❓ Frequently Asked Question or Initial Context
When reviewing your team's speed and efficiency metrics, it's common for some questions to arise:
Why do some records show a response time equal to 0?
Why are the times I calculate manually in Excel sometimes different from those shown in the platform report?
How should I interpret response times when conversations happen outside business hours?
This article will explain exactly how these times are measured and why the platform prioritizes the "median" as the main metric.
🤓 Explanation
Response time refers to the amount of minutes (or hours) an agent takes to answer a customer's message once the conversation has landed in their inbox.
This time can be broken down into two key indicators:
First response time: How long it takes the agent to send their first message.
Response time: The general time it takes the agent to respond to each subsequent message from the customer.
However, sometimes these values appear as 0 or don't match basic manual calculations. Let's see why.
⚙️ Troubleshooting: Step by Step
📌 Case 1: Response Times Equal to 0
Reason: In these conversations, no customer response was recorded after the agent's initial message.
🔍 Common Scenario: Imagine you send an automated campaign (flow) via API and the chat is immediately assigned to an agent, who greets the customer. If the customer reads the message but doesn't interact or reply, no real conversation cycle is generated. Since there's no second customer message that needs to be answered, the system has no parameters to calculate a "response time".
Expected Result: The report will mark a response time of 0. This is completely normal and does not represent a system error.
📌 Case 2: Differences in Manual Calculation (Use of Median)
Reason: Unlike many basic systems, the Service Level report in Atom uses the median (not simple average) as the statistical basis for calculating your team's overall times.
🔍 Real Example: In a metrics review, you noticed that the overall time you calculated in Excel by adding and dividing all responses (average) was 15 minutes, but Atom showed 5 minutes. This difference exists because the average is affected by outlier values, while the median seeks the real center of your operation.
✅ Why Does Atom Use Median Instead of Average?
The choice of median responds to high statistical quality standards for Call Centers and customer service:
Lower Impact of Extreme Values (Outliers): If 9 customers are served in 1 minute, but 1 customer was served in 60 minutes (because the agent went to the bathroom and forgot to pause), the average would spike, making it seem like the entire team is slow. The median isolates that error and shows that your actual time is 1 minute.
Better Representation of Typical Behavior: In chat support, there are usually many extremely fast responses and a few very slow ones (bias). The median gives you a more accurate view of what the "typical customer" experiences.
Easy to Interpret: It indicates exactly the midpoint of ordered times, offering a clear and consistent reference of your actual service level.
💡 Want to calculate the average anyway? If you need to use the simple average for your internal reports, no problem. You can export the Audit tab (in .xlsx or .csv format) and calculate the average formula directly on the times column in your spreadsheet.
📝 Additional Notes and Rules
Customer Interactions: Remember that cases where the customer doesn't interact (doesn't respond) will not generate valid response times.
Subsequent Messages: Response time only starts being calculated when there is a customer message sent after the chat has been assigned to the agent.
Hours: Times generated outside business hours will also be included in the total calculation, unless you use the top filters to limit the view exclusively to your "Business Hours".
Reassignments: When calculating an agent's metric, the system will take into account only that agent's current responses, ignoring how long previous advisors took when they had that same chat before reassigning it.
Measure your team's real speed with precision! 🚀
Understanding that cases without responses are marked as 0 and comprehending why the median isolates human errors will allow you to evaluate your agents much more fairly and improve your customer experience. ✅