Module 2: Your First OST

Build an Opportunity Solution Tree from a product brief and run a tree health check on a deliberately broken sample tree.

Time: 1 day Coding required: No. Prerequisite: Module 1 (working environment) and a product brief — the ShiftLoop sample brief for practice, or your own product’s brief if you’re going straight to the real thing (see the bring-your-own-product note below). You will produce: An Opportunity Solution Tree rooted in ShiftLoop’s activation outcome, plus a health check that catches the planted flaws in the provided seed tree.

Why this module

In Module 0 you traced one feature backward to an outcome. An Opportunity Solution Tree is that trace, scaled up and made permanent: one outcome at the root, the customer needs (opportunities) that block it, and the solutions you might build under each. It is the operating system of agentic PM — the single artifact every other activity reads from and writes to. Signals map onto it (Module 3), experiments hang off it (Module 4), and roadmap items that don’t trace to it don’t get built.

Get the structure right and everything downstream stays honest. Get it wrong — a solution smuggled in as an opportunity, two outcomes fighting at the root — and every decision built on it inherits the error. So this module is as much about detecting bad structure as building good structure.

Learning objectives

By the end of this module you will be able to:

Root a tree in a single, measurable, behavior-based desired outcome.
Frame opportunities as customer needs, and recognize a solution wearing an opportunity’s clothing.
Use the ost-workflow skill to build and extend a tree.
Run a tree health check and name which of the seven checks a tree fails.
Pick a focus branch using evidence-weighted prioritization instead of gut feel.

Concept reading

The four layers

An OST has exactly four layers, all rooted in one desired outcome:

Desired Outcome  (exactly 1 — a measurable change in customer behavior)
  └── Opportunity (N — unmet customer needs, pains, desires)
        └── Solution (N — ways to address the opportunity)
              └── Experiment (N — a test of one assumption inside one solution)

The rules that are non-negotiable

One tree = one desired outcome. Two outcomes means two trees. A second outcome creeping in at the opportunity layer is the most common way trees rot.
Opportunities are customer-centric, never company-centric. “Customers struggle to get their team loaded” is an opportunity. “Increase conversion” is your goal, not their need.
Solutions never sit at the opportunity layer. If a branch names a thing you’d build (“mobile app,” “redesign the wizard”), it’s a solution. Reframe it to the need it serves, then put the build under that need.
Every experiment tests a specific assumption inside a specific solution. (You’ll do this in Module 4.)

Frame opportunities like this

“Customers struggle to [X] when [context]” — or — “Customers need [X] but currently [workaround/gap].”

For ShiftLoop: “Managers struggle to get their existing team into the tool quickly, because their staff list already lives in a spreadsheet they have to re-enter by hand.” That’s a need with context — not a feature.

The health check (memorize this table)

Check	Red flag
Outcome clarity	Root is output-framed, vague, or unmeasurable
Opportunity framing	A branch sounds like a solution or a company goal
Coverage	Fewer than ~5 distinct opportunities
Evidence	Most opportunities have no attributed research
Focus	Team is actively working 3+ branches at once
Experiment velocity	No experiments closed in the last 2 weeks
Dead ideas	Abandoned paths still shown as active instead of archived

3+ red flags = the tree needs a reset before any new work.

Go deeper: ost-workflow skill (the full build + health-check + prioritization procedure), Agentic PM Playbook — OST as Operating System, and Tree Health Checks.

Hands-on exercise

Two parts: build a clean tree, then diagnose a broken one. The diagnosis half is where the skill of reading structure actually develops.

🎯 Doing this on your own product? The exercise below is written around the ShiftLoop sample so everyone practices on the same material. To run it on your real product instead, make three swaps: (1) open your product brief instead of ShiftLoop’s; (2) use your desired outcome as the root (the one in your pm-config.md) instead of 38% → 60%; (3) draw opportunities from your personas — however many you have. A single-persona B2C product has no “buyer/user split,” so all opportunities come from your one user; that’s expected, not a gap. Part B (the seed-OST health check) stays the same — it’s a shared diagnostic drill. Note that a tree you build on your own product here won’t line up with the ShiftLoop signals in Module 3; that’s fine — Module 3 has its own bring-your-own-signals note.

Part A — Build a tree from the brief (≈ 45 min)

Setup: Open the product brief. Do not open the seed OST or the signals yet — build from the brief alone so Part B and Module 3 have something to react to.

Steps:

In a thread, trigger the skill: “Build an OST for <product>. The desired outcome is in the product brief.” (Sample run: “Build an OST for ShiftLoop…”)

No ost-workflow skill? If you’re working without the skill active, apply the four-layer structure directly: describe your product and desired outcome in a plain Claude thread, then ask it to help you surface customer needs (opportunities) and candidate solutions. The skill enforces good structure automatically — without it, you’re the enforcer. Re-read the four-layer rules and the health-check table above and check your own work at each step.
Confirm the root with the skill: it must be your activation/desired outcome (for ShiftLoop, 38% → 60%), behavior-based and measurable. If you phrase it as a feature, let the skill reframe you — that’s the lesson.
Surface at least 5 opportunities, each framed as a customer need with context. Pull them from your personas — for ShiftLoop that’s managers, the buyer, and staff; for a single-persona product it’s all from that one user. Resist writing solutions.
Under your two strongest opportunities, generate 3 solution directions each (the skill will push for at least three, spanning incremental to transformative). Don’t evaluate them yet.
Save the tree where your notes live (per pm-config.md).

Deliverable: a saved OST — one outcome, ≥5 need-framed opportunities, solutions under the top two.

Part B — Health-check the seed tree (≈ 45 min)

Setup: Now open the seed OST. It is deliberately broken. Do not open the facilitator key yet.