Formula to Measure the After-Hours Unanswered Conversation Rate (with Weekly Tracking Table)

Most businesses know they miss messages at night or on Sundays, but they don't know how many. Without that number, any decision about schedules, on-call shifts or auto-replies is made by guesswork. This page solves one single thing: how to calculate the rate of unanswered conversations outside business hours and how to track it week by week with a table that takes five minutes to fill in.

This is not an article about 'improving customer service.' It is a formula, a set of prior definitions and a spreadsheet. If you apply the calculation with your own data, you will have a number you can actually discuss with your team.

What exactly this rate measures (and what it doesn't)

The rate of unanswered conversations outside business hours is the proportion of conversations that came in while the business was closed and never received a human or automated response within a window you define. It does not measure satisfaction, it does not measure lost sales in currency and it does not measure the quality of the response. It measures one single thing: how many conversations were left hanging.

It is worth separating it from two other metrics that are often mixed together and ruin the analysis: first response time (which is measured in minutes and applies to all conversations, not just those outside business hours) and the rate of unassigned conversations (which measures internal distribution, not the schedule). If you mix all three into a single dashboard, you won't be able to act on any of them.

• What it measures: conversations outside business hours that received no response at all.
• What it does NOT measure: how long the response that did arrive took.
• What it does NOT measure: whether the customer bought afterward or not.
• What it does NOT measure: whether the response was good, correct or useful.

Prior definitions: without these, the number is useless

Before calculating anything, three definitions must be set in writing. If they change halfway through, the historical series stops being comparable and the tracking table loses all value.

1. 1Define the business hours by days and time slots. Example: Monday to Friday from 9:00 to 18:00, Saturdays from 9:00 to 13:00, Sunday closed. It has to be written down somewhere, not just in someone's head.
2. 2Define what counts as 'out of hours'. It is every incoming message whose timestamp falls outside those time slots, in the business's time zone.
3. 3Define the grace window: how much time after opening the team is given to respond to what came in while closed. Example: 60 minutes from opening, or until noon of the next business day. Without this window, any message from Sunday night counts as 'unanswered' on Monday at 9:01, which is false.

The formula

The formula is a division. The only real work is in counting the numerator and the denominator correctly.

Out-of-hours unanswered conversation rate = (conversations that came in out of hours and did not receive a response within the grace window) ÷ (total conversations that came in out of hours) × 100.

Example with round numbers to check that you understood the calculation: if in one week 200 out-of-hours conversations came in and 34 remained unanswered within the grace window, the rate is 34 ÷ 200 × 100 = 17%. That 17% is the number that goes in the table.

How to count with the timestamp column (and why not to use created_at)

If you export data from a messaging platform, you will almost always find two date columns: one that records when the row was inserted into the database (created_at) and another that records when the event occurred according to the channel (timestamp). For this calculation you must use timestamp and nothing else.

The reason is concrete: when history is imported from a migration, all the rows are inserted within a few days, so created_at shows that 'everything happened on a Tuesday at 15:00' and any analysis of hours gives nonsense. With timestamp, on the other hand, the distribution reflects the real behavior of customers: peaks during business hours and a percentage of out-of-hours messages that usually falls in the range of 15% to 20% in customer service businesses. If you get 56% of messages in a single hour of the day, you are looking at the wrong column.

• Use timestamp for everything related to 'when it happened'.
• Use created_at only to audit internal data loading processes.
• Before publishing any distribution of hours, look at the histogram by hour: if it has a single absurd peak, the wrong column was chosen.
• Save the time zone you used for the calculation. If the team looks at the data from another time zone, the numbers will not match.

Weekly tracking table

This is the table to fill in every week. It is copied into a spreadsheet and filled in with five numbers per week. The rate column is calculated, not estimated.

The 'Note of the week' column looks decorative, and it isn't. Holidays, vacations of whoever covers the on-call shift, system outages, or mass campaigns that generate spikes explain jumps that would otherwise look like an attention problem and are something else. Without that note, three months later you won't be able to interpret the series.

How to read the result: four scenarios

The number alone says nothing. What says something is the number combined with the absolute volume and with the trend. These are the four typical scenarios and what to do in each one.

The most common mistake is looking only at the rate and not at the volume. 3% of 20 weekly conversations is 0.6 conversations: it doesn't justify setting up an on-call shift. 3% of 800 weekly conversations is 24 conversations: it does justify reviewing something. The rate orders, the volume decides.

What to do with the number (without overreacting)

Once you have the rate, there are three levers and it's best to evaluate them in order of increasing effort.

1. 1Auto-reply by schedule: an automatic message that confirms receipt and states when a reply will be given. It is the cheapest lever and does not require changing the team's staffing. It does not eliminate unanswered conversations, but it changes the customer experience and usually reduces complaints about silence.
2. 2Partial on-call coverage: a person available during the time slot where off-hours volume is concentrated. To define which slot, look at the hourly histogram of unanswered conversations: there is almost always one or two hours that concentrate most of them.
3. 3Extended hours: the most expensive lever. It is only justified if the rate remains high after the previous two and the volume supports it.

Any of the three can be implemented with a tool that suggests replies and allows auto-replies by schedule, such as Wando, but the rate calculation is independent of the tool: it can be done with an export to a spreadsheet and fifteen minutes per week.

The five mistakes that ruin the calculation

• Counting messages instead of conversations. A customer who writes four times at night counts as a single conversation.
• Using created_at instead of timestamp. The result will show that everything happens at the same time and you will not be able to draw any conclusion about schedule.
• Not defining the grace window. Without a window, everything that comes in at 20:00 and is answered at 9:00 the next day counts as 'unanswered', which inflates the rate and demoralizes the team.
• Changing the business hours halfway through the series without noting it. The series stops being comparable and you will not be able to see whether the measure worked.
• Looking only at the monthly average. One week at 40% and three at 2% gives an average of 11.5% that does not describe any reality and hides a specific problem.

How to know the calculation is done right

Three quick checks before trusting the number. If all three pass, the rate is defensible before the team and before anyone who asks for it.

• The hourly histogram of incoming conversations shows a reasonable business curve, with peaks during business hours and a percentage of off-hours messages consistent with the type of business.
• The numerator is always less than or equal to the denominator. If it is greater, there is a problem with conversation grouping.
• The sum of off-hours conversations plus in-hours conversations gives the total conversations for the period. If it does not add up, there are conversations without an assigned timestamp or with a different time zone.