Introducing the Propensity Scores Data App: Ship actionable churn and conversion scores with velocity

The ability to predict customer behavior is no longer a luxury. It’s a necessity to give your business a competitive edge. However, while the rapid injection of AI into our daily lives accelerated the push for every company to leverage ML to its full extent, most companies are struggling to deliver.

If your efforts to drive business outcomes with ML are stalled because of the challenges around creating and deploying ML models, you’re not alone. KDnuggets reported that 80% of ML projects fail before deployment.

Today, we’re launching our Propensity Scores Data App to address this problem head on. The app runs on top of our Profiles Product and leverages your customer 360 data from Snowflake to generate accurate churn and conversion scores. You get direct access to its code-based models, so you can configure them to meet your business requirements, and you can easily sync actionable outputs downstream to business teams for tactical activation.

Drive business outcomes fast with automated ML

When you get predictive analytics right and operationalize the results, the impact speaks for itself. Take customer churn, for example. Identifying churn risks and engaging customers before they leave can make a double-digit impact on revenue.

Driving these types of results, however, involves overcoming several technical challenges, and there’s no easy button. The closest thing to an easy button, SaaS solutions for predictive analytics come with significant limitations. These tools lack both the data access and flexibility needed to generate accurate, actionable insights. For example, an email tool may be able to predict a churn risk and recommend the right message to send, but its prediction lacks the context of the rich customer journey activity and relevant customer information that lives outside its walls. It’s unable to tap into the wealth of customer data in you have in Snowflake.

The DIY approach eliminates the restrictions of black box SaaS tooling, but it requires significant investment and involves several challenges. Drawing from our collective experience of building numerous models in production, we’ve found that the hardest part about building predictive analytics for many practical use cases is less about developing the models – it’s in the pre and post work.

First, you must collect and clean the data to train models. This prerequisite data engineering, which includes solving identity resolution, can often consume more time than the training of the model itself. Working with time series data, which is essential for most customer data use cases, adds additional layers of complexity. With time series data, the features for the training data set need to be calculated at a point in history (like when the customer churned). Plus, when definitions change, you have to recreate those event-based datasets, which are often quite large.

After collection and cleaning, you must train and deploy the models into a production environment. In production, monitoring for model drift, such as deviations from expected precision and recall as new data arrives, becomes essential. Once drift exceeds your predefined thresholds, you’ll have to retrain the models.

Each of these processes demands a significant investment in technical expertise and bandwidth.

Through close collaboration with customers, we've seen firsthand the frustrations arising from SaaS and DIY approaches. It’s clear that data engineers and analytics teams need a solution that combines power and flexibility – one that can harness the full potential of Snowflake without the overhead of complex MLOps infrastructure.

Enter RudderStack Profiles and the Propensity Scores Data App.

Propensity modeling on your Customer 360 data without the MLOps

Our Propensity Scores Data App makes it easy to generate propensity scores using your own churn or conversion definitions and automatically sync them to user profiles in business tools.

What makes it so powerful isn’t the app itself. It’s the tight integration with RudderStack Profiles and Event Stream. 

RudderStack Profiles enables you to power your business (and in this case the Propensity Scores Data App) with reliable, complete customer profiles. It works seamlessly with our Event Stream product, automatically creating an identity graph from your time-series data and enabling you to quickly build features on top to produce a comprehensive customer 360 with all of the relevant data in Snowflake.

With this foundation, you can run The Propensity Scores Data App on top of your customer 360 to ship actionable churn and conversion scores to your business in days, not months.

Here's what makes the Propensity Scores Data App stand out:

• No data silos, complete control – The Propensity Scores Data App operates within your existing Snowflake environment, giving you full access to and control over your data and eliminating the data silos often created by third-party SaaS solutions.
• Flexible declarative workflow: Our intuitive config-based workflow gives you granular control over model parameters and allows you to tailor models to fit your business needs without requiring extensive data science expertise.
• Reduced complexity: The Propensity Scores Data App takes full advantage of our platform and Snowflake’s Snowpark compute infrastructure to eliminate the need for complex modeling and MLOps. This makes advanced propensity modeling accessible to a wider range of organizations and enables data science teams to focus on more valuable exploratory work.

How the Propensity Scores Data App Works

The Propensity Scores Data App fits into our existing Profiles workflow, so you can build predictive features on top of your Customer 360 without changing your existing process. Predictive features in 5 steps:

1. Define the action you want to predict. This might be a lead conversion, a payer conversion, reactivation, or subscription churn.
2. Identify eligible users. Specify the user set you want to use for model training, such as active users for churn or unconverted leads for lead conversion.
3. Set the prediction window. Define the timeframe for your prediction. You use predict_window_days to set this according to your use case.
4. Define predictive features. Set the features you want to predict.
5. Run the model. You can run your project in the Profiles CLI or by uploading it to a Git repository and importing it into the RudderStack UI.

Once you run your project, you can view the training and prediction outputs. The model generates tables in Snowflake, and if you’ve imported your Profiles project to the RudderStack dashboard, you get a rich view directly in your UI.

Bring your own feature table

We designed the Propensity Scores Data App to work seamlessly with RudderStack Event Stream and Profiles, but it doesn’t limit you to feature tables created with Profiles. If you have existing feature tables created in a tool like dbt, you can easily define those as input models for Propensity Scores.

Delivering actionable data to your business stakeholders

With the output from the Propensity Scores Data App in Snowflake, you can automatically sync actionable data to any of our 200+ destinations with our Reverse-ETL pipeline. You can even leverage our Real Time Personalization Data App to make the data available via API to personalize web and product experiences.

Get started

The Propensity Scores Data App is more than just a tool – it reflects our commitment to empowering every data team to solve advanced use cases and ship high-ROI data projects to their businesses. You no longer have to choose between inflexible SaaS solutions or maintenance-heavy DIY for predictive analytics. Now, you can quickly generate predictive metrics based on all the relevant data from your Customer 360 and easily get actionable data into the hands of marketing, product, sales, and customer success teams for activation.

To learn more about how RudderStack can help you deliver actionable churn and conversion scores to drive smarter growth faster, check out the docs for Profiles and the Propensity Scores Data App. To see the Propensity Scores Data App in action, Request a demo with our team.