Quantifying Performance

By Ric Kosiba, Chief Data Scientist at Sharpen Technologies

Wow, have things changed. I’m writing from my basement “office” near Annapolis, Maryland.  Since our last On Target, the world, and the contact center industry has changed, in some ways a lot, and in some ways not all that much.

At the start of the pandemic work-at-home order, we made an effort to chat with several of our customers about how work-at-home affected their phone agents and any management issues they were having.  We learned a few things.

First, the transition to work-at-home went surprisingly
well, most switching to an at-home workforce in less than a day. We did an admittedly small survey, and here are the results:

This is great news—the transition seemed mostly painless.  When interviewed, most said that procuring hardware—laptops and solid WIFI—was the biggest impediment.  This is important news for our industry. Work-at-home has been more than doable, and companies that have shied away from at-home agents now have positive experiences with it.

Truth number 1: Four months ago, our industry often looked at work-at-home as an outlier. The great news is that we were able to innovate so our operations found the transition to an at-home operation to be not so tough.

Technology has made the transition easier and reporting and metrics have made management achievable. More on that in a bit.

The implication of this is that work-at-home is here to stay for many companies.  Why would you spend a huge capital cost for a physical building, when you could get away with a much smaller and cheaper administration building and a training center?  The financial math of a large contact center will be hard for many companies to justify, now that everyone knows that work-at-home isn’t that hard, and is for many of us, preferred.

Before chatting with our customers, we expected that work-at-home would provide distractions and that efficiency would favor the traditional call center… but it didn’t. Many—not all—of our contact center customers found that their agents were more productive, not less. Certainly, many agents preferred the work-at-home lifestyle.

Truth number 2: Companies today all know that they have many more options for configuring their contact centers.

We heard from one smart manager: “I’m missing the incidental information, glancing and noticing that an agent is away from their desk, or seeing an agent walk onto the floor in the morning.”  There is the intangible value of office work—hearing and seeing what is happening in the center.

We heard that management at work-at-home centers must be more proactive.  Management cannot rely on informal conversations; now every communication is on purpose. We also heard that simple tools can help.  One item, a performance dashboard, is particularly helpful.

Here is an example of a simple performance dashboard, visible on an agent’s desktop when they are not interacting with a customer. This serves a few purposes.  First, the choice of metrics, message, and goal, represents a management statement of what is important to the business. We don’t always think of our choices of metrics as a management statement, but they are. Our metrics are a regular reminder to our agents of management’s expectations of what good performance is.

Performance Dashboard
(metrics, message, and goal vary by staff group)

Second, it is a tool for agent self-management. In an environment where management is not physically visible or casually available, it is important to change the way people keep tabs on themselves.  Knowing expectations is important, but regularly noting whether you are meeting those expectations is as important, especially given the change in the working environment.

Finally, viewing your peer’s performance and your standing is a competitive motivator. Knowing that your performance is above-board is good; knowing your performance ranking compared to your peers will tell you at a glance what it takes to excel.

If social media has any lessons to the contact center it is this: everybody wants to know that they are doing well.  A happy agent is one who knows they are performing, and knows that others know it too.

Identifying Coaching Opportunities

I wanted to draw a couple of pictures that I believe are pretty useful.  First is a trick to determine who to coach and what the benefit of said coaching is. It is important that whatever agents see, management has a view into it as well.  The table below is a simple view:  a sortable table showing ranking of performance, on all metrics that are important to management. When an agent is coached by her supervisor, the supervisor should have a view into her performance, on the metrics that the operation views as critical.

Table: User group performance metrics

How do we make mathematical sense of this information? Let’s take a look at a simple graph of a stacked ranking of performance, where the X-axis plots agents, sorted from best to worst, and the Y-axis plots their performance on a single management measure.  The performance window can be a day to several months.  These metrics are really whatever you and your leadership thinks is important.

The simple act of sorting performance and viewing it on a graph can show some interesting features.  The shape of the curve will automatically show you your performance outliers.  Those agents falling on the very left side of the graph are your performance stars, and those falling on the extreme right side of the graph are your list of agents needing to improve.  If you overlay an “acceptable performance line” onto the graph, every agent whose performance is above that line is on the targeted training or coaching list.

There are some tricks in the math: the area under the curve represents the total value you would gain by improving performance, either via training, coaching, or reassignment.  The median arc on this “triangle” represents the average expected metric improvement possible if those agents scored “acceptably.”

One would expect this curve to have a very large, middle flat section, where your average performance resides.  This is normal.  The tail end of that curve, if it is sharply higher, means that there are a number of agents whose performance can be improved. If the end is only slightly above this middle, it means there is not much variance from mean performance.  It may be likely that agents cannot affect performance much at all (especially if the rank ordering changes often).

Another business question you can ask of your data is what is vexing your agents (and your customers)?  I recently was helping my mom change her compromised email password.  Because of the way her password was stolen, I had to call the company and speak to an agent to fix her account.  It was an interesting process, and in the end, I spoke with seven different agents, and had three live chats. I wondered if senior management knows that their agents cannot easily fix this simple problem? Does anybody know that I contacted them ten times?

Here is a simple graph that can help you determine which questions your agents cannot answer. Plot the distribution of callbacks, from the same phone number, over some short-ish time horizon (say 1 or 3 days).

In the callback distribution graph, you can see that there is a significant number of phone calls that required contacting the center more than two times. For many types of operations, like a mortgage broker, that’s expected. For many others (and my mom’s cable company), it isn’t. These contacts represent service failures, and simply listening to those specific contacts would yield a common problem that the phone agents don’t know how to service. Note that sophisticated solutions, like call analytics might not find these sorts of problems, since it was the callback, not the conversation, that triggered the training issue.

There are a few sources of ROI associated with reducing the number of callbacks.  First, there’s the simple cost of having to spend more agent hours answering and re-discussing an issue that should have been handled earlier. Second, there’s the loss of customer satisfaction. Every time I had to call my mother’s cable company—not a favorite thing for me to do—was an exercise in my losing confidence in that company.  Also, there’s the hidden cost of agent satisfaction.  The more times an agent has to work the same problem over—with possibly irate customers—the less satisfying the job.

Whether our work environment requires work-at-home agents or not, we know these truths will hold: our industry has many more options for configuring our work than we did before the current pandemic.

We also know that managing a distributed workforce isn’t the same as managing an on-site workforce.  It requires us to use our data in interesting ways.  By publishing our expectations, understanding our team performance, showing our agents their scores, analyzing that performance, and searching for our gaps in training, we can effectively manage our teams, even from our basement office.

Ric Kosiba, Ph.D. is a charter member of SWPP and is the Chief Data Scientist at Sharpen Technologies. He can be reached at rkosiba@sharpencx.com  or (410) 562-1217.

Wow, have things changed. I’m writing from my basement “office” near Annapolis, Maryland.  Since our last On Target, the world, and the contact center industry has changed, in some ways a lot, and in some ways not all that much.

At the start of the pandemic work-at-home order, we made an effort to chat with several of our customers about how work-at-home affected their phone agents and any management issues they were having.  We learned a few things.

First, the transition to work-at-home went surprisingly
well, most switching to an at-home workforce in less than a day. We did an admittedly small survey, and here are the results:

This is great news—the transition seemed mostly painless.  When interviewed, most said that procuring hardware—laptops and solid WIFI—was the biggest impediment.  This is important news for our industry. Work-at-home has been more than doable, and companies that have shied away from at-home agents now have positive experiences with it.

Truth number 1: Four months ago, our industry often looked at work-at-home as an outlier. The great news is that we were able to innovate so our operations found the transition to an at-home operation to be not so tough.

Technology has made the transition easier and reporting and metrics have made management achievable. More on that in a bit.

The implication of this is that work-at-home is here to stay for many companies.  Why would you spend a huge capital cost for a physical building, when you could get away with a much smaller and cheaper administration building and a training center?  The financial math of a large contact center will be hard for many companies to justify, now that everyone knows that work-at-home isn’t that hard, and is for many of us, preferred.

Before chatting with our customers, we expected that work-at-home would provide distractions and that efficiency would favor the traditional call center… but it didn’t. Many—not all—of our contact center customers found that their agents were more productive, not less. Certainly, many agents preferred the work-at-home lifestyle.

Truth number 2: Companies today all know that they have many more options for configuring their contact centers.

We heard from one smart manager: “I’m missing the incidental information, glancing and noticing that an agent is away from their desk, or seeing an agent walk onto the floor in the morning.”  There is the intangible value of office work—hearing and seeing what is happening in the center.

We heard that management at work-at-home centers must be more proactive.  Management cannot rely on informal conversations; now every communication is on purpose. We also heard that simple tools can help.  One item, a performance dashboard, is particularly helpful.

Here is an example of a simple performance dashboard, visible on an agent’s desktop when they are not interacting with a customer. This serves a few purposes.  First, the choice of metrics, message, and goal, represents a management statement of what is important to the business. We don’t always think of our choices of metrics as a management statement, but they are. Our metrics are a regular reminder to our agents of management’s expectations of what good performance is.

Performance Dashboard
(metrics, message, and goal vary by staff group)

Second, it is a tool for agent self-management. In an environment where management is not physically visible or casually available, it is important to change the way people keep tabs on themselves.  Knowing expectations is important, but regularly noting whether you are meeting those expectations is as important, especially given the change in the working environment.

Finally, viewing your peer’s performance and your standing is a competitive motivator. Knowing that your performance is above-board is good; knowing your performance ranking compared to your peers will tell you at a glance what it takes to excel.

If social media has any lessons to the contact center it is this: everybody wants to know that they are doing well.  A happy agent is one who knows they are performing, and knows that others know it too.

Identifying Coaching Opportunities

I wanted to draw a couple of pictures that I believe are pretty useful.  First is a trick to determine who to coach and what the benefit of said coaching is. It is important that whatever agents see, management has a view into it as well.  The table below is a simple view:  a sortable table showing ranking of performance, on all metrics that are important to management. When an agent is coached by her supervisor, the supervisor should have a view into her performance, on the metrics that the operation views as critical.

Table: User group performance metrics

How do we make mathematical sense of this information? Let’s take a look at a simple graph of a stacked ranking of performance, where the X-axis plots agents, sorted from best to worst, and the Y-axis plots their performance on a single management measure.  The performance window can be a day to several months.  These metrics are really whatever you and your leadership thinks is important.

The simple act of sorting performance and viewing it on a graph can show some interesting features.  The shape of the curve will automatically show you your performance outliers.  Those agents falling on the very left side of the graph are your performance stars, and those falling on the extreme right side of the graph are your list of agents needing to improve.  If you overlay an “acceptable performance line” onto the graph, every agent whose performance is above that line is on the targeted training or coaching list.

There are some tricks in the math: the area under the curve represents the total value you would gain by improving performance, either via training, coaching, or reassignment.  The median arc on this “triangle” represents the average expected metric improvement possible if those agents scored “acceptably.”

One would expect this curve to have a very large, middle flat section, where your average performance resides.  This is normal.  The tail end of that curve, if it is sharply higher, means that there are a number of agents whose performance can be improved. If the end is only slightly above this middle, it means there is not much variance from mean performance.  It may be likely that agents cannot affect performance much at all (especially if the rank ordering changes often).

Another business question you can ask of your data is what is vexing your agents (and your customers)?  I recently was helping my mom change her compromised email password.  Because of the way her password was stolen, I had to call the company and speak to an agent to fix her account.  It was an interesting process, and in the end, I spoke with seven different agents, and had three live chats. I wondered if senior management knows that their agents cannot easily fix this simple problem? Does anybody know that I contacted them ten times?

Here is a simple graph that can help you determine which questions your agents cannot answer. Plot the distribution of callbacks, from the same phone number, over some short-ish time horizon (say 1 or 3 days).

In the callback distribution graph, you can see that there is a significant number of phone calls that required contacting the center more than two times. For many types of operations, like a mortgage broker, that’s expected. For many others (and my mom’s cable company), it isn’t. These contacts represent service failures, and simply listening to those specific contacts would yield a common problem that the phone agents don’t know how to service. Note that sophisticated solutions, like call analytics might not find these sorts of problems, since it was the callback, not the conversation, that triggered the training issue.

There are a few sources of ROI associated with reducing the number of callbacks.  First, there’s the simple cost of having to spend more agent hours answering and re-discussing an issue that should have been handled earlier. Second, there’s the loss of customer satisfaction. Every time I had to call my mother’s cable company—not a favorite thing for me to do—was an exercise in my losing confidence in that company.  Also, there’s the hidden cost of agent satisfaction.  The more times an agent has to work the same problem over—with possibly irate customers—the less satisfying the job.

Whether our work environment requires work-at-home agents or not, we know these truths will hold: our industry has many more options for configuring our work than we did before the current pandemic.

We also know that managing a distributed workforce isn’t the same as managing an on-site workforce.  It requires us to use our data in interesting ways.  By publishing our expectations, understanding our team performance, showing our agents their scores, analyzing that performance, and searching for our gaps in training, we can effectively manage our teams, even from our basement office.

Ric Kosiba, Ph.D. is a charter member of SWPP and is the Chief Data Scientist at Sharpen Technologies. He can be reached at rkosiba@sharpencx.com  or (410) 562-1217.