Candidate behavior
A candidate behavior is a specific action users take after signing up, like inviting a teammate, finishing onboarding, or using the search bar within the first hour, that the team suspects might relate to whether those users eventually convert. The status is provisional: plausible enough to warrant measurement and testing, not yet confirmed as a real driver of the outcome.
Definition
A user behavior under hypothesis-testing as a possible correlate of the commercial outcome event.
Status between speculation and verified leverage
Candidate status names a specific epistemic position. The behavior is plausible enough to warrant quantification — someone has reason to think it might be doing work — but the correlation hasn't been confirmed through intervention. Teams typically hold many candidates in evaluation at once, with different candidates at different points in the lifecycle.
Candidate status is provisional; behaviors move out of it in one of two directions. A candidate either graduates to value milestone status through accumulated scale verification, or gets indicted as a non-lever through proxy failure.
The three lifecycle stages
A candidate behavior moves through three stages:
- Sourcing: where the candidate comes from. Either hypothesis sourcing (marketing-copy reading, onboarding-flow reading, feature-list reading, review reading, team hunches) or retracing (reading converter event logs to surface patterns inductively).
- Quantification: testing whether the correlation exists at the cohort level. Multiplier analysis is the primary operation — computing the conversion rate for users who performed the behavior against the baseline cohort rate.
- Verification: testing whether the correlation transfers under intervention. Scale verification tracks whether interventions designed to amplify the behavior produce predictable lift in the commercial outcome event.
Each stage gates the next. A candidate without sourcing has no interpretive grounding for why it should be tested. A candidate without quantification is a hunch with no measured strength. A candidate without verification is observational correlation that hasn't been tested as a lever.
Example
A team running a project-management SaaS reads through recent G2 reviews and notices several users describing how they invite their team before kicking off a project. They convert this into a candidate behavior: "invited a teammate within 3 days of signup." Multiplier analysis on the signup cohort shows users who performed the behavior converted to paid at 4.2x the baseline rate. The candidate is now quantified. The team designs an intervention — a contextual prompt during project creation — to amplify the behavior. Across the next two cohorts, the share of signups inviting a teammate rises (lift passes) and the conversion rate among performers holds within variation (hold passes). Repeated lift+hold across diverse interventions is what would accumulate into value milestone status.