https://support.tethr.com/hc/en-us/sections/202103128-Feature-spotlight Feature spotlight

# Average Handle Time (AHT) analysis

Section only

## What is an AHT analysis?

Tethr’s Average Handle Time (AHT) Analysis is designed to help you understand how categories (key events) found on your interactions impact handle time or interaction length. Utilizing advanced statistical methods and our newest machine-learning capabilities, we can accurately measure historical data and predict how much time will be added to your interactions when specific categories are found, as well as the potential cost associated with those categories.

## Elements of an AHT analysis

1. List of categories included in the analysis
2. Total annualized cost of the category, calculated using monthly volume and the average cost per minute
3. Time added to an interaction each time the category is detected
4. Volume of the total category hits found in analysis
5. Summary of the impact of the selected category on handle time. Click on a category in the analysis to display its summary
6. View interactions where the selected category was detected

## Technical overview

The AHT Analysis involves partitioning the handle times for a collection of calls among selected variables or categories. For example, a collection of calls may be divided into groups with Agent Confusion and those without. The average handle time is then calculated for calls in each group. The difference in average handle time in our example represents the amount of handle time attributable, on average, to Agent Confusion. This difference does not measure the time Agent Confusion is expressed in isolation but includes the totality of any associated activities, such as looking up information or asking additional questions. An additional benefit is the identification of categories associated with a decrease in handle time. For example, suppose an agent's behavior, such as Probing Questions, leads to a faster resolution of the customer’s issue. In that case, the decreasing handle time is thus represented as a negative time value.

In cases where the selected categories occur in the same call, an additional step is needed to determine each category's individual contribution. We accomplish this by factoring; separating the cases where the two categories occur together from where they occur separately. Resulting is a specific value for each category that represents its unique contribution to the average handle time. The sum of these values is a Least Mean Squared LMS estimate of the average handle time for a set of calls.

A statistical test is applied to each value to ensure the accuracy of these estimates. This test determines the likelihood that the calculated time estimate for a given category represents its contribution to handling time and is not merely a product of random variability within the set of calls. Those time values which pass the statistical test are said to be “statistically significant” and may be considered reliable for estimating handle time.

## Measuring cost

Tethr can then calculate a cost impact using the estimated handle time for each category. Looking at our Agent Confusion example, if Agent Confusion was determined to add an average of 30 seconds to a call, and using an average cost of \$1.50 per minute, then occurrences of Agent Confusion would, on average, cost 0.5 minutes) * \$1.50 per minute), or \$0.75 per call.

We would then calculate the percentage of calls Agent Confusion occurred on in the last month. Let’s say you have 50,000 calls, and Agent Confusion occurred on 12,500 of those calls; we can estimate that Agent Confusion is occurring at a rate of 25%. Then using an estimated yearly call volume of 600,000 calls, we can calculate the yearly cost of Agent Confusion. With 600,000 calls per year at a 25% rate of occurrence, we can estimate 150,000 calls a year with Agent Confusion. We then multiply our estimated cost per call for Agent Confusion of \$0.75 per call by 150,000 calls for the year and get a total estimated cost of \$112,500 for the year.