How to use Process Mining for intake channel analysis
This post is part of the Process Mining Use Case Series where we’ll focus on different techniques to identify process inefficiencies, non-conformant activities, and improvement opportunities. These posts will be broken into two sections – how to do the analysis and how to configure the Process Mining project.
Self-service channels such as portal or email allow quick customer or employee service. However, unlike phone, walk-up or a virtual agent experience it is mostly a one-way communicaton which makes intake more challenging. Not everyone knows how to describe their problem or what they need accurately or completely. Consequently, tickets created by self-service often have higher re-assignments, or more frequently require us to go back to the requestor for additional information.
These situations where additional information is required are frustrating for customer/employee and have a negative impact on satisfaction scores as well as service team productivity.
Process Mining can be used to help analyze work coming in view different channels and isolate opportunities to improve intake experiences as well as create new self service offerings which will increase the speed and productivity of the organization.
You can find other Process Optimization use cases here
How to do the analysis
This approach can be applied to any workflow(Incident, HR Case, Customer Service Case, etc) and does not require any additional/special configuration beyond the out of the box content packs offered with a given workflow
Open your project to the Analyst Workbench view
One of the first states to look for when you are analyzing a workflow and looking for opportunities to optimize is the "Awaiting Caller Info” state. Depending on your process or your workflow the state might be labeled differently, it could be “On Hold” or “Suspended”. Regardless of what is called, any time we need to pause the work to gather some additional information is an opportunity to improve and area we should focus on. (Note you may have to expand your connections within Model Options to see this state on your own visualized process maps).
In this case we have about 9600 incidents that have passed through this “Awaiting User Info” state and with 12600 total occurrences that is telling us that about 3000 incidents visited it more than once. That is definitely something we’d want to focus in on.
To isolate just those 9600 incidents that at some point went into “Awaiting Caller Info” we can use the Filter on transition option within the Awaiting Caller Info node.
We can decide to further focus and isolate incidents that went into awaiting caller info more than once, in this case, 2-9 times.
Within the Transitions filter interface, we can also build out specific paths we want to isolate. For example, if we wanted to focus in on things that went to Awaiting Caller Info to directly from New. This will allow a team to easily identify and investigate incidents that may not have the needed information to proceed.
Today, we are just going to focus on all incidents that went to "Awaiting Caller Info" once.
Once we click apply Filter on Transition within the Awaiting Caller Info node, the visualized process map will filter down to just show us the 9600 incidents that went through Awaiting Caller Info.
Now that we’ve narrowed the incidents down to the ones requiring additional information, we can focus on which channels are the biggest influencers.
We’ll use our breakdowns to do this.
Looking at our different channels, we can see that Self-Service, Email and Portal are the top three contributors from a volume perspective. But the thing that jumps out is that incidents coming in via Email are taking an additional 5 days on average longer than those other two options. Email being a very unstructured channel it would make sense that there is going to be some back and forth, but that doesn’t mean we shouldn’t see if we can improve in some way. Let’s focus on those first. We can select Email and click Apply.
Once we’ve isolated the data down to just the incidents coming in via Email that are going to Awaiting Caller Info we will now use our Requestor Intent - Cluster Analysis capability to analyze the short descriptions and descriptions of those incidents to see if there were any patterns in that unstructured data that would give us some additional direction
The Cluster Analysis will show us that we have a few different opportunities.
There is also a cluster of incidents where people are having issues with access to multiple applications and single sign on. 2. That is certainly something we should have a self-service option for (could be a virtual agent that leverages AI Agents) conversation. If we do have an existing option let’s look at how we can improve the intake experience or dive into which parts of the organization are not using the more structured intake channels for this and market the option a little bit more.
This same steps can be used to analyze other intake channels to look for improvements to the experience or potentially new self service options.
How to configure the Process Mining Project
This analysis should not require any special configuration. The out of the box project for each of the workflows will enable you to do this type of analysis.
If you have not created a Process Mining project before this short video will walk you through the process in minutes.
The one aspect of the configuration you’d want to confirm is that you have a Breakdown Definition configured for contact_type or channel and that it aligns to the field your organization uses to capture this information.
If you are looking for more in-depth training you can use the Process Mining Academy library of content
You can find other Process Mining use cases here
Additional Resources
Process Mining Fundamentals - ServiceNow University
