How Turo Replaced NPS with Future Purchase Intent (FPI)
tl;dr: Turo's NPS looked great for a decade and told them almost nothing. Replacing it took 12 months of work that most legacy survey platforms can't actually support. Here's what Kathy Lin (Staff Quant UXR at Turo) walked through, and why every team still running measurement on form-era tooling should be paying attention.
For most of the last decade, NPS was the answer. You picked it, you reported it, executives nodded, you moved on.
That stops working when your one metric stops responding to reality. Product launches don't move it. The market shifts, nothing. A global pandemic comes through, and still nothing. At some point the org looks at the score, looks at the actual business, and asks the question every research team eventually has to answer: if the metric never moves, what is it actually telling us?
That is exactly the question Kathy Lin (Staff Quant UXR at Turo) tackled in our webinar titled Beyond the North Star: Building Experience Metrics that are Actually Useful. As Turo’s first-ever quant research hire, Kathy stepped into her role with a mission that was simple on paper but brutal to execute: replace NPS with something truly useful.
Below are the three lessons the research community aren’t talking about enough.
Lesson 1: Prioritize Operational Focus Over Empirical Completeness
Kathy's first instinct, and honestly the right instinct for any researcher, was to refuse the single-metric trap. She presented three candidate metrics to her CPO (NPS, satisfaction, and a future-purchase-intent measure she'd been building) and laid out their tradeoffs.
After her presentation, Kathy’s CPO asked her a career-pivoting question:
"What are we going to do when these metrics don't all move the same way?"
That single line cuts right to the core of the issue. A research program isn't a research paper. The point is to give the business something it can actually use to make decisions. Three metrics meant three arguments, three explanations, three rounds of internal politics. One metric, even an imperfect one, meant operational focus.
Researchers default to complete, while executives need useful. Those aren't the same thing, and the second one is much harder.
Turo landed on Future Purchase Intent (FPI): "would you book again with Turo for your next rental need?" Determining one question that moves when the experience moves and correlates with actual rebooking behavior was the metric leadership could rally behind.
Aligning leadership around one operational metric is a massive win, but choosing the metric is only half the battle. Next comes the execution.
Lesson 2: The Mobile Survey UX Surprise (In-App vs. Email)
Email surveys at Turo were getting a 3% response rate. Kathy moved the survey in-app and the response rate jumped to 40%. When she started asking the same question in a different channel, the score moved 20 points.
This is the part that should make every team running NPS through Qualtrics or SurveyMonkey a little nervous.
That 20-point swing is not a rounding error. It’s the kind of change that should make you question every dashboard you've ever built. The diagnosis came down to mobile design. The initial in-app layout only text-labeled the extreme ends of the scale (the 1 and the 5), which accidentally biased users into defaulting to the top score. Explicitly writing out text labels for every single option on the scale instantly fixed the skew and brought back an accurate, realistic spread of data.
This story matters beyond Turo, because most legacy survey tools physically cannot run this experiment. You can't iterate survey UX in real time on a platform built to email a static form to a panel. Kathy found the problem because she could fire test variants against live users and watch the response distribution shift in days, not quarters. he era of "set up the survey, see you next quarter" is over.
As Kathy summarized it:
"Don't view measurement inconsistencies as roadblocks. Use them to build confidence."
However, that only applies if you have the tooling to actually run the diagnostic.
Lesson 3: Productize research through cross-functional alignment
A company-wide metric isn't a research project, it's a cross-functional rollout. Engineering owns the trigger logic. Data analytics owns the journey mapping. Data engineering owns the pipeline. None of those teams report to research, and none of them will help you if you show up with a finished metric and ask them to wire it in last. Building that cross-functional buy-in is the lesson most research talks gloss over.
Instead of walking into executive meetings with polished slides, Kathy brought in work-in-progress and asked for active input. By the time the metric was ready to launch, the leadership team already had a narrative for how they'd use it, which meant the internal rollout had effectively already happened.
Something I’ve been touting for years is that research has to be productized. You don't ship a product by perfecting it in a vacuum, rather you ship it by getting your distribution partners bought in before the launch.
From in the loop to on the loop: Enhancing Efficiency with AI Agents
The conversation ended where every research conversation now ends: AI.
It was inevitable that the session would turn to AI.Kathy built this program a year ago, before the widespread adoption of agents. She told the room that if she were starting again today, she'd run the entire validation phase through Cursor. Repeat correlations, distribution checks, and month-over-month diagnostics are all agent-eligible work now. She estimated that the validation phase that used to take 100% of her time takes up roughly 50% of her time today.
That's not a story about saving time, its a story about where the researcher's job is moving. Less time hands-on-keyboard and more time spent on narrative, strategy, and thinking about what the data is actually telling the business. Less in the loop, more on the loop.
How to Build a Modern Customer Measurement Program
Turo's NPS didn't lie; it told the truth about something nobody was still asking about. The score was stable because it was measuring loyalty among the customers who showed up to take a survey at all, while the actual experience was shifting underneath it.
NPS isn’t broken, but any measurement program built on top of legacy survey infrastructure is one product change away from telling you nothing. The teams that will get measurement right going forward are the ones who can change the question, the channel, and the analysis as fast as the product itself changes.
The reality is simple: organizations still running on legacy infrastructure don't realize the massive competitive advantages they are leaving on the table. Moving past those limitations is how you turn a passive metric into an operational engine.
Watch the On-Demand Webinar: Beyond the North Star and learn how Kathy managed segment sample sizes, built Turo's validation matrix, and handled overlapping survey conflicts.
.png)
