My work at Bolt

I was the researcher for Bolt’s most business-critical vertical: ride-hailing. Operating within a high-velocity, low-margin, global mobility ecosystem and two-sided marketplace, my scope covered a roadmap for millions of users across diverse international markets.

I lead research with creative approaches that had never been done at Bolt in order to bridge the gap between digital interface, the real-world operational service quality, and our business’s complex needs.

Case Study: Bolt OÜ

New Demand for Ridehailing

Overview

Mixed-methods research for Bolt’s New Demand Team on a new courier service (Bolt Send): discovery, concept evaluation, feature prioritization, service evaluation, sentiment tracking.

My work defined priorities for the product team and go-to-market strategy, identified workflow issues for customers and couriers, prioritized solutions.

Impact

  • Defined roadmap priorities and course-corrected go-to-market strategies for both app and service improvements.

  • Discovered new opportunities and evaluated solutions, ensuring maximum impact of short and long-term bets

  • Baselined and tracked shifts in customer sentiment in cadence with app releases to measure impact

  • Evaluated product-market-fit and identified opportunities for creating a business-focused service model to maximize our market share.

Research Highlights

  • Discovery Interviews (Moderated + AI-Moderated)

  • Feature Prioritization Survey (MaxDiff)

  • Post-Order Survey for Sentiment Tracking

If you want to evaluate your service, who do you talk to first?

Well…

Your customers, right?

As a two-sided marketplace, Bolt has two “customers”:

The riders
the demand / true “customer”

The drivers
the supply

I needed fast and comprehensive gut-checks on how this new service was really playing out in each market, so I went to drivers first.

Despite our usual playbook and my charter primarily being to the Rider Experience team, in this case drivers were a much richer data source.

  • Driver’s are the only ones with the full view of the Bolt Send journey. Unlike a typical ride, customers don’t actually experience the majority of the order; they hand-off the package and never actually witness the rest of the order process.

  • Drivers have more breadth and depth with the experience and overall market dynamics. While a customer telling us about their couple of orders is valuable; in order to gauge market-wide norms, typical processes, frequencies of issues, etc., a driver can tell us about the hundreds of orders they’ve fulfilled. They can tell us about the different delivery types (e.g. envelopes vs furniture vs food, apartments vs estates vs restaurants).

Prioritizing solutions for improving a barebones MVP.

Most MVP’s ship with tons of ideas for future improvements.

How do we know what’s worth building?

What if the experience is already good enough as-is?

What if all the ideas are really really good?

The approach

I approached this research with rapid, iterative phases instead of one big-bang. While it might be tempting to jump right to a multi-market, high N survey, this phased approach gave us much greater data validity and a much more acute ability to interpret the large-scale survey data.

My careful approach proved itself along the way as we uncovered new insights that revealed a massive interpretation gap between what we intended for concepts and how users interpreted them. If we had started with a large N survey, we’d have misinterpreted the data and made the wrong decisions.

Phase 0: Brainstorming based on my discovery research

After reviewing data from customer support tickets and my discovery research, we came up with over 30 different potential concepts together.

Phase 1: Concept Comprehension Checks

We put the concepts in front of users as just briefly written descriptions as a sort of “live MaxDiff” exercise. We listened to how they interpreted them.

  • What did they think this idea was?

  • What questions or concerns did they have, and why?

  • How did they think it would really work (or not) in real-life?

We started this with live-moderated sessions and then added AI-moderated interviews to cover additional markets.

This scrappy stage of the research revealed some very serious interpretation gaps between what we thought our descriptions meant, and how real users actually interpreted them. Things that if we didn’t know, would have caused us to make very wrong assumptions from survey data!

Phase 2: MaxDiff Survey

Using our refined feature descriptions, we sent this MaxDiff survey to our top Bolt Send users in 8 of our top markets.

Almost all the features came back with high preference. This wasn’t surprising to me; the current experience was an MVP and so there were a lot of good ideas to improve it. This is why MaxDiff was the perfect method here. Even though all the ideas were good, we still got to find out which ones had more share of preference.

Phase 3: Targeted Follow-Ups

The survey data showed us what was winning, but we needed to know more about why. I followed up with a series of interviews with the MaxDiff respondents to gain deeper insights into how they considered trade-offs between the features and to understand why some features scored so low for some individuals, but so high for others. Did they have different use-cases? Or just different preferences overall? Why?

The Impact

This gave the team strong data for defining their roadmap to leadership for our 2027 annual planning; including the scoped future phases of research/design required. Sure, we had a list of top features and a rough scope around them, but there was still a lot to be done to know exactly how to execute these things.

Tracking sentiment shifts to ensure our execution actually moved the needle.

It’s one thing to be diligent within prioritizing concepts, but execution is everything. How do we know if what we delivered is solving the problems we identified? How do we know if it actually shifted how they think about our service overall?

The Approach

I created post-order surveys which we used as a long-term health measurement; tracking how customer sentiments shifted in each market after major UX releases (~15 countries, 30+ cities).

The Framework

This was the first time our research team had sent post-order surveys like this at Bolt. Over about 2 months, I partnered with our Lifecycle Marketing and Engineering team to get this kind of system up and running. It wasn’t just “sending a push notification after order_finished event”. For example, we needed a system for sending it to only 1 out of every 100 orders placed, not sending to the same user for another 8 months, sending a next day follow-up email if they didn’t click the push-notification, etc. This was no small feat, but it is the type of long-term research infrastructure that pays off across all the verticals and for years to come.

Phase 1: Baselining

I started by measuring several sentiment factors to create a baseline in each market (e.g. product-market-fit (Sean Ellis), retrospective choice, CSAT, perceived effort, and more). I also used it to get us some rather simple but much needed order-profile data that we just totally lacked. Things like: what kind of things were being sent, and was it for personal or business purposes?

Phase 2: App Update Release Cadence

I then synced up the survey to be in cadence with our product team’s releases in order to measure if/how their releases were shifting perceptions (or not).

Outcome

We identified markets with unique use-cases, unique challenges, and unique overall sentiments. We were able to track how specific features impacted sentiments. Such. as how adding the recipient’s phone number impacted perceived effort in the order. Or how additional verification features impacted satisfaction.

Ultimately, it set the team up to measure their impact on user sentiments over time and set goals based on this new metric.

Summary

This research employed several methods that were entirely new to Bolt (post-order surveys, MaxDiff), was executed quickly in multiple phases to get the team what they needed when they needed it without sacrificing velocity, and gave the team lasting clarity on both their short-term and long-term vision for this vital bet for the business.