Engine Works
Under the hood of Alteryx: tips, tricks and how-tos.
Today Alteryx announced that Alteryx Analytics Cloud (AAC) is now available on Databricks Partner Connect, along with other new cloud capabilities that make it even easier to use Databricks with AAC. The net-net: You get set up with Databricks faster.
Let’s walk through a fun customer churn example to see how the new Databricks connectivity helps you achieve a use case from start to finish.
The use case: Analyze gym membership churn
We're just a couple of months into 2024, which means our New Year’s resolutions are hopefully still alive! But perhaps not for long. Gyms typically see a membership boost in January – but how many of those new members will stick around?
"Quitter’s Day” is officially January 12, which is a real bummer. It only takes 12 days for most of us to abandon our fitness goals?
Not if data-driven gyms have something to say about it!
In this scenario, I’m the head of memberships at a gym chain and I want to combat customer churn before it happens. To spot the members likely to cancel and identify proactive actions to take, I’m going to explore, combine and prepare data from my gym’s Databricks environment.
Here’s our order of ops:
- The data engineer builds the data pipeline, landing the raw data into Databricks. At this point, the data is in a raw state, which we’ll call bronze. It still needs to get cleaned and prepped before it’s usable.
- Then you, the Alteryx user, can actually get those datasets from within the Alteryx Analytics Cloud interface.
- Once you have the data, you can work with it in Designer Cloud. And you can also use Databricks as your execution engine.
- Finally, you can use that dataset to uncover insights in Auto Insights. And now that your dataset is in Unity Catalog, someone in Databricks who might be more technical, like building a machine learning model, can use that same dataset that you so nicely prepared for them.
Launch the Alteryx Analytics Cloud from Databricks Partner Connect
If you’re a Databricks user but not yet an AAC user, you can now launch a free trial directly from the Databricks interface. Your workspace admin will need to do this part. Navigate to Partner Connect, click on the Designer Cloud tile and fill out the form.
Once the trial launches, your organization’s existing Databricks connection will be ready to go in AAC. Just like that!
Access the relevant datasets from AAC, using Databricks Unity Catalog
Now that I have my Databricks connection set up, I can quickly get to the data.
A more technical role like data engineering has probably already landed the data in Databricks. Databricks Unity Catalog allows users to do their own data discovery and identify useful datasets.
Here is what the gym membership datasets look like when you view them in the Databricks UI:
...but thanks to Alteryx support for Unity Catalog, if I’m in a business role and don’t want to mess around in the Databricks UI, I don’t actually need to go into Databricks to find those datasets!
I can look up those same datasets from AAC in the Data tab. I’ll click on Import Data, select my Databricks connection, and find the gym datasets.
Prep and explore the data in Designer Cloud
I’ll open a new workflow in Designer Cloud and get to work cleansing and prepping those datasets. We’re looking at membership records, gym usage and check-in data from the front desk, and survey data from an exit survey we send out to members who cancel.
Survey data is notoriously messy, especially if the survey isn’t well-designed. One question in my gym’s exit survey had a free text box, so people typed in their own “Yes” or “No” responses...causing lots of trailing spaces! I’ll fix those and do some other prepping, such as updating the gym location names with Find/Replace.
Once I join the datasets together and I’m happy with the quality of the data, I can do some customer segmentation to personalize marketing offers.
For instance, I applied a filter to separate out customers who are employed and visited the gym in the afternoon – they’re likely to be interested in information about the gym’s new co-working area, so it’s easier to exercise during the workday.
I’ll also filter for customers who canceled due to not using the gym enough and listed “stress relief” as their health goal. Now I can target them with an email about our new yoga and meditation class schedule and see if that’ll get them back and using the gym more.
Execute the workflow with the Databricks engine
Time to output the final dataset! You can run the job in the Databricks engine, which is especially useful when you’re working with large datasets or want to minimize data movement. Databricks is designed to process ungodly amounts of data.
I’ll drop in an output tool and configure it to output to my Databricks connection. When I Run Job, I’ll make sure the Databricks option is selected.
You can see the job details in both AAC, and in Databricks. That means both the Alteryx user (like a business analyst) and a Databricks user (such as the data engineer or IT admin) can see the same job results. Everyone gets visibility without leaving their preferred UI. A win-win all around.
Another bonus: With your final dataset available in Databricks Unity Catalog, another member of the team – let’s say a more technical data scientist – can now easily find and use that dataset as well.
Use the final dataset for analysis in Alteryx Auto Insights
Now that we have our polished, high-quality dataset, we’re ready to use AI on it. To be specific, we’re going to have Auto Insights read the dataset and tell us what to look into. Which factors make a member likely to cancel? What proactive action can I take to prevent cancellations?
Once I build my Auto Insights mission, I can find some answers. Here are a few insights I discovered:
- In the second half of the year, people canceled because they found a better gym. Now, I know this is just a hypothetical gym, but still. OUCH. My fake gym members like another fake gym better?!
- Gym visit duration – the amount of time a member spent per visit – saw a drop-off during the spring and summer months, then picked back up in the fall. Also, members who did CrossFit had longer gym visits, which tracks.
- The health goal “stress relief” increased at the end of the year and drove longer gym visits. The busy holiday season is prime time for getting into meditation and yoga.
These findings are useful for my gym’s marketing team, so I’m going to use Magic Documents to send them an email about it. (in the customizable settings, I asked for a “Witty” tone). Every problem is a dumbbell waiting to be lifted!
By the way, I was inspired by a real-life example for this use case: Peloton noticed that people tended to cancel their subscriptions during the summer months – when indoor cycling loses its appeal – so it adjusted subscriptions to offer a free tier of equipment-free classes.
Also, I used Auto Insights Playbooks to generate the hypothetical data for this example. Check it out. It comes in clutch for experimenting with use cases.
Even though this was a gym example, Alteryx users at all kinds of businesses will need to analyze churn. With the deeper Databricks integrations, AAC gives you a path to investigating your data and getting to the answers more quickly.
If you’re ready to use Alteryx with your organization’s Databricks environment, reach out to your workspace admin to launch a trial from Databricks Partner Connect!
To see the new Databricks connectivity with Alteryx Analytics Cloud in action, save your seat for our upcoming DIY Data episode, where we'll walk through this very scenario step by step. Register today!
Resources:
