That's the insight that sits at the center of how Pendo thinks about Pendo Predict, our AI-powered predictive analytics product. And it's an insight that changes the conversation completely, once you understand it.

Jonathan Rice, a Data Analyst on Pendo's internal data science team, put it plainly in a recent conversation: "I think a data team with a good head on their shoulders can build a churn model, and probably already has."

So if good data teams can already build predictive models, what's the point of Predict? It turns out, the model is only part of the problem, and not the most important part.

The missing piece of homegrown data models

Here's the part of the predictive analytics conversation that tends to get glossed over: a model sitting in a data warehouse doesn't save a single customer.

"The model itself doesn't actually change the outcomes of your business," Jonathan explained. "It's people driving outcomes from the model predictions that actually changes things."

Data teams know this tension well. You spend weeks, sometimes a full quarter, building something you feel genuinely good about. The predictions look right. The recall is solid. And then... it just doesn't move net recurring revenue (NRR), because there's no clean path from "here's a high-risk account" to "here's a CSM doing something about it."

The question is, who owns that work? Is it the data team's job to drive action, or does responsibility hand off to RevOps, or enablement, or Customer Success (CS) leaders? The answer is usually "everyone sort of owns it, so no one really owns it."

Predict is built to solve this problem not by replacing the model, but by doing everything that comes after prediction scores.

What Predict actually does (and what it doesn't try to)

Predict isn't a black box that claims to know your business better than the people who've been living inside it, it's a platform that compresses the time between "we have a model" and "our team is doing something with it."

Where Predict really shines is in everything that happens after model training: getting predictions into Salesforce and Slack (or wherever your GTM team lives), building Pendo segments from score tiers, attaching playbooks so CSMs know what to do when an account surfaces as high risk, and retraining automatically so your scores don't quietly go stale.

That last piece matters more than it might seem. Jonathan described Pendo's previous churn model as something that had gone years without being rebuilt or retrained. The business had changed significantly in that time, but the model was still running on old signals. 

"We were going to need to rebuild the model," he said. "The problem was, it’s not an easy task. It would have taken a data scientist the better part of a quarter."

Predict flips that timeline. With automatic retraining enabled, the model continuously rebalances feature weights based on observed outcomes, so your scores reflect how your business actually works today, not how it worked when someone had time to build a model two years ago.

The time shifts, not the work

Before Predict, the model-building phase wasn't necessarily the longest part of the process. Jonathan broke it down this way: gathering data and engineering features takes a long time. Training the actual model is fast. But then figuring out how to get predictions where they need to go, writing custom code to push scores into Salesforce, or surface them in a dashboard, or trigger a Slack alert, takes a long time again.

"Predict shifts how much time you're spending in those three areas," he said. That's where you see the huge time savings. And now, data teams can invest more time and effort in gathering better data."

The Pendo team saw this firsthand when they started experimenting with external data enrichment for their account tiering model. They wondered whether pulling in additional signals about prospects (like whether a company was visibly using AI) would improve the model's accuracy. Within one day of gathering the data and feeding it into Predict, they had their answer: yes, it was informative, and they pushed out a new version of the model. Everyone downstream just received better predictions. No one had to know anything had changed.

The unlock for data teams

Probably the most underappreciated benefit of Predict, from Jonathan's perspective, is what it does for data teams' visibility inside the business.

"Data teams need to communicate their ROI," he said. "Being able to show that the models you're building and the actions you're driving off of those are bringing dollars to the business, or saving dollars from exiting the door, is a huge unlock for data teams."

He drew a direct comparison to what Pendo does for product managers: PMs have a tool that shows them whether the things they're building are being adopted, and helps them make those things better. Data teams haven't had that. Predict is built to be that tool.

And because Predict ties directly into Pendo's broader platform, including segments, Guides, and Orchestrate, the path from model score to in-app action is much shorter than it would be if a data team were trying to wire all of this together themselves. When every account with a high churn score can be automatically added to a Pendo Orchestrate journey, and you can track engagement with those guides and emails the same way you'd track any other Pendo data, the loop from prediction to action to outcome becomes measurable.

Want to see how Pendo Predict can help your team turn predictions into action? Get a demo.