Sales metrics were mostly designed for phone and email teams: calls made, emails sent, meetings booked. When your team sells over text and WhatsApp, those numbers don’t fit. A rep might handle forty conversations in a day without making a single call, and the deal might close in a thread that spans three weeks and two channels.
Messaging teams need metrics that reflect how chat actually works: conversations arrive at all hours, customers reply in bursts, and the difference between a won and lost deal is often whether someone followed up.
Here are seven metrics worth tracking, with definitions, how to compute them, what good looks like and the traps to avoid. We’ve deliberately left out benchmark numbers. Response expectations vary by industry, channel, hours and customer, and an industry average you found online is less useful than your own trend over time.
Before you measure: define a conversation
Every metric below depends on what counts as a “conversation”. Decide this up front and write it down. A common approach:
- A conversation is a thread with one customer. A new conversation starts when the customer messages after a defined period of inactivity (for example, 7 days), or when the previous conversation was closed.
- Business hours are the hours your team is expected to reply. Decide whether metrics count only business hours or the full clock, and be consistent.
- Automated messages (auto-replies, away messages) don’t count as responses.
Without these definitions, two people will compute the same metric two different ways.
1. First response time
What it measures: How long a customer waits for the first human reply after starting a conversation. See the glossary entry on first response time.
How to compute it: For each new conversation, take the timestamp of the first human reply minus the timestamp of the customer’s first message. Report the median across conversations, plus the share of conversations answered within your target (for example, within your stated business-hours window).
What good looks like: Consistent and predictable. A team whose median is steady and whose slowest replies aren’t wildly slower than the rest is in good shape. Improvement shows up as the slow tail getting shorter.
Pitfalls:
- Averages hide problems. A few overnight messages can make an average look terrible, or a burst of instant replies can hide a handful of leads that waited a day. Use the median and look at the slowest 10% separately.
- Auto-replies inflate the result. “Thanks, we’ll get back to you!” is not a response.
- After-hours rules. Decide whether a message at 10 PM is measured from 10 PM or from opening time the next morning.
Speed matters most for new leads. Our article on lead response time goes deeper into why.
2. Time to useful response
What it measures: How long until the customer gets a reply that actually moves things forward: an answer, a quote, a booking option, a real next step.
How to compute it: This one needs a human definition. A practical approach: have reps tag the message that answers the customer’s question, or measure time until a key event (appointment proposed, quote sent, question answered). Sample 20 conversations a week if full tagging is too much.
What good looks like: A small gap between first response and useful response. If the first reply is fast but the useful reply takes days, your team is acknowledging customers without helping them.
Pitfalls:
- Hard to automate fully. Accept a manual sample rather than skipping it.
- “Let me check and get back to you” can hide in first response time. This metric exposes it.
3. Unanswered threads
What it measures: How many conversations are currently waiting on your team, where the customer sent the last message and nobody has replied.
How to compute it: Count open conversations where the last message is from the customer. Break it down by age (under 1 hour, 1–4 hours, same day, older) and by owner.
What good looks like: A small number, mostly young, with nothing old and nothing without an owner. This is a snapshot metric: check it several times a day, not once a month.
Pitfalls:
- Conversations that don’t need a reply. “Thanks!” or a thumbs-up can count as unanswered. Let reps close or mark these.
- Unassigned threads. If a thread has no owner, it doesn’t show up in anyone’s list. Track unassigned conversations separately. Clear conversation assignment rules help.
4. Follow-up completion
What it measures: Of the follow-ups your team planned, how many actually happened on time.
How to compute it: Take all follow-ups that were due in a period (reminders, scheduled tasks or conversations where your team sent the last message and the customer went quiet). Divide the number completed by the due date by the total due.
What good looks like: High and steady, with misses concentrated in specific, explainable situations (a rep out sick, a holiday week) rather than spread everywhere.
Pitfalls:
- Only measuring follow-ups someone remembered to schedule. If your system doesn’t create follow-ups automatically when a customer goes quiet, the metric only covers the disciplined reps.
- Rewarding volume. A follow-up that just says “Any update?” counts the same as a helpful one. Look at a sample of messages, too. See sales follow-up cadence for what good follow-ups contain.
5. Conversation-to-deal rate
What it measures: The share of new conversations that turn into a deal in your pipeline, or into a won deal, depending on what you want to see.
How to compute it: Count new lead conversations started in a period. Count how many of them became a deal (or a won deal) within a defined window, for example 60 days. Divide.
What good looks like: Stable or improving over time, and similar across reps handling similar leads. Big differences between reps are worth a conversation, not a judgment.
Pitfalls:
- Mixing lead and customer conversations. Existing customers asking about appointments shouldn’t be in the denominator. Tag or filter new-lead conversations.
- Window length. If your sales cycle is long, a 30-day window will undercount.
- Deals created outside chat. If some deals come from phone calls, decide whether to count them.
6. Conversion by channel
What it measures: How conversation-to-deal rate differs by channel: SMS, WhatsApp, Instagram, web form.
How to compute it: Apply metric 5 separately to each channel where the conversation started.
What good looks like: There’s no “right” answer here. The point is to understand your channels. You might find WhatsApp leads ask more questions but convert well, or that Instagram DMs bring lots of inquiries that rarely book.
Pitfalls:
- Customers switch channels. Someone may DM on Instagram and then text. Credit the channel where the conversation started, and be consistent.
- Small numbers. With few leads on a channel, a couple of deals can swing the rate a lot. Look at several months before drawing conclusions.
- Channel isn’t the only variable. Different channels may bring different kinds of customers.
7. Workload per rep
What it measures: How much conversation work each person is carrying.
How to compute it: For each rep, count open assigned conversations, new conversations assigned per day and messages sent per day. Look at the distribution across the team.
What good looks like: Balanced relative to each person’s role and hours, with nobody consistently overloaded. When workload is balanced, response times usually are too.
Pitfalls:
- Message counts reward chattiness. Ten short messages aren’t more work than one careful quote.
- Using it to rank people. Workload is for planning and fairness, not leaderboards. If one rep handles the hardest customers, their numbers will look different.
- Ignoring complexity. A remodeling quote thread and an appointment confirmation are both “one conversation”.
Putting the metrics together
You don’t need all seven on day one. A practical sequence for a small team:
- Start with unanswered threads and first response time. They’re the easiest to act on today.
- Add follow-up completion once you have a follow-up routine.
- Add conversation-to-deal rate and conversion by channel once you have a couple of months of pipeline data.
- Use workload per rep whenever response times slip, to see whether it’s a process problem or a capacity problem.
Review them weekly, together, in a short meeting. Look at trends, not single weeks, and always look at a few real conversations behind the numbers.
Where inChat fits
inChat is a conversation-first CRM for small teams that sell over text and WhatsApp, and it’s in development. Analytics for response time, volume by channel and conversion are part of what’s being built, alongside response timers and follow-up reminders. If you manage a sales team that lives in chat, you can request early access to see how these metrics are designed to work.