Path investment
Path investment is how a team decides how much to put into building out a particular path. The principle is to start with the smallest, cheapest move that would produce signal about whether a richer version would work, then let that signal decide whether to deepen, drop, or try a different angle. The point is to avoid pouring resources into a fully-built path before evidence has shown the depth will pay off.
Definition
Path investment is the discipline of building paths in small, leveraged moves whose depth escalates only when prior moves produce evidence that deeper investment will be rewarded.
Depth escalates on evidence
Each move into a path is small and leveraged for both impact and learning. The previous move's signal decides whether the next move happens at all, and shapes what it looks like. This inverts the more common pattern of designing a full path up front, shipping it, and finding out later whether the investment was warranted.
The form of a move varies. A move might be pure segmentation (instrument the detection, observe distribution and conversion, change nothing else). It might be a copy variant, a reordered step, an intent-specific email, a painted-door CTA, a differentiated first-action prompt. What's constant across forms is the ratio: low investment, fast feedback, a clear decision point about whether to evolve.
The colloquial name is "baby pokemon" — start unevolved, evolve based on what happens.
The first move is detection only
Detection is reliably the first move. The team adds intent detection — typically via a sorting hat — and instruments it. Every segment still routes to the OSFA experience. The output is data on how intent is distributed across the user base and how conversion performance varies by segment.
Nothing about the offering changes yet, which makes this the cheapest move available. It's also what makes everything downstream possible: differentiated investment without per-segment conversion data has no basis on which to direct itself.
The discriminating question
Advancing to deeper investment requires evidence that deeper investment will pay off. The discriminating question after any move is whether this move produced that evidence.
An intervention producing strong intent-matched lift is evidence the segment is real and responds to the lever the intervention varied — signal to deepen along that lever. An intervention producing no lift is evidence either that the segment isn't real (revisit the segmentation), or that the lever varied isn't what the segment responds to (reason backward from intent and the critical route to identify a different lowest-investment test before deepening). Either way, the next move's design responds to what the previous move surfaced; teams don't pre-plan the sequence at the start of path-building.
Example
A SaaS team suspects its signup cohort contains two distinct intents: people setting up tracking for their own work, and people trying to coordinate a small team. The first move is to add a single onboarding question that classifies the two — pure detection, no differentiation. Every signup still goes through OSFA. Three weeks of data shows team-coordinators converting to paid at half the rate of solo trackers.
That's the signal to escalate, but not all the way. The next move is the cheapest meaningful test of what team-coordinators need that OSFA doesn't deliver — a differentiated first-action prompt that asks them to invite a teammate before they log their own work. If that moves team-coordinator conversion meaningfully, the team escalates further along that lever, perhaps toward a fuller differentiated onboarding flow. If it doesn't move conversion, the team reasons backward from team-coordinator intent again to identify a different cheap test before committing to anything bigger.