Incrementing vs. Iterating
Incrementing and iterating are two ways a team can ship a series of experiments, distinguished by whether each test responds to what the last one revealed. When the tests respond, they compound and the team is iterating; when they don't, the team is incrementing, even if it ships something daily.
Definition
The diagnostic that distinguishes iterating — where each move's strategy is shaped by what the previous move's performance revealed, so moves compound into a learning trajectory — from incrementing, where moves ship without compounding because each was decided independently of the last.
The diagnostic test
Path investment specifies iterating as the working mode: each move small, leveraged for learning, its design conceived in response to what the previous move surfaced. Incrementing vs. iterating is the check on whether a team is actually working that way or just looks like it.
The test runs at one moment — when the team strategizes move N+1. Does that strategy respond to what move N's performance revealed? If yes, the team is iterating. If no, it's incrementing.
The "if no" splits into two distinct causes, and naming them is how the diagnostic does its work. A move can fail to respond either because its strategy was fixed before the last move produced any signal, or because it shares no thread with the last move at all.
Two shapes of incrementing
- Plan-execution. The strategy for the move predates the previous move's findings. The team decided the full sequence in advance and is now executing it. Later moves can't respond to earlier signal because they were specced before the signal existed.
- Drift. The move comes off an unconnected backlog — a discrete fix, an ad-hoc request — that shares no learning thread with what came before. Nothing compounds because nothing connects.
Both produce small moves that ship. Neither produces a learning trajectory. The output looks like iteration from the outside and isn't.
Cadence and iteration are independent
A team can ship daily and still be incrementing. Cadence describes how often moves go out; iteration describes whether each move answers the last. Two teams shipping at identical frequency can sit on opposite sides of the line — one's moves respond to the prior move's signal, the other's don't.
This is where the diagnostic earns its keep, because two common slips dress fast shipping up as iteration. A team decomposes a big-picture plan into stages and labels the stages iterations (cadence-as-iteration). Or a team works through a backlog of discrete fixes release by release and calls the cadence iteration (shipping-as-iteration). Both ship steadily; neither compounds.
The same test applies wherever a team builds on a model that evidence can correct — paths, affordances, intent theories, value milestones, pricing. At every layer, iterating is what lets a correction at move N reshape move N+1.
Example
Two teams each ship an onboarding change every week.
Team A instruments intent detection in week one and finds that signups pursuing one intent — connecting a data source first — convert at twice the base rate. Week two's move responds: a copy change that surfaces the data-source step earlier for that segment. Week three responds to week two's lift, deepening the segment's path or abandoning it on weak signal. Each move's strategy answers what the last move revealed.
Team B built a ten-week onboarding roadmap last quarter and ships one stage per week, calling each a sprint iteration. Week six's stage was specced in the roadmap; it doesn't respond to anything weeks one through five surfaced. Same cadence as Team A, zero compounding.
Run the diagnostic at the moment each team strategizes its next move. Team A's next strategy responds to the last move's performance — iterating. Team B's next strategy was fixed before the last move shipped — incrementing.