rather than a mystery number.
Two more numbers worth adding once the basics are steady
Once you have a few weeks of the four core numbers, two supporting metrics add useful context without turning the dashboard into a spreadsheet nobody reads.
| Metric | Formula | Why it helps |
|---|---|---|
| First-response time | Time from message received to first reply sent | A rising number, even with high containment, can mean flows are getting slower or a knowledge base lookup is timing out. |
| Repeat-contact rate | Conversations from a contact who messaged about the same topic within 7 days ÷ total conversations on that topic | A high repeat-contact rate suggests the first answer did not actually resolve the query, even if it was marked 'contained'. |
Neither number needs its own dashboard: track them monthly, alongside the weekly four, rather than daily.
Setting your own targets instead of chasing a benchmark
There is no fixed 'good' number for containment, handoff or CSAT that applies across businesses, since a bank's chatbot and a restaurant's booking bot have entirely different enquiry mixes. The useful exercise is setting a target against your own first month's baseline, then reviewing it quarterly as your flows and knowledge base mature.
- Run the bot for four weeks and record all four metrics without changing anything.
- Pick one metric to improve first, usually the one with the clearest gap (a high handoff rate on one specific topic, for instance).
- Make one flow or knowledge-base change, then measure again for two weeks before judging the result.
- Repeat quarterly, resisting the urge to change multiple things at once, since it becomes impossible to tell what caused a shift.
This turns the dashboard into a working tool for decisions, rather than a report nobody acts on.
Common measurement mistakes beyond the obvious
- Measuring over too short a window. A single day's numbers can swing widely due to one unusual conversation or a campaign send; use at least a week before drawing conclusions.
- Not segmenting by flow or topic. An overall containment rate of 65% can hide one flow performing at 90% and another at 20%; segment by flow name or trigger where your analytics allow it, so you know exactly where to focus.
- Ignoring conversations that were abandoned, not resolved. A customer who simply stops replying is neither 'contained' nor 'handed off'; count these separately so your containment number is not inflated by silence.
- Changing multiple flows in the same week you're trying to measure impact. Isolate changes so a shift in the numbers has one clear explanation.
Example conversation
Check your metrics on Convo360 Analytics
Start free in minutes — no code, no credit card. Bring your own AI keys and pay no per-message markup.
Frequently asked questions
What is a good containment rate for a chatbot?
There is no universal target — it depends on how repetitive your enquiry mix is. Track your own trend over time rather than comparing to a number quoted elsewhere.
How do I calculate revenue per chat if I don't use a CRM?
You can approximate it manually by tagging orders or deals that originated from a chat conversation in your sales register for a sample period, then dividing total attributed revenue by total conversations in that period.
Should containment rate and handoff rate always add to 100%?
Roughly, yes, minus any conversations that were simply abandoned with no clear resolution either way. If they consistently do not add up, check how your platform is classifying conversation outcomes.
How often should I review these metrics?
Weekly is a practical cadence for spotting trends early, with a deeper monthly review to connect metrics to business decisions like staffing or flow redesign.
Does a high handoff rate always mean the bot is failing?
Not necessarily — some businesses deliberately design flows to hand off complex or high-value conversations to a human quickly, which is a strategy choice, not a failure. Read handoff rate alongside CSAT and revenue per chat before judging it.
Related reading
Available across India — see WhatsApp & AI calling automation by state and city.