Which metrics to track in WhatsApp customer service and what a reasonable value is for each
Which metrics to look at in WhatsApp customer service and what is a reasonable value for each one
If you serve customers on WhatsApp, at some point you will wonder if you are doing it right. The problem is that most guides give you generic metrics designed for call centers or social media, and reasonable values on WhatsApp are not the same. This page is a reference table: what to measure, how to calculate it with data you already have, and what number is healthy in each case.
The values listed here do not come from a marketing manual: they come from measuring the real behavior of WhatsApp Business conversations in Latin American businesses. They are ranges, not promises. Your business may be outside them and still do well, but if you are far from these numbers, there is something to review.
The metrics that matter (and those that do not)
On WhatsApp there are dozens of metrics you can look at. Most are not useful for making decisions. These are the ones that do matter, because they measure something the customer notices and that you can improve:
- First response time: how long it takes a customer to receive a human response after writing.
- Resolution time: how long it takes for the conversation to close with a yes, a no, or a solution.
- First contact resolution rate: what percentage of conversations are resolved without the customer having to write again.
- Bounce rate: what percentage of customers write and never receive a response.
- Messages per conversation: how many exchanges are needed to resolve a topic.
- Peak demand hours: at what times messages are concentrated and whether your team can cover them.
Metrics that are NOT useful: total number of messages (it says little if you do not know how many conversations there are), likes or reactions (on WhatsApp they are not a signal of anything), and average response time calculated from your system's database if you do not separate human responses from automatic ones.
The reference table: reasonable values per metric
| Metric | Reasonable value | Alert | How it's calculated |
|---|---|---|---|
| Time to first response | Less than 5 minutes during business hours | More than 15 minutes | Add the time between the customer's first message and the first human response, divide by the number of conversations. |
| Resolution time | Less than 2 hours for simple queries | More than 24 hours | Add the time between the first message and the last message of the conversation, divide by the number of conversations. |
| First-contact resolution rate | 70% or more | Less than 50% | Count conversations that were closed without the customer writing again after the first response, divide by the total. |
| Bounce rate | Less than 5% | More than 10% | Count conversations where the customer wrote and never received a response, divide by the total. |
| Messages per conversation | Between 4 and 10 | More than 15 | Divide the total number of sent messages by the number of conversations. |
| Peak hours coverage | 80% or more of messages answered in less than 5 minutes | Less than 60% | Identify the hour with the most messages and measure the first response time in that time slot. |
These numbers assume there is a human on the other side. If you use automatic responses for the first contact, the first response time will be almost zero, but that doesn't mean the service is good: the customer notices the difference between an automatic response and a human one. The metric that measures real quality is the first-contact resolution rate.
How to calculate these metrics with your own history
You don't need an expensive tool to get started. If you export the WhatsApp Business conversation history (or have it in a database), you can calculate the main metrics with a spreadsheet. The key step is to separate customer messages from agent messages, and use the correct timestamp column: the one that marks when the message was sent, not when it was loaded into your system.
- 1Export the conversation history with the date and time of each message, and who sent it (customer or agent).
- 2Separate the conversations: each message thread that starts with a customer is a conversation.
- 3For the first response time: subtract the time of the customer's first message from the time of the agent's first message.
- 4For the resolution time: subtract the time of the customer's first message from the time of the last message in the conversation.
- 5For the bounce rate: count conversations where the customer wrote and the agent never responded.
- 6For the first-contact resolution rate: count conversations where the customer did not write again after the agent's first response.
Watch out for a common mistake: if you imported the history from another tool, messages may have two different dates (when they were sent and when they were loaded). Always use the send date. If a single hour or day concentrates more than a quarter of your messages, you are likely looking at the load date, not the real one.
What to do if your numbers are out of range
Each out-of-range metric points to a different problem. This table tells you what to check first:
| Metric out of range | What might be happening | What to check |
|---|---|---|
| High first response time | Not enough people covering the schedule, or messages get lost in a messy inbox | Number of agents per shift, inbox order, notifications enabled |
| High bounce rate | Messages come in but nobody sees them, or the system marks them as read without replying | Inbox settings, alerts for new messages, schedule coverage |
| Low resolution rate | Replies are generic and don't solve the issue, or the customer has to insist | Quality of replies, whether there are templates or guides for common cases |
| Very high messages per conversation | Missing info in the first reply, or the customer doesn't understand what's offered | First reply scripts, available product or service information |
| Low peak hours coverage | The highest demand schedule doesn't match the support schedule | Team schedules, out-of-hours auto-replies, shifts |
The general rule: first fix the bounce rate (if nobody replies, the rest doesn't matter), then the first response time, and only then the resolution rate. A conversation that takes time but gets resolved is better than a quick one that solves nothing.
The special case: auto-replies and AI
If you use auto-replies or an AI assistant, the metrics change meaning. The first response time will be almost zero, but what you need to measure is something else: what percentage of conversations the tool resolves on its own and how many end up escalating to a human.
A reasonable value for a well-configured assistant is to resolve between 40% and 60% of simple queries (hours, prices, location, order status) without human intervention. If it resolves less than 20%, it's more of a nuisance than a help. If it resolves more than 80%, you're probably losing queries that needed human judgment and you don't know it.
To measure this, count the conversations where the customer didn't ask to speak to a person and the issue was closed with an auto-reply. That's your automatic resolution rate. If you're not measuring it, you don't know whether the tool is saving you time or creating angry customers.
Common measurement mistakes
- Mixing automatic messages with human ones: if you count auto-replies as if they were human, the first response time will always look good and you won't see the real problem.
- Using the upload date instead of the send date: if you imported a history, messages may all appear loaded on the same day even though they were sent on different weeks.
- Measuring only the average: if one day you didn't reply at all and another day you replied in seconds, the average can look good. Also look at the median and the worst day.
- Not separating by query type: a billing query may take longer than a hours query. Measure separately if you can.
- Comparing your number with a call center's: on WhatsApp the customer expects immediacy, not a wait of minutes. Standards from other channels don't apply.
How often to review these metrics
Support metrics are not reviewed once a year. The minimum recommended frequency is weekly for first response time and bounce rate (these degrade the fastest), and monthly for resolution rate and messages per conversation (they need more volume to be stable). If your business has seasonal peaks, always compare against the same period of the previous year, not against the previous month.
A single week with bad numbers is not a crisis: it could be a holiday, a technical issue, or a campaign that brought more inquiries than usual. Only when the number stays bad for three consecutive weeks should you intervene.
The metric that is most neglected and the one that says the most is the bounce rate. If a customer writes and no one responds, you not only lose that sale: you lose the next one, because that person will not write again. A 10% bounce rate means that one out of every ten customers who contact you leaves without a response. That number should be the first one you look at every Monday.
If you want to automate part of these measurements, tools like Wando can help you organize your inbox and provide visibility into response times, but the metrics can still be calculated with a spreadsheet and an exported history. No paid tool is needed to start measuring.
Frequently asked questions
What is a good response time on WhatsApp?+
During business hours, a first response time of less than 5 minutes is reasonable. More than 15 minutes is already an alert. These numbers apply to human responses: if you use automatic ones, the time will be almost zero but that does not measure quality.
How do I calculate the bounce rate on WhatsApp?+
Count the conversations where the customer wrote and never received a human response, and divide by the total number of conversations. A reasonable value is less than 5%. More than 10% is a serious alert.
What does it mean if the first-contact resolution rate is low?+
It means the customer has to write again because the first response did not solve their issue. A reasonable value is 70% or more. If it is below 50%, review the quality of your responses: they are probably generic or incomplete.
Do automatic responses improve WhatsApp metrics?+
They improve the first response time, but not the quality. If you use automatic responses, measure separately what percentage of queries they resolve on their own. Between 40% and 60% of simple queries is reasonable. Less than 20% means the tool is not helping.
How often should I review these metrics?+
Weekly for first response time and bounce rate, monthly for resolution rate and messages per conversation. Only intervene when a number stays bad for three consecutive weeks.