Teams often ask whether they have “enough traffic” for a product experiment. The more useful question is whether the planned experiment window can support the lift they claim to care about.
Start from the decision, not the dashboard
If shipping the change requires a 2% relative lift on activation to be worthwhile, write that number down before choosing a calendar date. Product experiment measurement only helps when the bar for action is named in advance.
Translate lift into people
Take the baseline conversion you already observe, apply the lift you would act on, and estimate how many unique exposures each variant needs. Bangkok commerce apps with sharp weekday peaks usually need longer windows than a smooth global traffic curve suggests.
Leave room for guardrails
A primary success metric can look healthy while a guardrail — refunds, support tickets, or time-to-first-value — drifts. Size the experiment so those secondary measures have a fair chance to speak.
When waiting is the measurement choice
If the expected window spills into a festival week or a major campaign spike, waiting can be the responsible call. Distorted traffic does not become cleaner because a sprint board says “launch.”
When you want a second pair of eyes on the plan, our Measurement Plan Design engagement walks through these choices with your squad.