Formula to Estimate the Cost of an Uncovered Peak Hour

What this formula is for

An uncovered peak hour is not measured by what the team did, but by what it did not manage to do. When more inquiries come in than can be answered, three things happen at the same time: some inquiries never get a response, some customers write again (retry), and the team ends the hour more exhausted than it started. All three have a cost, and all three can be estimated with the same handful of numbers you already have on hand.

The formula does not replace a staffing decision or a conversion analysis. It serves another purpose: to put a number on the table when someone asks whether it is worth adding a person, moving a shift or changing the service hours. It is an order-of-magnitude calculation, not a promise.

The three parts of the cost

The total cost of an uncovered peak hour breaks down into three addends that should be calculated separately, because they are measured differently and corrected differently.

• Lost inquiries: those that came in during the peak hour and never received a response, or received it so late that the customer had already resolved it elsewhere.
• Retries: the same customer who writes again because they were not answered, plus the work generated by handling the same inquiry twice.
• Team strain: the effect of sustaining a peak hour without reinforcement on the service provided during the following hours, which tends to be the most expensive addend and the least measured.

The formula, step by step

First, the unanswered inquiries from the peak hour are estimated. Everything that came in during that hour is taken and what the team was actually able to answer in that same hour is subtracted. The difference is the queue that went uncovered. If the team kept answering afterward, that queue is spread between the following hour and the next day: it doesn't disappear, it shifts.

Then a value is assigned to each inquiry. The value is not the price of the product or the average ticket: it is the margin you expected to earn from that conversation multiplied by the probability that the conversation ends in a sale. If you don't have that probability measured, use a range and do the calculation twice, with the floor and with the ceiling. An honest range is more useful than a made-up number.

The third step is the retry. Every customer who didn't get a response writes again, and that second inquiry takes the same agent time as the first. It is estimated as a percentage of the unanswered inquiries. That percentage is not the same for everyone: it depends on the industry and the time of day. If you don't have it measured, calculate the scenario with a low value and a high one, and write down the assumption.

The fourth step is the drain. Here there is no closed formula, but there is a way to estimate it without making things up: the time the team spends answering the backlog in the following hour is taken, which is time it is not spending on that hour's new demand. That time has the same hourly cost as the rest of the service, and it is the cost you pay without seeing it.

An example with round numbers

Let's suppose a peak hour in which 120 inquiries come in and the team manages to answer 90. The uncovered queue is 30 inquiries. If 40% of those 30 are resolved elsewhere before someone answers, the lost inquiries are 18. If the expected margin per closed conversation is 8,000 pesos and the channel closing probability is 25%, the value of each inquiry is 2,000 pesos, and the 18 lost inquiries cost 36,000 pesos.

Retries are calculated on the 30 unanswered. If 30% write again, that is 9 repeated conversations. Each one takes the same agent time as the original, so the cost is that of 9 extra services. The drain is estimated separately: if answering the backlog consumes 40 minutes of the following hour, that time is paid at the team's hourly cost.

The total for the hour is the sum of the three. What matters is not the final number, but the comparison: that total is placed next to the cost of covering the peak hour with one more person or with a tool. If the cost of covering is less than the cost of not covering, the decision stops being a discussion of opinions.

How to measure the uncovered queue without fooling yourself

The most common mistake is measuring the queue with the wrong clock. If you count when each message was saved in the database instead of when the customer sent it, any analysis of time slots will give you a false picture: you will see a peak where there was actually a bulk data load. To know what hour the demand comes in, you have to use the message timestamp, not the row timestamp.

The second mistake is counting an inquiry that was answered outside the peak hour as answered. That response exists, but it did not cover the hour. For the queue calculation, the valid response is the one that occurs within the window you are measuring.

• Separate incoming volume from response volume by hour, always using the message timestamp.
• Define the peak-hour window before looking at the data, not after.
• Count conversations, not individual messages: a customer who sends three messages in a row is one inquiry.
• Write down the retry and closure assumptions in the same place where you store the calculation, so you can review them later.

Burnout: the addend that appears in no spreadsheet

Burnout does not show up in the message report because it is not a message. It shows up in the following hour: the team arrives with backlogged work, new demand waits longer than usual, and some inquiries from that hour are also lost. It is a cost paid twice, and that is why it is worth measuring as a chain and not as an isolated event.

A simple way to estimate it is to compare the response rate of the hour following a peak against the response rate of an equivalent hour without a peak. The difference in unanswered inquiries during that following hour is the peak spillover. If the spillover is large, the problem is not the peak hour: it is that there is no capacity buffer.

It is also worth looking at turnover and absenteeism, but carefully: these are indicators that move for many reasons and cannot be attributed to a single hour of the day. They serve as context, not as proof.

When the problem can no longer be fixed with the formula

The formula is useful for sizing. It stops being useful when the problem is structural: if every week there is a peak hour that overflows, if the spillover eats into the following hour, if the team starts prioritizing by urgency instead of by order of arrival, or if lost inquiries always concentrate in the same time slot. In those cases the calculation has already given you the answer and what is missing is a coverage decision.

That is where the options that are not about adding people come in: shifting the shift, extending the auto-reply schedule, using templates for the most repeated inquiries, or relying on AI suggestions that the team approves with one click. Wando works like this: the AI suggests the reply and the team approves it, and it learns from the business's real replies instead of depending on programmed flows. That does not eliminate the peak hour, but it changes how many inquiries remain uncovered.

Checklist to apply the calculation in your operation

• Choose a single peak hour to start with, the most obvious one.
• Extract the inbound and reply volume for that hour, with the message timestamp.
• Calculate the uncovered queue.
• Define the expected margin per conversation and the channel's closing rate. If you do not have them, use a range.
• Estimate the retry percentage with two scenarios, low and high.
• Measure the spillover minutes in the following hour.
• Add the three components and write down the total with the date.
• Compare that total against the cost of covering the hour with one more person or with a tool.
• Repeat the calculation one month later with the same assumptions and compare.

Frequently asked questions

The questions that always come up when applying this formula have to do with what counts as a lost inquiry, how the value of each conversation is chosen, and how often it makes sense to repeat the calculation. The short answers are below; the details of each one are in the sections above.