Rebuilding research at Weedmaps — from a specialist function to how the company understood its customers

I believe research is too important to leave under-resourced, poorly organized, or easy to ignore — even when defending it costs political capital. This is the story of rebuilding Weedmaps' research practice from the ground up: resetting who led it, changing how the rest of the organization accessed it, and, eventually, staking my own credibility on putting a politically abandoned question back in front of the CEO.

Resetting the team and installing the right leadership

When I took responsibility for research, the function wasn't structured the way it needed to be. Two of the three researchers on the team had grown deeply resistant to the direction the practice needed to go, and that resistance had started to spill into private conduct that was undermining trust on the team — enough that the third researcher came to me directly, uncomfortable with the dynamic and unsure he wanted to stay on the team at all. Because the behavior was happening privately, it wasn't a simple performance conversation; it required moving carefully but decisively before it did more damage to the team's culture and to the one researcher I most needed to retain.

I made the decision to exit both, quickly and quietly, and reset the team around a single hire: Kellyn Dunn, who I brought in to lead the practice — not just to manage research projects, but with a mandate to change how the rest of the organization learned from customers.

That was an important shift in philosophy. Research would no longer be something teams requested from researchers. It would become a capability researchers helped the organization build for itself.

From a bottleneck to a hub-and-spoke

Historically, research access follows a familiar agency-style, limiting, pattern: a small team of researchers fields requests from an entire product organization, prioritizes them, and delivers findings back later. Kellyn and I moved deliberately away from that model, positioning researchers as coaches and force multipliers rather than the sole executors of every study.

That required actually teaching people how — framing research questions, writing interview guides, interviewing without leading the customer, distinguishing anecdote from evidence, and translating findings into product decisions. We trained the entire design team, and roughly half of them reached the point of running their own validation studies independently before the practice was cut short by a company-wide layoff.

We were deliberately more conservative with product managers. Our VP of Product judged that her PMs weren't yet ready to take on research execution directly — a reasonable call given that we still had dedicated research capacity to lean on. Rather than force self-service everywhere, designers who hadn't yet run their own studies, along with PMs, partnered directly with our researcher, often through the Thursday swarms, participating as active observers rather than running sessions solo. Researchers stayed responsible for the rigor of the practice; the goal was building capability where it made sense, not removing the researcher everywhere at once.

Building a rhythm

We moved research from something that happened periodically around big initiatives to something that happened every week. First Tuesdays became customer days — teams observed and interviewed real users. Third Thursdays became research swarms, where cross-functional teams worked live product questions together instead of waiting weeks for a specialist team to get to them.

To remove the biggest friction point in continuous discovery — recruiting participants — we built our first standing customer research group using a tool called Ethnio: a panel of more than 1,000 volunteer Weedmaps customers we could reach quickly for interviews and surveys, with built-in incentive management (gift cards and other compensation) so participation was sustainable rather than a one-off ask each time. This augmented our robust use of User Testing for “new/non customer” research.

Alongside that, we ran both qualitative and quantitative surveys directly through the research team, and I made a deliberate technology choice in Sprig for in-product feedback — specifically because it was one of the few tools that worked natively across both mobile app and web. Given how app-heavy Weedmaps' usage was (65%+), that cross-platform coverage mattered more than it would have for a web-only product.

Earning buy-in

Before I joined, Weedmaps had already run a large segmentation study — a survey of nearly 15,000 customers that suggested five distinct customer segments. The findings hadn't gained traction with leadership at the time, and by the time I inherited the research function, the prevailing wisdom was to leave the topic alone rather than reopen it without a stronger case.

I didn't think that was the right long-term call for the business. Any company serious about its customers needs some working model of who they are — but I also recognized that resurfacing a shelved idea would need the right moment and the right framing, not just a repeat ask.

That moment came when the company was considering a content strategy built around creators and influencers producing cannabis-related content to drive shopping behavior. I made the case directly to our CEO: generic content wouldn't move the needle, because no single piece of content motivates every customer the same way — the most effective content would need to speak differently to different segments. Without the segmentation work we'd set aside, the content strategy itself was going to underperform. He saw the logic immediately, agreed, and personally authorized restarting the work.

From there, we ran it properly. The original 15,000-person study hadn't retained email addresses, so we couldn't validate the segments against actual purchase behavior. We commissioned a segment-typing instrument from Qualtrics — effectively a validated personality-style assessment built from the original segment definitions — and fielded it to Weedmaps customers, ultimately getting 5,000 responses.

The typing tool held up: it reliably sorted respondents into the expected segments. But when we layered in RFM analysis (recency, frequency, monetary value) across those 5,000 validated respondents, the segments didn't actually differ in how they spent. They were a fairly homogeneous population of high-intent cannabis customers — different in category experience, not in commercial behavior.

That negative result was itself the finding. It told us the growth opportunity wasn't hidden inside our existing customer base — it was outside it. The work pivoted toward external market research, looking beyond Weedmaps' existing demographic for new growth segments entirely. Kellyn had begun that market study at the time a company-wide layoff ended the chapter.

Why this is the story

It would have been easier to let the segmentation question stay shelved — it hadn't landed the first time, and there was no obvious pressure to revisit it. I pushed to bring it back, because I think that's what taking research seriously actually requires: not just running studies, but knowing which findings still matter, and reading the moment when the organization is finally ready to hear them.

That's the same conviction behind restructuring the team, installing Kellyn, building the panel, and picking tools for what our product actually needed rather than what was easiest to buy. Research doesn't get taken seriously by accident. Someone has to organize for it, make room for it, and occasionally spend their own credibility defending it.