Sign up to save your podcasts
Or
Share #012 w/ Trisha Price: We Spent Years on an AI Feature—And Still Got It Wrong
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
#012 w/ Trisha Price: We Spent Years on an AI Feature—And Still Got It Wrong
Trisha is the CPO of Pendo.io, one of the most widely used product experience platforms. Trisha brought us a gem of a failure story.
Her team has been working on a powerful machine learning future for a long time.
The feature was aimed at answering one question:
What are the features that drive user retention?
Trisha’s failure story
Over the years, Pendo had gathered mountains of product usage data, such as clicks, sessions, outcomes. They even knew which users stayed and which churned. So the idea was to build a machine learning model that identifies what behaviors drive long-term retention.
Early results were promising. Some insights made total sense (“creating shared dashboards” was a big one). Others were… surprisingly boring (“visiting the settings page” meant high retention - obvious for administrators). Still, the team pressed on.
They iterated on the interface. They brought in new designers. Even the CEO jumped in. But no matter how much effort they put in, something wasn’t working.
“We got it wrong. We got it wrong. We got it wrong again.”
—Trisha Price, CPO @ Pendo
Eventually, they launched it. And while some users did get value from the feature, adoption was low. Most PMs loved the idea—they just didn’t come back to use it. Trisha and her team were left scratching their heads.
Was it the UI? Was it the model? Was it just… too static?
Turns out, just knowing which feature drives retention isn’t enough. It needs to be timely, actionable, and trusted. If the insight changes too often, users don’t believe it. If it doesn’t change at all, it’s not useful beyond the first visit.
A few things we loved from the conversation
🧠 “AI can’t replace taste.”
Trisha talked about what Megan Quinn once called taste—the combination of intuition, signal processing, and experience that defines a great PM. AI can help with execution, but it can’t develop taste. That’s why the best PMs are probably sticking around.
🎯 “There is no such thing as product strategy.”
Or rather—your product strategy is your company strategy. Especially in product-led SaaS companies, the two are often inseparable. Trisha emphasized that if your company strategy is to sell more to happy customers, your product strategy better be about creating those happy customers.
💡 “We don’t believe in fixed innovation percentages.”
Instead of saying 20% of R&D goes to new ideas, Trisha prefers a more fluid approach: start small, see what sticks, then double down. That’s how they scaled features like Session Replay. It’s also how they’re thinking about future churn prediction features.
What else we talked about
* Why junior PMs might not actually be PMs yet—and why that’s okay
* The changing PM-to-engineer ratio in the age of AI (some teams run on 1:1:1!)
* Why product marketing can’t save a weak product, and how PMs sometimes try to offload differentiation
Listen to the full episode
🎧 Spotify
📹 YouTube
🍎 Apple Podcasts
🧠 Follow Trisha: LinkedIn
🎙 Follow Christoph: LinkedIn
🎙 Follow Stefan: LinkedIn
🔥 Follow Hotfix: Linktree
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thehotfixpodcast.substack.com
View all episodes
By The Hotfix PodcastTrisha is the CPO of Pendo.io, one of the most widely used product experience platforms. Trisha brought us a gem of a failure story.
Her team has been working on a powerful machine learning future for a long time.
The feature was aimed at answering one question:
What are the features that drive user retention?
Trisha’s failure story
Over the years, Pendo had gathered mountains of product usage data, such as clicks, sessions, outcomes. They even knew which users stayed and which churned. So the idea was to build a machine learning model that identifies what behaviors drive long-term retention.
Early results were promising. Some insights made total sense (“creating shared dashboards” was a big one). Others were… surprisingly boring (“visiting the settings page” meant high retention - obvious for administrators). Still, the team pressed on.
They iterated on the interface. They brought in new designers. Even the CEO jumped in. But no matter how much effort they put in, something wasn’t working.
“We got it wrong. We got it wrong. We got it wrong again.”
—Trisha Price, CPO @ Pendo
Eventually, they launched it. And while some users did get value from the feature, adoption was low. Most PMs loved the idea—they just didn’t come back to use it. Trisha and her team were left scratching their heads.
Was it the UI? Was it the model? Was it just… too static?
Turns out, just knowing which feature drives retention isn’t enough. It needs to be timely, actionable, and trusted. If the insight changes too often, users don’t believe it. If it doesn’t change at all, it’s not useful beyond the first visit.
A few things we loved from the conversation
🧠 “AI can’t replace taste.”
Trisha talked about what Megan Quinn once called taste—the combination of intuition, signal processing, and experience that defines a great PM. AI can help with execution, but it can’t develop taste. That’s why the best PMs are probably sticking around.
🎯 “There is no such thing as product strategy.”
Or rather—your product strategy is your company strategy. Especially in product-led SaaS companies, the two are often inseparable. Trisha emphasized that if your company strategy is to sell more to happy customers, your product strategy better be about creating those happy customers.
💡 “We don’t believe in fixed innovation percentages.”
Instead of saying 20% of R&D goes to new ideas, Trisha prefers a more fluid approach: start small, see what sticks, then double down. That’s how they scaled features like Session Replay. It’s also how they’re thinking about future churn prediction features.
What else we talked about
* Why junior PMs might not actually be PMs yet—and why that’s okay
* The changing PM-to-engineer ratio in the age of AI (some teams run on 1:1:1!)
* Why product marketing can’t save a weak product, and how PMs sometimes try to offload differentiation
Listen to the full episode
🎧 Spotify
📹 YouTube
🍎 Apple Podcasts
🧠 Follow Trisha: LinkedIn
🎙 Follow Christoph: LinkedIn
🎙 Follow Stefan: LinkedIn
🔥 Follow Hotfix: Linktree
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thehotfixpodcast.substack.com