Phase-Bounded Big Bets
Dream big, but chunk the bet into named learning phases with quarterly go/no-go gates
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 92%
Zhang's reconciliation of visionary ambition with execution discipline. Rather than choosing between 'shoot for the moon and figure it out' and 'do the math first', you break the big vision into explicit, time-boxed phases, name what each phase is meant to learn, and run a go/no-go at every milestone. This makes killing a bet at one quarter emotionally survivable, instead of sunk-costing into year two.
Origin
Zhang's post-mortem lesson from Airbnb Plus, the inventory-inspection programme she worked on and eventually left, and reinforced by her year at WeWork during the 2019 collapse.
Core principles
- 01A big vision is necessary to innovate — the failure mode is not ambition, it is unbounded ambition.
- 02Be explicit with the team about which phase you are in; 'we're in a learning phase' is a statement that changes behaviour.
- 03Unscalable, manual things are legitimate in the prototyping phase — but only if the phase is named.
- 04Sunk-cost fallacy compounds with time; short intervals starve it.
- 05Unit economics must be interrogated at the start, not assumed to resolve at scale.
- 06Hiring is gated by the same milestones — over-hiring against a vision you haven't validated ends in layoffs.
How to run it
- 1
Interrogate the unit economics before the bet, not after
Do the math on whether this can ever be a business that makes money. Refuse the magical thinking that says the economics will work out once you get to scale X.
Watch out 'When we get to the scale of X it'll all work out' is the tell. Zhang does not believe the Airbnb Plus unit economics ever would have worked, and says they should have dug into that at the very beginning.
- 2
Declare the phase you are in, out loud, to the team
Say explicitly: we are in the learning phase for three (or six) months. Nothing else about the work reads correctly to the team without this label.
- 3
Name the specific things this phase will learn
List the concrete hypotheses under test — e.g. are the signals telling us communication with hosts is the real problem, does listing type or multi-property hosting change the picture. Vague learning is not learning.
Pro tip Do unscalable things in this phase deliberately — they're cheap and they generate the signal fast.
- 4
Articulate success and the milestones in small intervals
Define what success looks like at the end of the quarter and each quarter after. Make each milestone an explicit go/no-go point.
Watch out Without the milestone, the alternative isn't a shorter bet — it's a two-year bet you only notice at the end.
- 5
Gate hiring on the same milestones
Tie headcount unlocks to passing the gates and showing the results. Do not staff a team for the size of the vision before the bet has cleared its learning phases.
Watch out Zhang says WeWork 'hired beyond our skis on the tech side' — laying off half a team you personally recruited is the price of skipping this.
In the wild
Facing competitive fear of managed marketplaces, Airbnb went solution-first: inspect the inventory, manage the inventory. The real problem — people wanting to know what they were getting into — could have been attacked far more cheaply and specifically. A lockbox is cheaper than deploying an inspector; a local cleaner partnership priced into the fee is cheaper than an inspection; the guest review system is essentially free and scales. Zhang eventually moved off the team and onto the review system.
→ A blunt, capital-intensive instrument was applied to a set of problems that each had a cheaper targeted solution, and the unit economics never worked.
WeWork was operationally excellent, and its real lever was inventory and inventory management — making it easy for sales and operations to get inventory onto the platform. Instead the company invested deeply in a large tech team building platform-aware, futuristic capability users didn't care about.
→ Over-hiring against an unvalidated technology thesis, and layoffs. Zhang's takeaway: you can still dream big without hiring big for a vision the milestones haven't earned.
Common mistakes
Magical thinking about unit economics
Assuming that scale will fix an economics problem you can already see. If the math doesn't work at the start, be specific about which scale effect fixes it, and test that.
Letting a bet run long enough for sunk cost to take over
After two years of investment the argument becomes 'let's just go a little bit longer, maybe next quarter'. After one quarter, killing it is emotionally cheap — you invested a quarter, not two years.
Is it for you?
Best for
Product leaders and founders running an ambitious new bet inside a company that prizes visionary thinking and struggles to kill anything
Not ideal for
Core-product execution work with known economics and known scope, where the go/no-go ceremony is pure overhead
From the transcript
“so I think it's totally fine to be like hey we are going to try X for six months three months you know whatever it…”
“you can do what I call like unscalable things in that prototyping phase”
“you should articulate what success looks like and the Milestones you want to hit in the small intervals”
“and every single point it was a go and no go”
“you have like magical thinking around unit economics”
From the episode
Building minimum lovable products, stories from WeWork and Airbnb, and thriving as a PM
Jiaona Zhang (Webflow, WeWork, Airbnb, Dropbox)