Growth Is a Game of Inches
Don't chase the step function. Ship 20 bets at 20% instead of 4 perfect ones — then raise the hit rate.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 90%
Liu's model of growth as a shots-on-goal machine rather than a search for a magic bullet. The math: a team that ships four carefully-planned things gets four wins at best; a team that ships twenty at a 20% success rate gets the same four wins plus sixteen lessons — and if it lifts the hit rate to 30%, it gets six. Growth is therefore a learning engine, and it operates as an optimization layer wrapped around a working core product, not as a replacement for one.
Origin
Deb Liu, from running payments growth at Facebook, the ads product, Marketplace, and now growth at Ancestry. She cites the internal Facebook lore that adding a 'create an ad' link — just a link — was one of the biggest growth drivers, because people simply didn't know how to reach the ads flow.
Core principles
- 01You get to 3x one percent at a time. Points of growth compound.
- 02It is absolutely okay if 80% of your ideas don't work.
- 03Throughput beats planning: more shots at a constant hit rate produces the same wins plus the lessons.
- 04Raising the hit rate from 20% to 30% is a 50% output increase in the same calendar time.
- 05Growth is the icing, not the cake — it sits on top of a core product that already works.
- 06The wins are often embarrassingly small things: a button, a link, a placement.
How to run it
- 1
Verify the core engine works first
Growth is optimization on top of a functioning core. At Ancestry, search and hints are core: separate teams keep uptime good, hints working, and search delivering. Growth then makes the flow faster, surfaces hints better, and improves acceptance.
Watch out Applying growth tactics to a broken core is polishing a product nobody wants. The core teams are a prerequisite, not an alternative.
- 2
Build a list of a hundred hypotheses
Liu's payments growth team kept a running list of a hundred things they were working on as hypotheses. The list is the asset — it removes the pressure to find the one perfect idea.
Pro tip Include the ideas that seem silly. The 'create an ad' link seemed trivial and was one of Facebook's biggest growth drivers.
- 3
Ship in batches of ten through sprint cycles
Pick the first ten, work them, then the next ten, then the next ten, running them through sprint cycles. Same pattern she used on payments, ads and Marketplace.
- 4
Optimise for throughput, then for hit rate
Compare a team that ships four things to one that ships twenty at 20% success: same four wins, plus sixteen lessons about what doesn't work. Then push the hit rate from 20% to 30% — suddenly six things work instead of four, a 50% output increase in the same time.
Pro tip Throughput is the first lever because it also generates the learning that raises the hit rate. They compound.
- 5
Treat the failures as the learning machine
The 80% that don't work aren't waste — they're what you reiterate on and change. Growth is 'an engine of a learning machine of what doesn't work'.
Watch out This requires a culture that doesn't punish the 80%. If failed experiments cost people their promo, throughput collapses.
In the wild
Liu recounts the Facebook story that one of the biggest growth drivers was simply adding the words 'create an ad' as a link — because people didn't know how to get to the ads flow at all.
→ An apparently silly change that bent the curve, and the canonical argument for keeping unglamorous ideas on the hypothesis list.
The payments growth team maintained a list of a hundred hypotheses, picked the first ten, worked them through sprint cycles, then the next ten, then the next ten. Liu ran the same pattern on the ads product and on Marketplace.
→ Small things adding up to big things across three separate billion-dollar-scale products.
Common mistakes
Hunting for the magic bullet
People treat growth as one step-function move. Liu: you get to 3x one percent at a time, five percent at a time — single-digit growth compounded.
Overthinking the perfect battle plan
PMs try to plan the perfect set of four bets. The team that ships twenty imperfect ones matches their output and learns five times as much.
Confusing growth with the core product
Growth is the wrapper — faster flows, better surfacing, better placement. It cannot substitute for a core engine that delivers value.
Is it for you?
Best for
Growth PMs and product leaders with a working core product who are being pushed to find a single step-function unlock.
Not ideal for
Pre-product-market-fit teams whose core value proposition doesn't yet work — optimization on top of nothing yields nothing.
From the transcript
“it's like if you can move things 1% a little bit faster every single week think about the amount of growth you get at the…”
“we had a list of a hundred things we're working on hypotheses and then we would pick and we would grow them by picking the…”
“imagine you're a team and you can ship I don't know like let's say four things but what if you're a team that can ship…”
“and by the way it is absolutely okay if like 80% of them don't work”
“growth is a is is actually optimization on top”
From the episode
Succeeding as an introvert, building zero-to-one, and why you should PM your career like you PM your product
Deb Liu (CEO of Ancestry, ex-Facebook, PayPal, eBay)