The Business Value of Clustered Customer Journeys

Nearly every company is actively looking at ways to improve satisfaction and reduce costs by enabling customers in web and mobile channels rather than one-on-one agent phone calls. On the surface, these journeys appear easy to assemble. But you do have options that can affect your analytics and business decisions: static merge, rolling merge, and clustered merge.

The Basics of Digital Containment

Digital containment is one of the mainstay use cases of customer journey analytics. In its simplest form, the digital containment use case requires just two data sources: the digital channel (web or mobile) and agent phone calls. When the customer has a web interaction without a related agent call, we consider that web interaction to have been "contained."

For many businesses, the website's containment rate often becomes a key performance indicator (KPI). We can easily generate a daily dashboard displaying the percentage of web visits that did not have a corresponding phone call. The business can track this regularly and discuss initiatives to improve the measure.

But which agent calls are related to the web visit? To keep things simple, we'll define this as a simple time delay, say a 24-hour window. Any interactions by the same customer inside of a 24-hour window are assumed to be related to one another. Events that occur outside the 24-hour window are not.

Static Merge: Fixed Dates with a Fixed Window

Our data-savvy audience is now imagining two SQL tables: one for the web visits and another for the agent calls. If we select Monday's web visits and Monday's phone calls, we can join these by customer identifier and easily calculate the web containment rate.

This is a nice, easy method. The logic is simple and easy to execute. However, it does have one major flaw: Any interactions separated by a midnight boundary will not be associated with one another.

For example, consider customer Ms. Smith searching the web site for payment options Monday at 10pm who then makes a phone call 4 hours later. If our fixed date range is Monday, we won't see the phone call from Tuesday. The web interaction clearly has a related phone call, but the call falls outside our fixed date range. 

Even if we extend our analytic date range, examine 7 days rather than 1 day, the error rate will drop but may never entirely disappear. How bad is the error rate? That depends upon how interactive your international customers are at your dataset's midnight boundary. 

Rolling Merge: Flexible Dates with a Fixed Window

SQL experts in the audience are excitedly solving this problem: For a 24-hour window, use a date range 2x larger and include the window in the join criteria. Okay… what?

Let's start with Monday's web visits. Next to this, let's also get phone calls from both Monday and Tuesday. As before, join the datasets by customer identifier, but only if the two interactions are within 24-hours of each other. Ms. Smith's web interaction at 10pm Monday can easily be joined to the phone call at 2am Tuesday, giving a more accurate containment rate for Monday's web interactions. 

Problem solved! Right?

Let's consider another customer. We find that Mr. Jones had a web interaction at 9am Monday, another web interaction at 11am Monday, and finally an agent call at 10am Tuesday. What is Mr. Jones's web containment rate? 

The first web visit has no associated phone calls within 24 hours, but the second web visit clearly does. But these web visits are only 2 hours apart! If web and phone interactions within the window are related, then surely two web interactions in the same window are also related. Shouldn't Mr. Jones's first web visit be associated with the phone call by proxy? 

Clustered Merge: Full Flexibility

At ClickFox, we encourage our clients to leverage a "clustered merge": Let the customer's interactions cluster into related buckets. Starting with the first interaction, add new interactions to this cluster as long as they occur within 24 hours of the cluster. If we see a full 24-hour break in activity, the next interaction creates a new cluster. The cluster of interactions may end up being quite large, but every interaction in the cluster is within 24 hours of at least one other member of the cluster. 

So, Mr. Jones starts his interaction cluster at 9am Monday. His 11am Monday web visit is clearly within 24 hours, so it's added to the cluster. Similarly, the agent call on Tuesday is within 24 hours of the second web visit and also gets added to the cluster.

Great! Where's the containment rate? We can now examine each cluster for specific sequences. Any web interaction that was not followed - in that same cluster - by an agent call counts as a contained interaction.

In the end, any of these merge strategies can help the business make better decisions. The more capable and more accurate solutions obviously carry increased complexity - unless you have the right tools. Request a Fox demo to see examples of how ClickFox makes journey analytics easier for businesses around the world. 

REQUEST A DEMO

Written by Al Mays

Al Mays is the Chief Product Officer. In this role, Mays is responsible for the software strategy and development direction for ClickFox products. During his tenure at ClickFox, he has also held several management positions in information architecture and pre-sales solutions architecture. Prior to joining ClickFox, Mays spent seven years at Ticketmaster where he held various positions beginning as an IVR developer and became the Director of Voice Applications where he set strategy and lead development, infrastructure and managed day-to-day operations of their complex voice self-service platforms. He holds a B.S. in Computer Science and Mathematics.

Subscribe to the ClickFox Blog

  • Be the first to know about what's trending in Journey Science
  • Stay up to date with our latest Journey Analytics Products
  • Be a part of the Journey Science movement!