Ten Paths (And Then Some) You Should Be Exploring

Originally published at https://aster-community.teradata.com/community/bso/aster-guided-analytics/blog/2016/11/15/ten-paths-and-then-some-you-should-be-exploring.

You have a ton of great event-based data but aren’t sure where to begin exploring. Path analysis is one of the most common uses of event data. Think “what happens before X” or “what happens after Y.”

As more enterprises adopt Aster’s Path Analysis Guided Analytics Interface for business-friendly journey analysis, we come across more and more paths that companies from a variety of industries should be exploring. Some of these paths are obvious. Others, we probably wish had been more obvious. And some are downright creative.

With that in mind, I want to share with you several paths that you should analyze and explore.

1. Paths to churn

If you work for a telecommunications company, insurance company, retailer or other business dependent on subscriptions or repeat purchases, this path is probably top of mind for you. Good news — the Path Analysis interface provides several visualizations that may help you understand your churn paths in ways that previously were not so intuitive.

2. Paths to cart abandonment

For retailers — and pretty much anyone who runs a website — this is also a very common path to try to understand. Through such an analysis, you may uncover issues in your online checkout process between the shopping cart and payment completed pages. Or you may realize you ask prospects to jump through too many hoops to buy anything.

3. Paths from sign-up

Think of this similar to the cart abandonment use case above. When a prospect applies to open an account, for example, what happens next? Which events are included on a path to a successful application? Are there paths that do not yield the desired result, perhaps due to no fault of the prospect? Insurers and financial services providers are particularly interested in these paths.

4. Multi-channel paths after opening an account

So your customer has completed the sign-up process, and you have approved their application. Now what happens? For a financial services company, what do these paths tell you about how easily customers can transfer funds from an external account to fund their checking account? Or perhaps you see that some customers immediately use your credit card at a retailer but then struggle to pay their first month’s bill online. Does some segment of customers forget about you at all despite prodding via email, text and from your customer support center?

A tree diagram shows common multi-channel paths after a customer completes a bank account application online.

5. Paths through a website starting at the home page

When a visitor hits your home page, what are the next two or three pages they hit? Are some visitors searching for information while others want to immediately purchase products? Do visitors explore several pages before contacting you directly?

6. Paths to purchase

What are the simplest paths for a customer to complete a purchase from you? Are there complex paths that you could simplify? Should you be allocating marketing dollars to drive customers toward specific paths? This analysis can yield some surprising insights into what customers must do to buy from us.

7. Paths to upgrade

This could be described, along with the previous example, as a subcategory of “paths to conversion” in many cases. The idea is pretty much the same, except you have an existing customer converting to a newer or “better” product rather than a new customer converting as in the previous example.

8. Paths to renewal

The events that lead to becoming a first-time customer and a repeat customer are often quite different. Don’t take the differences in these paths for granted.

Now let’s think a little bit outside the box. Chances are, the data for the following use cases exists somewhere within your organization. Maybe it isn’t being used for path analysis, but the opportunity to generate value by leveraging this data for path analysis is significant.

9. Paths including website and call center

Every time a customer reaches a customer support agent, you are paying money to solve a problem. On the low end, some of our customers estimate that the average cost of a support call is more than $6. This may sounds small by itself, but in aggregate it becomes tens or hundreds of millions of dollars each year. This path analysis may help you identify areas to improve your website and reduce those expensive calls significantly.

10. Paths from interactive voice response (IVR) to agent

Similar to our analysis above, are there simple questions your IVR could handle that would help reduce the number of calls that require a live customer support agent?

11. Paths to equipment failure

The Internet of Things / Analytics of Things provides us with immense amounts of data in near-real-time. Manufactures, automotive companies, airlines and many others love to use this data for early failure detection. While this is a great use case for IoT data (hopefully saving lives and making companies more efficient), there is plenty of value to be gained in understanding the paths that precede such failures. (Pro tip: In addition to path analysis, consider applying techniques such as Naïve Bayes and Support Vector Machine to predict failure of equipment and components.)

12. Paths through a physical store

Retailers have spent decades trying to design the perfect store layout. But they have never had access to the amount of location-based data now available. When you understand how customers move through your store, you can help customers more quickly find the items they seek, and you can leverage these paths to introduce them to new items they might have missed otherwise. (Pro tip: Look at collaborative filtering, pSALSA and other recommendation techniques alongside this path analysis.)

And to hammer home the point that opportunities for path analysis are all over the place, let’s look at this final example:

13. Paths to surgery

In the medical world, a patient’s history often reveals how they got where they are today and can contain clues about how they might respond to specific treatments such as surgery. Consider what you could learn by looking at those patient paths in aggregate. Are there opportunities to improve outcomes by proposing surgery earlier after specific diagnoses? Are there common paths where alternative forms of treatment are not adequately explored? How do paths differ for various patient populations? (Pro tip: Another popular path to explore in healthcare is paths to readmission.)

This sigma diagram shows which events are commonly associated with a heart transplant.

I’m sure I left many great use cases off the list. Feel free to comment below with your recommended use cases. If we generate enough examples, I’ll write a follow-up post.

And of course, if you’re interested in using Aster’s Path Analysis Guided Analytics Interface to better understand your paths of interest, please send me a note at ryan.garrett@teradata.com.