A change that helps one side of your marketplace can starve the other by Tuesday
Pricing and matching do not have an isolated blast radius. Move a parameter and supply reacts, then demand reacts to that. Expose the change one region at a time and let it stop itself when the balance tips.
Creating an account with Google shares your name, email address and profile picture with us, so we can identify you and contact you about your account. Our stated retention period is the life of the account plus 30 days after you close it.
You find out from the supply side, three days late
The change looked good on the metric you were watching. The metric you were not watching is the one that decides whether couriers keep opening the app next week.
The effect arrives out of order
Demand responds within hours. Supply responds over days, once people have decided the economics changed. By the time the second effect shows up, the change is everywhere.
A global rollout has no control group
Shipping to every region at once means the comparison you need is the same region last week, against a different day, a different weather, a different set of promotions.
The dashboard is not a circuit breaker
Someone has to be looking at it, recognize the shape of the problem, and decide to act. At 2am on a Saturday, all three of those are optimistic.
What changes when the rollout watches both sides
Same parameter change, contained to a blast radius you chose.
One region at a time, reusable as a segment
Define the region once as a segment and reuse it across every rollout and experiment, so the group you exposed last time is provably the same group this time.
Guardrails on the side you are not looking at
Put the supply-side metric on the guardrail, not just the demand-side one you were optimizing. The rollout stops when either moves the wrong way.
The result is attributable
Exposure is recorded per user, so the effect on each side is measured against who actually received the change rather than against the calendar.
How it works
Define the region as a segment
Targeting rules live in configuration, not in a deploy, so widening or narrowing the exposed region does not need an engineer.
Guard both sides of the market
Attach one guardrail to the metric you expect to improve and one to the metric you are afraid to break. Either can stop the rollout.
Promote only on evidence
A stage advances after its window passes clean. The decision, and the numbers behind it, stay attached to the rollout.
Change the algorithm without betting the whole marketplace
Fifteen minutes from signup to a region-scoped rollout with guardrails on both sides.
Start on GrowthCreating an account with Google shares your name, email address and profile picture with us, so we can identify you and contact you about your account. Our stated retention period is the life of the account plus 30 days after you close it.
Reusable segments, guardrail auto-rollback and unlimited experiments included.