Value milestone
A value milestone is a specific user behavior, like sending an invoice, publishing a post, or shipping a commit, that a team uses as a lever to move metrics like conversion or retention. The behavior earns milestone status by holding up under testing: when the team ships changes that get more signups to do it, the business metric moves with it, and that pattern repeats across different attempts.
Definition
A behavioral proxy that functions as a reliable intervention lever — interventions amplifying the behavior produce predictable lift in a commercial outcome event.
Two qualifying conditions
A behavior earns milestone status by satisfying both:
- Scale-tested: the correlation has survived interventions designed to amplify the behavior, not just observational correlation in historical data.
- Intent-aligned: the behavior is anchored to a theorized user intent, not just statistically correlated with the commercial outcome.
The qualifiers do different defensive work. Scale-testing is empirical — accumulated evidence that interventions on the proxy actually move the outcome. Intent-alignment is conceptual — a theory of why the behavior correlates, which gives the team something to check against when correlation moves.
A behavior with strong observational correlation that hasn't been intervened on yet is a candidate behavior, not a milestone. Promotion happens through intervention testing.
How status accumulates
A candidate behavior accumulates milestone status when interventions targeting it pass two checks:
- Lift: the intervention increases the rate at which signups perform the behavior.
- Hold: the conversion rate among performers stays within the metric's natural variation band.
Lift without hold means the team manufactured the behavior without producing the commercial outcome — the proxy didn't survive being intervened on. Lift and hold together means the correlation survived an attempt to scale it.
Milestone status is the accumulated state of repeated lift+hold passing across diverse interventions, not a one-time qualification. A single intervention failure is intervention-quality variance — the experiment may have been weak — and doesn't by itself indict the proxy. Accumulated hold failures across diverse interventions where lift passed is what indicts it; that pattern is proxy failure.
Plural by structure
A user base with multiple intents needs a portfolio of milestones, one per theorized intent — not a singular activation metric. A singular metric averages across intents and loses resolution as the intent mix shifts; per-intent milestones preserve resolution and make drift diagnosable when correlation moves. See portfolio of milestones.
The portfolio doesn't prevent any individual milestone from suffering correlation drift; it makes drift legible by giving each milestone its own theorized referent.
Example
A B2B invoicing tool runs multiplier analysis on its signup cohort and finds that users who send a first invoice within 14 days convert to paid at roughly 8x the cohort baseline. The team's intent theory: these are users who came to bill clients, and the first invoice marks crossing into actually using the product for that purpose. That gives them a candidate behavior — strong observational correlation, theorized intent — but not yet a milestone.
The team ships a redesigned empty state that prompts new users toward sending their first invoice. The intervention lifts the first-invoice rate from 22% to 34% of signups, and the conversion rate among performers stays in the 7-9x range. Lift and hold both passed.
Three more interventions follow over the next two quarters: contextual reminders, a sample-data starter, an onboarding email sequence. Each passes lift+hold within natural variation. After repeated passing across diverse intervention types, "first invoice sent" has accumulated milestone status — the team can now intervene on it expecting predictable lift in first paid conversion.
If a fifth intervention had passed lift but failed hold — say, an aggressive prompt that pushed users into sending throwaway invoices that didn't reflect the underlying intent — that single failure would be intervention-quality variance. Repeated hold failures across diverse interventions would indict the proxy.