This site uses different types of cookies, including analytics and functional cookies (its own and from other sites). To change your cookie settings or find out more, click here. If you continue browsing our website, you accept these cookies.
Read Alteryx customer stories to learn how they transform their organizations into becoming a data-driven business.
Announcing Alteryx + Snowflake | Alteryx and Snowflake make analytics and data science fundamentally easier. With the new integrated starter kit, you can push down data prep transformations and more into Snowflake for faster data quality and analytics output. Learn More
MindBody is an American software-as-a-service company headquartered in San Luis Obispo, California. Founded in 2001 the company provides cloud-based business management software for the wellness services industry. MindBody has been serving about 35 million consumers located across 330 countries and territories. Their Data Science and Analytics teams leverage Alteryx Designer and Python together to build a predictive risk prevention model and they use Alteryx Connect for all their Data governance requirements.
Describe the business challenge or problem you needed to solve
I began to realize MindBody had a few problems. We had different teams across the business that had similar missions but no one’s priorities aligned. We took all these teams and re-aligned them a different way so that all the different functions fell under one respective group. We now have a matrixed organization. We’ve also had a bizarre software model, we have one database per customer, and we have hundreds of thousands of databases. We spend 80 to 90% of our time every day just gathering the data, getting it structured, getting it cleaned.
With all that data collection we’ve found certain customer types carry higher churn and risk than others. Fortunately, a lot of that churn is preventable but there's churn that's completely unpreventable. If a business fails and they go out of business, we can't prevent that. The problem with these is that we are also an independent sales organization (ISO), so we have a lot of credit card payments and if a business closes their doors and people come back and say, "Well, I paid for this service. Who's going to refund it to me?" With some of our contracts, we're the people that are caught in the middle that have to pay them. It's what we call a chargeback on services. So if a business fails and they have $1 million worth of chargebacks on there of guaranteed services, in some cases, we're the ones who have to pay for that. So that's a huge risk.
Describe your working solution
To assist with risk reduction, we use Alteryx for data science with Python. We built a churn prediction model in Python that basically uses Scikit-learn to train and observe those that have churned from these business failures, and then against all of the existing churn customers, to basically build a predictive model that we gather. We gather all the source data from all of our different systems, train it in Python, push it up to Amazon S3 with Pickle, which is a clever little Python package, then pull it down and do a quick logistical regression on the results of the model against the current data set of the month we're looking at.
We then set a logistical barrier of entry, basically, a score, and then output that out in email to the front line team, so that every week, they get a list of the highest-risk customers for churning out due to business failures. The power we have with that risk of the chargebacks is that we can shut their account down. We say, "You are high risk. You have high amount of volume left on your services that it could be redeemable, but we think you're going to go out of business. So we're going to pause this, have a phone call, see how it's actually going, and move on from there."
Then with Alteryx Connect we govern all of this. All of this source data, manipulation, blending, and then the eventual models. We have a data governance department that utilizes Connect. These are the four main principles of data governance. It's data governance body of knowledge 101. It's building a culture of data, of governance, of ownership, having trust between the business, the people that own the data and those that are reporting on it, controlling access, and then having good policies around data.
And people have to know what they're looking at. If they just get a spreadsheet and they see a column that's named one way, they can assume something. But without a system of record, without dictionaries, glossaries, so on, they're going to be lost. These are the features we use out of Connect to apply these data governance principles. The glossary's obviously great for our metric dictionary, for storing links to our single sources of truth, our gilded data sources, and metadata workflow.
Of all these models that we have automated in Alteryx Connect allows us to monitor them, collect the metadata, and understand how they're performing and where their things break. We have at least five different SQL platforms that people like to program on and we have tons of different Python platforms. Alteryx serves a specific need, and we like to push people in that direction so that we can be a bit consolidated on these highly dependent automated solutions.
Describe the benefits you have achieved The beauty of Alteryx is that it's also language agnostic. We have some data scientists that know how to create a model in Python. We have others that know how to do it in R. And we have others that don't know how to do it at all that can use the built-in functionality of Alteryx.
What's great about Alteryx, is that we don't have to bring in a data scientist that knows a specific language, or a language at all. We can bring somebody right out of their program that might have some SQL skills, and if they can do SQL, you're likely going to be able to get them to train up on some other language. But Alteryx is great because we can get them immediately into the data and have them start building some predictive models.
Alteryx has been a huge gift to us in the sense that we don't have to find those perfect data scientists out there in the world. We can bring the people that we like, that have the personality, and bring them up into different languages, but also give them a tool that's not going to make them feel like they're useless for six months, but that they're able to start building and creating value for themselves. We've got a pretty comprehensive plan that we utilize, and I really enjoy, but it's helped everybody on our team start developing in a really strong way.
The entire PowerPoint presentation can be found here.