Seven Lenses to Kill a Feature: Observation Haki as a Prioritization Tool

Most PRD frameworks are broad and shallow. Observation Haki is narrow and deep — seven lenses, each forcing one question you’d rather skip.

What it is

Observation Haki lives inside Shaka, the PRD satellite in Stella Protocol — my open-source AI-PM methodology. Shaka is one of Vegapunk’s six satellite bodies in One Piece — the analytical one. Haki is a perception ability; characters with Observation Haki sense things others miss.

The metaphor is load-bearing. The lens pass isn’t about “covering all bases.” It forces perception where intuition has blind spots. Each lens is one sharp question you’d prefer to answer vaguely, and the lens refuses the vague answer.

Why seven

I tried shorter cuts. A 3-lens version (Pain / Shape / Fog) works for ideation but leaves scope and architecture unanchored. A 5-lens version (add Victory + Skeleton) closes scope but ships products with broken first screens because you never thought about the first pixel.

I tried longer. A 10-lens version turned into ceremony — by lens 8 I was restating lens 3 in different words.

After four shipped projects, seven was the empirical floor that caught blind spots without diluting focus. Not a magic number, just what held up.

The seven lenses

Each lens: name → one-sentence definition → signature question → concrete example from a real project.

1. Pain

Definition: The specific friction or cost the user feels today without your product.

Question: What is the user doing right now, in the absence of your product, that hurts?

Example (Stoka): User walks into the kitchen at 6pm, stares at the open fridge, can’t decide what to make, orders takeout. Cost: ~$15, 30-min wait. Frequency: 3–5 times a week.

Not “users waste food.” That’s a slogan. The lens demands specific action, specific cost, specific frequency.

2. Victory

Definition: What “won” looks like — as an observable state, not a vague goal.

Question: What exact behavior would I see a successful user doing?

Example (Amal Najib, a compliance tool I built): A halal counselor publishes three assessment question flows in a single week without asking a developer for help. That’s Victory.

“Counselors are empowered” is not Victory. It’s a slide deck. If you can’t point a camera at a successful user and film Victory happening, the lens isn’t done.

3. Boundary

Definition: What you are explicitly not building.

Question: What would a reasonable person assume is in scope that actually isn’t?

Example (House of Riddle V1, a riddle game): No social feed. No DMs. No streaks. No friend list. We are a riddle game, not a social network.

This is the lens where most PRDs lie by omission. If a feature isn’t mentioned, people assume it’s coming. Boundary forces you to name the ghosts in the room and banish them in writing.

4. Shape

Definition: The single feature that is non-negotiable for the user to hit the AHA moment.

Question: Which one feature, if removed, means the first user never feels the “oh, this is useful” feeling?

Example (Stoka): AI meal suggestions. This is the lens that killed receipt OCR in P0 — full story in the receipt scanning post. Short version: my instinct said OCR (impressive, automated). The honest answer was recipes (actual AHA). Lens 4 caught the gap.

Shape is the most dangerous lens because the intuitive answer is almost always wrong. Your instinct names the coolest feature; the lens wants the AHA feature. Rarely the same.

5. Skeleton

Definition: The minimum architecture that supports Shape and Victory.

Question: What’s the most boring stack that makes this work?

Example (Stoka): Next.js App Router + Supabase + Dexie.js. No microservices. No event bus. No queue. No Redis. A single database table for items, a single table for users, one Vercel deployment.

If your Skeleton answer has more than five nouns in it, you’re probably over-engineering. The lens rewards boring.

6. Surface

Definition: What the user sees first — the actual pixel real estate at T=0.

Question: What is on screen the first time a user opens the app, before they do anything?

Example (Stoka onboarding): Three screens, each skippable. Pre-filled with five sample items (milk, eggs, tomato, chicken, bread) so if the user skips to recipes, the AI has input and returns something useful. No empty state. No “add your first item” dead end.

Surface forces commitment to the empty state, onboarding, home screen. PRDs that skip this lens ship a beautiful feature list and a broken first screen.

7. Fog

Definition: What you don’t know yet, and the cheapest way to find out.

Question: What assumption am I making that I’d pay money to verify before building the feature?

Example (Stoka): Will users grant camera permission to photograph items? That’s fog — I didn’t know. The expensive way to find out is to build the camera feature and measure permission grant rate. The cheap way: ship without a camera, add a “photo this item” button that opens a feedback form, and measure clicks. Demand signal for ~0 engineering cost.

Fog is where the 7-lens pass earns its keep. Every unknown you name gets a cheap experiment. The ones you don’t name become expensive surprises in month two.

How to run them

Not all at once. Sequential, each one a short note — 3 to 5 bullets, not an essay.

• Pain → Victory: establish why. If you can’t fill these two, stop. There is no product.
• Boundary + Shape: establish what. Boundary names the no’s; Shape names the single must-have yes.
• Skeleton: establish how. Boring stack. If it’s exciting, rewrite it.
• Surface + Fog: establish unknowns. What does the user see on day one; what do you still need to learn?

The whole pass takes 60–90 minutes for a fresh product. Faster for a refinement. Slower if you’re honest and Lens 4 makes you delete half your PRD.

Why not Lean Canvas, RICE, or Story Mapping?

Each is good at something the lenses aren’t. None force the questions the lenses force.

• Lean Canvas — nine boxes, broad and shallow. Excellent for communicating a business hypothesis. No forcing function for scope. Nothing asks “which single feature defines AHA?”
• RICE scoring — prioritizes candidate features against each other. Useful when you already have the right list. Doesn’t question whether items belong on the list. You can RICE-rank wrong features and feel productive.
• User Story Mapping — great for sequencing and visualizing a journey. Answers “in what order?” not “should we build this at all?”

Observation Haki is narrower. It assumes you have ideas and need to kill most of them. It doesn’t replace those frameworks; it sits upstream. Run the 7 lenses, then RICE the survivors.

When it’s overkill

Stella Protocol has two tracks: Grand Line (new products, long horizon) and East Blue (hotfixes). Observation Haki is a Grand Line tool.

For an East Blue hotfix — “recipe card overflows on iPhone SE,” “expiry sort is reversed” — seven lenses is ceremony. Scope is already tiny; no AHA to relocate, no Fog to de-risk. A two-line diff and a screenshot is the artifact.

Using the lenses on trivial work burns trust in the process. Reserve them for work where shipping the wrong thing costs weeks, not hours.

Lesson

Broad frameworks feel thorough but let noise through. Seven forced viewpoints beat twenty vague ones. The lenses work because each one refuses the vague answer — and the vague answer is the one your instinct wants to give.

Key Takeaways

1. The lens only works if it refuses the vague answer. “Users waste food” is not Pain. “Counselors are empowered” is not Victory. If the answer could fit in a slide deck, the lens isn’t done with you.
2. The coolest feature is rarely the AHA feature. Shape (Lens 4) is the most dangerous lens because the intuitive answer usually points to what’s impressive, not what brings users back on day 2.
3. These lenses sit upstream of RICE and Lean Canvas, not alongside. Run Observation Haki to decide what belongs on the list. Then RICE or story-map the survivors. Using RICE on the wrong list is how you feel productive while building the wrong product.

Satellite: Shaka (PRD) · Morgans (this post) · Pipeline: DISCOVERY — Observation Haki → Morgans