Formula to calculate daily workload per WhatsApp operator: active, pending and closed conversations

Why the number of open chats says nothing

If you measure an agent's workload by counting how many chats they have open in the inbox, the number will almost always lie to you. An open chat may be waiting on the customer, may be waiting on a colleague, may be open because nobody dared to close it, or may be a conversation that already ended and was left unchecked. The total number of open chats mixes real work with phantom work.

Daily workload is calculated by separating three distinct states: active conversations (where the agent has the ball), pending (where someone else has the ball: the customer, a colleague, or a system), and closed (where nothing is left to do). Only with those three numbers separated can you know whether the agent is overloaded, whether they are waiting, or whether the problem is not the number of people but how conversations are distributed.

This page does not give you a universal number of chats per agent, because that number depends on your industry, your average ticket, and how long your team takes to respond. What it gives you is the formula to calculate it with your own data and the table to track it day by day.

The base formula: daily workload per agent

An agent's daily workload is measured in active conversations handled during the day, not in messages. Messages are a bad indicator because a conversation can have 3 messages or 40, and in both cases it is a single conversation that has to be opened, understood, and resolved.

The formula is: Daily workload = Active conversations of the day + Pending conversations carried over from the previous day. Closed conversations are counted separately, as a production indicator, not a workload indicator. If you mix closed with active, you are measuring two different things in the same number.

A simple example: an agent starts the day with 8 pending conversations they did not close yesterday. During the day, 35 new conversations are assigned to them. Their daily workload is 43. If they closed 30 and left 13 pending for tomorrow, the day's workload was 43, production was 30, and the carryover is 13. All three numbers matter and mean different things.

How to classify each conversation without thinking twice

The practical problem with this formula is that classifying conversations by hand is slow and nobody sustains it for more than a week. The solution is to define a simple classification rule and always apply it the same way. The rule must be able to resolve any conversation in less than five seconds.

1. 1Active: the last message in the conversation is from the customer and has not yet received a reply from the team.
2. 2Active: the last message is from the team but the customer replied with something that requires a concrete action (a piece of data, a confirmation, a referral).
3. 3Pending: the last message is from the team and the customer did not reply. The ball is in the customer's court.
4. 4Pending: the conversation is waiting on a third party (a colleague, an area, a supplier) and there is a note or tag indicating it.
5. 5Closed: there is an explicit resolution (sale, answer delivered, query resolved) and a tag or status that marks it.
6. 6Closed by abandonment: the customer did not reply after a period you define (for example 72 hours) and the conversation is archived with a tag different from the resolved one.

The daily tracking table per operator

With the five indicators you build a table that is filled out once a day, at the close of the shift. No more frequency than that is needed. The table can be kept in a shared spreadsheet or directly in the statistics panel of the tool you use, as long as you can separate active, pending and closed.

Looking at that table shows what an aggregate total does not. Operator C has the highest load (56) and the highest carryover (34): they are receiving a lot and are not able to close. Operator D has the lowest load (24) and closes almost everything that comes in (21 out of 24). Operator A is in the middle. The decision is not 'hire someone' or 'push C to hurry up': it is understanding why C accumulates. It could be that they get the hardest conversations, that they are doing tasks that are not theirs, or that their shift is poorly placed relative to the message peak.

How to read the numbers: the four warning signs

A table without reading criteria is useless. These are the four patterns that indicate different things and that should be reviewed before making any staffing decision.

• Growing carryover several days in a row: the team is receiving more than it can close. It is not a problem of one operator, it is a problem of total capacity.
• High load with high closed: the operator is saturated but performing. Here the conversation is about sustaining the pace, not about intervening.
• Low load with low closed: the operator is not receiving work or is taking too long per conversation. The distribution of assignments must be looked at before individual performance.
• Inherited pending that never goes down: there are conversations nobody wants to close. This usually happens when closing requires an uncomfortable decision (saying no, charging, escalating).

None of these signals is read in a single day. They are read over a week of the table. A bad day is a bad day; five days with the same pattern is a process problem.

How long each conversation takes: the missing data in the formula

The daily load tells you how many conversations there are, not how much time they consume. Two operators with 40 conversations each can have very different real loads if one handles price inquiries and the other resolves complaints. To capture that, the formula is complemented with an average duration per conversation type.

The calculation is: Service hours = Daily load × Average duration per conversation. If your average daily load is 40 conversations and each one takes 6 minutes of effective service, that's 240 minutes, or 4 hours of real service. That number is compared against the available shift hours and tells you whether the operator has room or is at the limit.

The important takeaway from this table is that an operator with 18 complaints has a higher real load than one with 45 price inquiries, even though the number of conversations is much lower. If your tracking table doesn't distinguish conversation types, you'll distribute work poorly and reward the one who was lucky enough to receive easy inquiries.

How to use the formula to decide if you need more people

The question 'do I need to hire someone?' is answered by the team's total load, not by one operator's load. The calculation is: Team capacity = Sum of available service hours of all operators. Team demand = Sum of service hours required by all the day's conversations. If demand consistently exceeds capacity, you need more people or you need automation of the repetitive tasks.

There's a third path before hiring: reduce the average duration per conversation. That's achieved with saved replies, templates for the most frequent inquiries, and an inbox that suggests replies instead of forcing you to write them from scratch. It doesn't always suffice, but it's faster and cheaper than adding a full shift, and it's measured with the same indicators.

Common mistakes when measuring workload per operator

These are the mistakes that make the formula produce numbers that are useless for deciding anything. If you are making one, fix it before drawing conclusions.

• Counting messages instead of conversations: it inflates operators who handle chatty inquiries and hides those who resolve in two messages.
• Counting open chats as workload: it mixes active with pending and with conversations that already ended but nobody closed.
• Not separating resolved from abandoned: it counts as output what was actually lost.
• Measuring only the team average: it hides the overloaded operator and the idle operator. The table is read per person.
• Changing the classification rule mid-week: the numbers stop being comparable. The rule is defined once and sustained.
• Not recording duration by conversation type: it makes workloads that are not equivalent look the same.

Frequently asked questions

The answers to the most common questions about how to apply this formula in a real team.