Research in the dark.

Fog rose thick against the chill, rolling low from a doorway hung with black drapery. A new group of trick-or-treaters pushed slowly past the fabric, into the dim ochre light beyond. Each muffled voice nudging the next to go first. Three. Two. One. Screams, then laughter. Right on cue.

The screams mean the scare landed. The laughter means I stopped short of breaking them. I'd been listening for both, watching the reactions, making mental notes of what worked and what fell flat. Adjusting on the fly, already planning the next haunt. And then last year, up on a ladder securing an eight-foot rafter, soot from the charred boards streaked on my cheeks, I realized I'd been running user research in the dark for eighteen years.

My research usually starts with a plan. I decide who represents the users, then spend weeks recruiting them. Workshops, focus groups, user interviews, a survey when the question needs counting. Then I wrap it all into a report and a deck. All to answer one question: what makes someone do the thing I'm designing for? None of that exists at the haunt. The craft is still there.

The haunt started as a decorated condo balcony. Kids and parents alike went straight to the glowstick-lit candy bowl and left slightly happier. That's when I realized I needed more than decoration to make them feel the kind of emotion I did as a kid in a community-run haunted house. I needed more control than hanging plastic sheets gave me, a real canvas to work from.

The following years saw black fabric-wrapped walls and darkness-defined spaces. Sightlines shrank to hide the payoffs, and motion-activated props were placed at just the right spot to maximize the scare without giving themselves away. When you capture someone's eye and ear with the obvious, the well-lit, it gives you a chance to scare from the other direction. If you hear a group bunching up at a closed door, that's where you send an actor in from behind. When parents refused to even come up the stairs, it was a point of pride. I introduced instant ramen to the candy table, and the high take rate surprised me. So I expanded the offerings the next year. The fabric gave way to framed panels, charred black, strong with the smell of smoke. Setting a circus theme one year taught me a lot of people really fear clowns. At least one clown-focused vignette pops up in each haunt I build.

Fog, music, lighting, and sound effects all set the mood. A consistent mood is what sustains the dread long enough for someone to forget it isn't real. Eighteen years of practice. The kids who came through as little ones show up as teenagers every Halloween. And they still flinch at the right spots.

The same toolkit I use to build products, the same triggers, reversed and repurposed. Instead of focusing on smoothing friction and earning delight, I focus on instilling dread and triggering fight-or-flight. The craft is the same wherever it's practiced. Only the apparatus changes.

I read interviewers and arrestees in their plexiglass-separated holding cells while wearing a city-issued ID. I read rooms of stakeholders and their opinions on every template in a global platform's design system and made it work for all of them. I read two groups who wanted different things from one screen, a reviewer building a case and a lender who wanted the loan out the door.

The apparatus enables the craft to scale beyond what one person can hold in their head. The craft remains. The tools deliver it to a Fortune 50 medtech company or a cabinet-level federal agency, across teams that never share a time zone. They're how the work survives long after the original team is gone. The haunt doesn't run on any of it because it doesn't have to. I'm in the middle of it every night.

The research settings, conference rooms, and development sprints I can't be in every day are where I lead, and where the craft has to work without me in the room. That's the job: getting it there, through the team.

This October, the haunt goes up again. I'll be back in the dark, listening for the screams and the laughter.