The most important project at Lyft was stuck, and the room had twenty people in it.
This was the project meant to unify Lyft's transportation modes in a single app: rideshare, bikes, scooters, transit. Roughly twenty designers, researchers, and product managers had produced plenty of good concepts and no strategy. Executives were unhappy. The CEO was in the room, which at that size of company is the fire alarm. I was asked to take it over.
Options as avoidance
The first thing I learned was that nothing was missing. Not talent, not research, not ideas. Twenty capable people can generate options indefinitely, and options had become the deliverable, because producing another direction is safer than arguing for one. Accountability had diffused to the point where no one person would have called any part of the work theirs, which meant no one could be wrong, which meant nothing could be decided.
Twenty capable people can generate options indefinitely. Options are how a room avoids deciding.
A large team on a stuck project is usually read as commitment. Sometimes it is the cause of it: every person added raises the coordination cost and lowers the personal cost of being wrong.
Subtraction
I cut the team to seven. Not a layoff. The other thirteen went to work that needed them, which mattered, and which I would still handle better today. Being moved off the company's flagship reads as a verdict unless you spend real effort making it not one. I spent some, and it was not enough for everyone.
Choosing the seven was the real design decision. What I selected for was not seniority in the title sense (two of the seven were mid-level) but the willingness to hold a position in a room where the CEO might disagree, and the appetite to be accountable for it afterwards. Judgment, in other words, and the nerve to spend it.
Three directions
We cut the concepts down to three, argued them against each other honestly, and picked one. The mechanics were not exotic: fewer people, each holding more of the problem, with disagreement made cheap early instead of expensive late. In user testing of the resulting tab view, use of the modes beyond rideshare (bikes, scooters, transit) rose by half. The team's work traveled further than the app; it featured in the roadshow materials for the March 2019 listing.
The smaller room changed the argument before it changed the design. With twenty people, disagreement had been expensive: a dissent in that room cost social capital in front of an audience, so dissent arrived late, softened, or in the hallway afterwards. With seven, arguing was simply the work. We put the three directions on the wall and required each person to attack the one they preferred, which sounds like a workshop trick and functioned as insurance: by the time the CEO saw the recommendation, its weaknesses had been found by the people who liked it most.
Speed followed for a reason that had nothing to do with effort. Coordination is quadratic and judgment is not. Seven people who trust each other cover more decisions per week than twenty people covering for themselves.
Judgment density
The rule I took with me: on a stuck project the constraint is almost never capacity. It is judgment density, the ratio of people who can decide to people who can produce. Adding producers dilutes it. If a project matters enough to be staffed like a department, it usually matters enough to be run like a founding team instead.
The organization followed the same logic later. Director-level peers and I redesigned the structure in a stretch with no VP in place, around the app's new surfaces, and the incoming VP kept it. Shrinking the team was never the point. Deciding was, and the team's size was what stood between the room and a decision.
I led product design at Lyft through the multimodal transition, one of the senior directors running the function in a stretch without a VP. The counter-argument I take most seriously: this is survivorship. Small teams fail too, and when they do there is no slack to absorb it. Size should follow the shape of the decision to be made. Ours needed seven. Yours may well need twenty.
© 2026 Renato Valdés-Olmos