Path analysis with SAS Visual Analytics

Introduction

Understanding the behavior of your customers is key to improving and maintaining revenue streams. It is a an important part when crafting successful marketing campaigns. With SAS Visual Analytics 7.1 you can analyze, explore and visualize user behavior, click paths and other event-based scenarios. Monitoring the customer journey by visualizing all touch-points in your organisation will help you to identify gaps and improve the overall customer experience. Flow visualizations will help you to best understand hotspots, highlight common trends and find insights in individual user or aggregated paths.

In path analysis you are typically trying to determine a sequence of events in a particular time window. For example you pay attention to paths more frequently used than others in order to understand what path prospects take before they become new customers. Path analysis works best with linear event streams such as customer life cycle (1. prospect, 2. trial subscription, 3. customer, 4. product upgrade, etc.) but is also commonly used for web usage analysis. As a data scientist you may look for optimal paths to compare with paths customers have actual taken. This often reveals interesting insight and opportunities for revenue improvements.

Challenges

Path analysis can be challenging especially when analyzing web usage. There are often many ways customers can navigate on a website, so even if we determine the optimal path in this scenario, it’s very likely that just a very minor number of users will actually take the optimal path. This means you must pay special attention to the path analysis results in order to gain the right insights. You may for example compare the least and most used paths in terms of sequence count or the number of drop off’s (e.g. customers who left a session and therefore didn't complete an order).

It can be useful to apply segmentation to path analysis (more details below), as this will greatly reduce the number of steps in a path and may represent a better aggregated view about paths taken. In most cases, you are after the number of people taking the optimal path to reach your goal such as purchasing a product. Once you are understanding common paths you can try to influence the customer behavior by redesigning the web page or starting a marketing campaign, for example.

Path analysis in SAS Visual Analytics

Let’s start with a very basic example about path analysis to explain the basic steps. Consider the following simple data set:

vae_path_analysis_00

The table structure shows our customers (John, Jane and Bob) and the visited web pages (item column) per session (transId column). As you can see the customer “John” visited our web page twice at different times. The sequence column is just used to maintain the order of the events. Typically you would take a date/timestamp here.

Since this is very simple data set you can easily see what paths each customer has taken:

  • John: ABC, ADE
  • Jane: BDEED
  • Bob: AFD

Visualizing this example in SAS Visual Analytics provides the following Sankey diagram:

vae_path_analysis_01

The diagram is colored by path indicating that there are 5x different paths including a drop off (path 2, red). This already gives interesting insights such as partly shared paths (John/green and Jane/turquoise share event D & E) as well as a common start event (A).

As part of the path analysis in SAS Visual Analytics you can also change the link aggregation and colorization. Switching the aggregation to color links by event shows the following:

vae_path_analysis_02

Again highlighting the common partial path in yellow. By default the diagram uses the sequence count or frequency as default link width. However, you may want to weigh paths by a given measure, such as purchases or revenue as this better reflects the impact a path may have. The following example shows a currency measure assigned as path weight:

vae_path_analysis_03

Let’s look at a more advanced data set with a few more events to analyze. Note, that this data set is just a small extract of a real web site access log file. You will see how quickly the number of paths increases and things like ranking and segmentation will play an important role:

vae_path_analysis_04

Similar to the first data set we are looking at customers visited specific pages on our website over a period of time. The increased number of potential pages or events also mean an increased number of potential paths a customer can take. Let’s look at a first visualization of this data source:

vae_path_analysis_05

Not surprisingly most customers enter our web page via the welcome page. This could be mainly driven by the fact that users typically click on the first link in search engines rather than one of the sub categories. As you can see the paths taken are very long making this diagram very wide – tools such as the overview panel or path selection help navigating in the diagram:

vae_path_analysis_06

Segmentation for path analysis

One of the methods to reduce the overall number of events is to group events. SAS Visual Analytics provides methods to create custom categories. In our example we are going to group a number of events into groups such Buy, Search and Product:

vae_path_analysis_07

Applying this new custom group item to the Sankey diagram provides an aggregated and simplified view of paths taken:

vae_path_analysis_08

Once you have determined a particular path of interest you often want to further analyze the related group of customers having taken this path. For instance to include the group of individuals in your next marketing campaign. SAS Visual Analytics allows you to narrow down the selection by either filtering or merging into a new visualization.

Path filtering

Path filtering is done by selecting one or more events and either include or exclude the items by various conditions:

vae_path_analysis_09

In this example we are only interested in paths starting with the welcome page:

vae_path_analysis_10

Note, that the user can go ahead use the current filtered selection to create new visualization for further analysis.

Ranking paths

Given the high number of potential paths a custom can take you may also just concentrate on the top or bottom ranked paths. SAS Visual Analytics provides a number of options to filter and rank paths shown in the following property panel:

vae_path_analysis_11

Given the new top 5 ranking settings and the selected vertical layout the diagram renders as follow giving you great understanding of the flow users take in the 5 most common paths.

vae_path_analysis_12

Conclusion

SAS Visual Analytics provides a robust platform for analytic discovery on your data. Path analysis is important when trying to understand your customers' behavior online. From basic web usage to campaign and attribute analysis, gaining insights from your data will help drive your next email campaign or paid search or banner ad. Even in customer life cycle monitoring as part of acquisition and retention analysis you can quickly see how customer touch points such as email offers, call center sessions or branch visits pay out.

tags: customer lifecycle, data visualization, path analysis, Visual Analytics

2 Comments

  1. Posted August 19, 2014 at 11:32 am | Permalink

    Some readers might find this article by searching for "path analysis in SAS." The name "path analysis" also is used in connection to structural equations modeling, which in SAS is carried out by the CALIS procedure in SAS/STAT. PROC CALIS has the ability to construct path diagrams as described in Yung (2014), "Creating Path Diagrams That Impress: A New Graphical Capability of the CALIS Procedure,"

  2. Michael Pawlak
    Posted September 2, 2014 at 2:29 pm | Permalink

    Great information Falko. Nice to see SAS Visual Analytics can support path analysis. Another area where this type of analysis is highly valuable is in IVR/VRU usage. IVR (Intelligent Voice Response) and VRU (Voice Response Unit) are the service organizations transactional systems that most customers despise, as they are the systems that a caller first hears (press option 1 for XX department, etc.). The challenge most service organizations struggle with is understanding what options customers select in the IVR/VRU and why. Service organizations for years have tried to understand what options are taken, in what order, and what customer behavior is driving the option selection and the option selection order. With path analysis in SAS Visual Analytics, organizations can finally learn the "why" behind customer behaviors in these systems, and develop more logical business rules to more direct and easier to use experiences for customers. This will also have significant impact on overall satisfaction levels, as a customer's experience in the IVR/VRU system is their first impression of an organization. As we all have probably know very well, these experiences are not pleasant. The key to understanding the "why" behind customer behavior in the IVR/VRU, and the "what" to change in the IVR/VRU business rules engine, is SAS Visual Analytics.

Post a Comment

Your email is never published nor shared. Required fields are marked *

*
*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <p> <pre lang="" line="" escaped=""> <q cite=""> <strike> <strong>