Analyzing the Customer Perspective

It happens often. We focus so much on improving processes that we forget why we’re improving them in the first place. If our first goal is to serve our customers, and our processes were designed to support those customers, then shouldn’t the process data we analyze also account for that customer perspective? I’m not just talking about analyzing customer surveys, but ensuring our metrics are customer-centric, i.e., they reflect the customer’s behavior. Here’s an example that yielded unexpected insights.

Some background info about the analysis

This study was done for a large corporation that has thousands of store locations across the United States. If a customer called one of these locations and the store was closed or no one was available to answer, then their call was forwarded to a centralized call center. These call center agents often operated like an answering service because they weren’t as skilled to answer the customer’s issues like the employees at the local store. I wanted to analyze these forwarded calls to assess what was the customer demand for each store’s hours of operations and therefore understand how the stores were able to meet that customer demand.

How was the data analyzed?

To analyze this, I pulled a sample of those customer calls over a period of several months and plotted the average % of weekly volume by hour of day for each day of the week. Using this approach I can see how the volume comes across the week from day to day and how it comes in for each hour of each day. 

For example, if they averaged 100K calls per week, then I looked at how those 100K calls came in each day of the week and each hour of those days. So if an average of 10K of those 100K calls were on Monday, then the whole week of calls add up to 100% and Monday’s call portion adds up to 10%, and that 10% portion for Monday is sub-divided by each hour in Monday. The chart below illustrates how the volume of these calls flowed by day and by hour throughout each week:

From this chart we can conclude several interesting things:

• Most of the weekly call volume occurs on Mondays.
• The call volume slightly declines through the remaining weekdays and significantly declines on weekends.
• The call volume tends to peak most days around the 11 a.m. or 12 p.m. hours.
• The majority of call volume for each day follow a similar pattern of occurring between 9 a.m. and 3 p.m. (illustrated by the shaded area in the chart).

Some wrong assumptions

The similar hump-shaped pattern for each day’s call volume looks like a bell curve typically associated with a normal distribution. However don’t be fooled by that, it’s merely coincidental. This data is not meant to reflect a distribution like in a histogram.

Since most call volume occurs during the mid-part of each day especially near lunch time, we could deduce that customers are calling when it’s most convenient for them, such as during their lunch break. However upon further analysis below, this is also a wrong assumption.

What I learned in analyzing this data is that the timestamps recorded for the calls are based on the location for the corporation’s headquarters which resides in the Central time zone of the U.S. Since the corporation uses multiple call centers located across and outside of the U.S., it’s very helpful from an operational standpoint to synchronize the data to a single time zone. By doing so, we can know how to staff the call center to support when customers are calling. But how does that affect the customer’s perspective?

What is the customer’s perspective?

Customers don’t care where the corporation is headquartered nor where the call centers are located. Customers call when it’s important to them, so it’s our job to ensure we understand when it’s important to those customers so we can meet their demand. To do that, we need to account for the time zone when the customer called. 

To do this, I pulled the predominate time zone for each state in the U.S. (there are websites that easily post this data) and I added or subtracted an hour for each time zone’s distance from the Central time zone. For example, calls in the Eastern time zone had 1 hour added, calls in the Pacific time zone had 2 hours subtracted, and so on. 

Since some states have more than one time zone, I used the time zone that covers the largest portion of the state (or at least where it's most populated). Although I could've pulled the time zone for the respective zip code or area code, it would’ve required more time and work than necessary. In addition, the benefit of doing statistical analysis means we don’t need perfect or complete data – we only need enough data to draw valid statistical conclusions that influence meaningful business decisions.

What were the results?

By accounting for the customer’s time zone, it revealed a significant shift in the call volume by hour as shown in the chart below:

From this chart we can conclude a few interesting things:

• The call volume by day and the decline through the week is the same as before (I didn’t expect that to change).
• The call volume sharply peaks most days at 9 a.m. when most stores open.
• The majority of call volume for each day falls into a narrower window between 9 a.m. and 12 p.m. (illustrated by the shaded area in the chart).
• Although Sunday has the lowest call volume, most stores don’t open on Sunday until 12 p.m. or 1 p.m., so the call pattern peaks around that timeframe.

The biggest surprise from this customer perspective is how these customers want to contact a local store as soon as it opened, no matter what day of the week it was. Since this call volume represents the calls that rolled over to the centralized call center, it further means those stores weren’t able to meet the customer demand in those earlier hours and were forwarded to a phone agent who is less equipped to address the localized needs for those customers. 

In other words, rather than the call center focusing on improving their average handle time (AHT) or issue resolution (IR) metrics, what’s in the best interest of the customer is to not have them routed to the call center in the first place. So we needed to find ways the stores could meet that customer demand like opening each store earlier, or ensuring they’re appropriately staffed in that early part of the day, or instead of forwarding the call to a call center they could forward the call to another local store. 

Final Conclusion

Many organizations attempt to get the customer’s perspective by sending them surveys. However this kind of insight on customer behavior and demand is rarely discovered from a customer survey. Surveys are often designed around what we think the customer cares about, rather than letting the customer’s behavior reveal what they truly care about. 

Although it’s not always possible to measure customer behaviors, we need to creatively explore ways we could analyze our data to account for that customer perspective. And when we do, we may find valuable customer-centric insights whereby we could anticipate and meet our customer’s needs before they even realize they have a need – now that is delighting customers!

About Matt and StatStuff

Matt is a Lean Six Sigma (LSS) Master Black Belt (MBB) who's been leading continuous improvement teams for over 20 years across several industries and across many departments within several large corporations. Matt developed and led several enterprise-wide LSS training and certification programs and founded StatStuff as the only FREE source for complete LSS training. StatStuff is highly endorsed as quality LSS training from leaders at top companies like Apple, eBay, Pepsico, Bank of America, Dell, Sprint, BP, Honeywell, etc. Many companies, training organizations, and universities use StatStuff for their training curriculum.