LLenny's Podcast
← All frameworks
InnovationBenjamin Lauzier (Lyft, Thumbtack, Reforge)

Rank, Don't Filter

Every user-facing filter silently carves out your supply — turn preferences into ranking signals.

Difficulty
Moderate
Time to result
~weeks to results
Steps
4
Confidence
95%

Users ask for control, and product teams dutifully ship toggles and filters. Each one fragments the supply pool in ways users cannot see and do not actually want. Lauzier's rule: treat stated preferences as ranking inputs rather than hard filters, so a preference reorders results instead of deleting 95% of your marketplace.

Origin

Benjamin Lauzier's rule, drawn from Thumbtack's 'smoke machine' incident on the events vertical and from watching Sidecar fragment its ridesharing supply with user-facing car and driver filters while competing with Lyft and Uber.

Core principles

  • 01Users optimise their stated preferences without seeing the liquidity cost of doing so.
  • 02A filter is a supply deletion; a ranking signal is a supply reordering.
  • 03The system should be intelligent enough to know a stated preference is usually not a deal breaker.
  • 04Hyper-fragmenting supply harms the very users who asked for the control, via longer waits and worse matches.
  • 05The inverse also applies: when market health is weak, open up your supply walls and offer users things tangential to what they said they wanted.

How to run it

  1. 1

    Audit every filter for its supply cost

    For each user-facing filter or toggle, compute the percentage of supply it removes when checked. Thumbtack's smoke-machine checkbox looked innocuous and removed 95% of DJs.

    Pro tip Rank filters by (usage rate x supply carved out) — that product is your fragmentation exposure.

  2. 2

    Ask users whether it is actually a deal breaker

    Talk to the users who check the box. Thumbtack's users said they did not care that much about the smoke machine and had no idea it was removing most of the supply.

    Pro tip The gap between the checked box and the stated indifference is exactly the space where ranking beats filtering.

    Watch out Do not over-listen to feature requests here — users describe preferences, not constraints.

  3. 3

    Convert the filter into a ranking signal

    Rework the control so it affects ranking, not filtering. Supply matching the preference sorts to the top; the rest of the marketplace stays visible and matchable.

    Pro tip This preserves the perception of control while removing the liquidity penalty — the user sees their preference honoured and still gets a fast match.

    Watch out Reserve true hard filters for genuine constraints (accessibility, licensing, legal requirements), not taste.

  4. 4

    Reduce cognitive load rather than adding controls

    Default to fewer options. Users do not need more toggles than are required to be successful and happy; more control usually degrades both liquidity and experience.

In the wild

The Thumbtack smoke machine

On Thumbtack's events vertical, wedding DJ requests included a smoke-machine checkbox. Users enthusiastically checked it, unaware that only 5% of DJs had a smoke machine — the checkbox carved out 95% of supply. When asked, users said they did not really care about it.

The fix was to make the checkbox affect ranking rather than filtering, preserving both the preference and the supply pool.

Sidecar's control-first differentiation

Competing with Lyft and Uber, Sidecar differentiated by giving riders extensive filters — minimum car year, driver criteria, and even driver-set prices. On paper it was choice; in practice it hyper-fragmented supply, so a rider who filtered for a 2020+ car lost ten minutes waiting while a great driver in a 2018 Honda Civic sat unmatched.

Liquidity suffered badly and Sidecar eventually shut down; Lauzier reads the control-first design as a core strategic mistake.

Common mistakes

Building the toggle because two user segments asked for it

User research surfaces one group that wants new cars and one that does not, so the team ships the toggle. Nobody models the liquidity impact, and the supply pool fragments in a way that hurts everyone including the users who asked.

Letting supply set its own dimensions of differentiation

Sidecar let drivers choose their own prices on top of rider filters, multiplying the fragmentation. Every extra degree of freedom on either side shrinks the overlap between what supply sells and what demand buys.

Is it for you?

Best for

Marketplace product managers designing search, filtering and matching experiences under liquidity pressure.

Not ideal for

Marketplaces with genuine hard constraints (regulated services, accessibility requirements) where filters are non-negotiable.

From the transcript

and it turns out a lot of people are checking this you're like yeah hell yeah I want a smoke machine in my wedding and…

31:00

work on ways to like make this sort of checkbox affect the ranking but not the actual filtering

31:30

and I think the mistake is that you uh you know unknowingly fragment your supply in a way that has a much more meaningful impact…

29:30

From the episode

How marketplaces win: Liquidity, growth levers, quality, and more

Benjamin Lauzier (Lyft, Thumbtack, Reforge)