Multiplier analysis
Multiplier analysis measures how much more likely users are to convert when they perform a specific behavior, relative to the signup cohort as a whole. Compute the share of all signups who ended up paying (the baseline), then compute the same share among the subset who performed the behavior, and divide. A 6x multiplier means users who did the behavior converted at six times the cohort's overall rate.
Definition
A quantitative method for measuring how strongly a candidate behavior correlates with the commercial outcome event, computed as the ratio between the conversion rate of the signup-cohort subset that performed the behavior and the cohort's baseline conversion rate.
How the multiplier is computed
Compute the baseline conversion rate first — the rate at which the full signup cohort reached the commercial outcome event (for example, first paid conversion within a 30-day window). Then, for each candidate behavior, compute the conversion rate for the subset of the cohort who performed it. The multiplier is the ratio of subset rate to baseline rate. If the baseline cohort converts at 4% and users who performed the candidate behavior convert at 24%, the multiplier is 6x.
The baseline is computed once and reused across all candidates being tested. The exception is platform- or tier-restricted candidates: when a candidate can only fire on web, or only for users on a paid tier, the baseline has to be filtered to match the candidate's restriction. Otherwise the multiplier conflates the behavior's effect with the underlying platform or tier conversion difference, and reads as if it's the behavior when it's actually a composite.
The signup cohort as baseline
Multiplier analysis runs against the signup cohort specifically, because the patterns distinguishing profitable customers from missed conversions do most of their differentiation work in the early lifecycle window. A baseline that spans mature lifecycle stages dilutes the signal across regimes where leverage is structurally smaller — the analysis still produces numbers, but the numbers mix the high-leverage early window with later periods where the same behaviors carry less signal.
Two analytical reads
Each candidate produces two numbers worth interpreting separately:
- Multiplier: how much more likely a user who performed the behavior is to convert, relative to baseline. Strength of signal per user.
- Percent-of-converters: what share of actual converters performed the behavior. Aggregate coverage.
The two diverge meaningfully. A high multiplier with low percent-of-result is a niche lever — strong individual signal but limited reach. A low multiplier with high percent-of-result is widespread but weakly distinguishing, since most non-converters perform it too. A high multiplier with high percent-of-result is both strongly predictive and covering a large share of the conversion population. Consulting only one read produces lopsided investment decisions — chasing niche levers that can't scale, or optimizing behaviors that don't actually distinguish converters from non-converters.
Where candidate behaviors come from
Candidates come primarily from offering review: what marketing promises, what onboarding asks users to do, what customers describe doing in reviews and forums, what the team's hunches are. The analytics tool's event catalog is a legitimate technical inventory of what's instrumentable, but a candidate list built from the catalog alone produces a ranking of events whose connection to user intent is unknown — events get instrumented for engineering convenience, debugging, and attribution reasons unrelated to user intent.
Retracing is the complementary generative method: rather than testing a pre-specified list, retracing starts with the converter population and surfaces patterns inductively from their event logs. Multiplier analysis then confirms and quantifies the candidates retracing surfaces.
Example
A SaaS offering measures first paid conversion within 30 days as its commercial outcome event. Baseline conversion rate across the signup cohort is 4.2%. The team tests several candidate behaviors sourced from onboarding-flow review and customer interviews.
"Imported existing data within 7 days" converts at 24.4% — a multiplier of 5.8x, with 38% of all conversions coming from users who did this. "Invited a teammate within 7 days" converts at 13.0% — a multiplier of 3.1x, with percent-of-result at 19%. "Viewed the pricing page" converts at 5.9% — a multiplier of 1.4x with percent-of-result at 67%, widespread but weakly distinguishing because 67% of converters view pricing and nearly half of non-converters do too.
The output is a ranked spreadsheet with both reads visible against the baseline. Strong-multiplier candidates with meaningful coverage become candidates for value milestone qualification via scale verification.