LLenny's Podcast
← All frameworks
InnovationJason Droege

Wide Aperture, Then Coalesce

Keep considering many ideas — including ones that look bad — and test cheap versions until the signals converge on one

Difficulty
Moderate
Time to result
~months to results
Steps
3
Confidence
85%

Droege's idea-selection process: hold a deliberately wide aperture on candidate ideas for as long as possible, including ideas that seem bad at first, and keep digging or cheaply testing versions until the evidence coalesces on the winner. He explicitly disclaims being 'the smartest person in the room' and lets many small experiments do the deciding.

Origin

Jason Droege's method for choosing what business Uber should enter, which produced Uber Eats after ~15 tested versions of adjacent ideas.

Core principles

  • 01Ideas that look bad at the start may be wrong-looking, not actually bad — keep digging to find out
  • 02Stay open far longer than feels comfortable; narrow only when everything coalesces
  • 03Cheap real-world tests beat armchair judgment about what will work
  • 04Let signals across many attempts, not a single hunch, pick the winner
  • 05Global/broad launch surface can be an asset when testing which idea pops

How to run it

  1. 1

    Hold the aperture open

    Maintain a very wide set of candidate ideas for as long as you can, deliberately including ones that seem bad, and resist premature narrowing.

    Pro tip Adopt it as a standing philosophical principle so you don't collapse to the obvious idea too early.

  2. 2

    Run cheap real versions

    Instead of debating, launch small, low-cost tests of each candidate to see whether a 'bad-looking' idea is actually bad or just misjudged.

    Watch out Skipping basic research on a category (e.g. what a convenience store really sells) will make a test fail for the wrong reasons.

  3. 3

    Read the signals until they coalesce

    Compare results across all the tested versions and commit only when one is clearly 'popping off on all signals' and the unit economics work.

    Pro tip Speed of the early signal is itself information — a winner often shows up almost immediately.

    Watch out Don't force convergence; wait for the evidence to actually point one direction.

In the wild

Fifteen versions before Uber Eats

Droege's team tried convenience-item delivery vans (10 trucks, 250 SKUs in DC — 'crickets'), grocery, and point-to-point delivery, roughly 15 versions in all, before food delivery clearly popped on every signal with workable unit economics.

Uber Eats launched in Toronto Dec 2015 and did ~$20,000 in sales within two hours, confirming the coalescing signal; it grew to ~$20B in 4.5 years.

Common mistakes

Dismissing bad-looking ideas without digging

An idea that seems bad at the start may just be misjudged; closing the aperture early means you never test whether you were wrong about it.

Testing without doing the category research

The convenience-van test flopped partly because they 'hadn't really done the research on what convenience stores really were' — no cigarettes, beer, or slurpees meant no traffic, so the test failed for reasons unrelated to the core idea.

Is it for you?

Best for

Teams inside a resourced company scanning many possible new business lines

Not ideal for

Focused startups that must concentrate all resources on a single bet and can't afford parallel experiments

From the transcript

I keep a very, very wide aperture on ideas for as long as I can until I'm like, okay, everything is coalescing

53:00

you just keep digging and see if you're right that they're bad or you're wrong

53:30

we tried like 15 versions of all these things before we eventually just said okay the food delivery thing is just popping off on all…

55:00

From the episode

First interview with Scale AI’s CEO: $14B Meta deal, what’s working in enterprise AI, and what frontier labs are building next

Jason Droege