STARTIFY — Product Strategy Evaluation Framework
How STARTIFY's eight weighted dimensions decide "what deserves to exist": the frameworks it sequences, how it's scored and read, and what the review against the PM canon changed.
STARTIFY is an eight-dimension product-strategy evaluation model created by you. It chains internal prioritization frameworks and external strategy frameworks into one sequence for deciding what deserves to exist. It's an upstream "should this exist" filter: narrative first, scored second, sitting beside RICE and WSJF rather than replacing them. Earlier name: STRATA. STARTIFY = start + stratify.
It didn't invent new ideas. It organized proven ideas into a better sequence.
STARTIFY is a framework you made up, with no data behind it yet. In AI Agent for PM it runs in shadow only (v1.5): it scores features but never ranks them, until shadow results over several cycles show it adds something (decided 2026-11-18).
The five layers and the frameworks behind them
The order is the contribution: internal value → market direction → strategic position → compounding advantage → sustainability. Narrative comes first, then the five layers, with weighted scoring at the end to summarize, never replace, the thinking.
Layer 1 · Internal value
What should we build, and in what order?
- RICE = (Reach × Impact × Confidence) ÷ Effort. Impact uses a multiplier scale, from 3 (massive) down to 0.25 (minimal).
- WSJF = Cost of Delay ÷ Job Size, where Cost of Delay = user/business value + time criticality + risk reduction/opportunity enablement. "RICE's cousin."
- Kano sorts features by the shape of their satisfaction curve: Basic, Performance, Delighter, Indifferent, Reverse. A roadmap balances the three productive types.
All three assume the thing is worth building and don't question market direction or defensibility.
Layer 2 · Market direction
Is the market moving this way?
- Gartner: the Hype Cycle (five stages) and the Magic Quadrant (completeness of vision vs. ability to execute) gauge category maturity.
- Forrester (the Wave) leans toward CX and customer-centric analysis.
Layer 3 · Strategic position
Competing better, or competing differently?
- Blue Ocean: value innovation (differentiation and low cost) and the Four Actions Framework: Eliminate, Reduce, Raise, Create.
- Play Bigger: Category Design and the "Category King". Whoever defines the problem defines the solution.
Layer 4 · Compounding advantage
Does it get stronger with use?
Competing in the Age of AI: data flywheels compound while traditional advantage erodes. This layer made the A dimension non-optional rather than a bonus category.
Layer 5 · Sustainability
Can we keep winning after we've won?
Porter's Five Forces (1979) is STARTIFY's final reality check.
Context, not lineage
Maslow's hierarchy
Maslow (1943: physiological → safety → belonging → esteem → self-actualization) is the conceptual ancestor of Kano's tiered needs. Kano is empirically grounded (the Kano questionnaire), while Maslow is weakly supported and mostly used as a heuristic.
The eight dimensions
| Letter | Dimension | Question | Drawn from | Category | Weight |
|---|---|---|---|---|---|
| S | Strategic Differentiation | Value competitors can't easily replicate? | Blue Ocean, Play Bigger | Foundational Leverage | 1.3 |
| T | Trend & Category Momentum | Aligned with or shaping an emerging market? | Gartner, Forrester | Momentum Builders | 1.2 |
| A | AI / Data Flywheel Reinforcement | Gets better the more it's used? | Competing in the Age of AI | Momentum Builders | 1.1 |
| R | Relevance to Customer Experience | Would customers notice if it disappeared? | Kano (loosely Maslow) | Core Validation | 1.0 |
| T | Time-to-Impact | How fast to validate / see results? | RICE, WSJF | Operational Execution | 0.9 |
| I | Industry & Analyst Validation | Do respected outside voices confirm? | Gartner, Forrester | Core Validation | 1.0 |
| F | Force Field Stability | Survives competitors, suppliers, substitutes? | Porter's Five Forces | Momentum Builders | 1.1 |
| Y | Yield & Sustainability | Still creating value years after launch? | Competing in the Age of AI, Porter | Foundational Leverage | 1.2 |
| Total weight → maximum score on a 0–5 scale is 8.8 × 5 = 44 | 8.8 | ||||
Y's lineage comes from the frameworks doc's lineage table. The article left it unnamed. Weights are unchanged in v2 and v3.
Weighting philosophy: strategic gravity
Weights measure strategic gravity, not importance. The source doc also calls this strategic inertia.
| Component | What it asks |
|---|---|
| Strategic half-life | How long does the decision keep creating value? |
| Path dependency | How hard is it to recover if it's wrong? |
| Strategic leverage | Does it amplify every other investment? |
| Category | Dimensions | Weights |
|---|---|---|
| Foundational Leverage | S, Y | 1.3, 1.2 |
| Momentum Builders | T-momentum, A, F | 1.2, 1.1, 1.1 |
| Core Validation | R, I | 1.0, 1.0 |
| Operational Execution | Time-to-Impact | 0.9 |
Scoring
Rubric (0–5)
| 0 | Not addressed, or contradicts strategy |
| 1 | Very weak evidence |
| 2 | Limited strategic value |
| 3 | Solid but unremarkable |
| 4 | Strong strategic contribution |
| 5 | Exceptional strategic advantage |
R evidence anchors (v3)
| 1 | Internal belief |
| 2 | Stated interest |
| 3 | Problem confirmed via past behavior |
| 4 | Observed behavior (time, money, workarounds) |
| 5 | Commitment to the solution (payment, pilot, LOI, repeated use) |
From The Mom Test and Continuous Discovery Habits.
S evidence anchors (v3)
First name the alternative customers use today.
| 2 | Parity features |
| 3 | Confirmed but replicable within a year |
| 4 | Hard-to-copy advantage with switching rationale |
| 5 | Changes the comparison |
From Obviously Awesome.
R and S are the dimensions most easily scored on conviction. Score the highest level the evidence supports, and any 4 or 5 on R or S must name its evidence.
Stage lenses (v3)
Every stage scores with the default weights (sum 8.8, max 44). What changes by stage is which dimensions the discussion challenges hardest.
| Stage | Push hardest on | Hold loosely |
|---|---|---|
| Startup | R, S | F, Y |
| Growth | A, F, Y | Time-to-Impact, I |
| Mature enterprise | F, Y | Time-to-Impact, T-momentum |
v1 and v2 adjusted weights by ±0.1 to ±0.2. A 0.1 change moves the max by 0.5/44 (about 1%), which is less than one point of scorer disagreement (about 1.0), so the profiles couldn't change a decision.
History: v1's growth and enterprise profiles pushed the max to 45.5 and 45. v2 rebalanced them to 8.8, and v3 dropped the weight math. Lenses were chosen over bigger shifts (±0.3 or more) because they fit narrative-first, and they're easy to reverse.
Reading the score
| Score | Class | Meaning |
|---|---|---|
| 80–100% | Prime Category Feature | Strong differentiation, sustainable advantage, excellent timing |
| 60–79% | Differentiated Play | Good opportunity, moderate strategic or execution risk |
| 40–59% | Incremental Value | Valuable optimization, unlikely to reshape the market |
| < 40% | Tactical Improvement | Limited strategic impact; redesign or deprioritize |
The v1 scoring doc totaled the example as 36.6 / 40 (91.5%). v2 corrects it to 36.6 / 44 (83.2%), which now agrees with the article and the project notes. Early docs gave a max of 40 and inflated the example percentages.
Worked example: AI personalized checkout
The useful signal was the weak lines, not the total. Time-to-Impact is a delivery problem, and Force Field Stability is a moat problem.
Review against the PM canon (2026-10-19)
A review of STARTIFY against the books in your PM reading list. Two changes were adopted and are now in v3. The unit-of-analysis question was decided with no change, and the category-creator question is still open. The book characterizations come from Claude's general knowledge, since none of the books is in the wiki yet, so treat them as medium confidence. The arithmetic is exact.
Where the canon supports it
- Good Strategy Bad Strategy (Rumelt): strategy starts from diagnosis, not a checklist, matching "the score summarizes the conversation". A candidate reference.
- Crossing the Chasm (Moore): supports scoring Trend & Category Momentum as timing.
- Blue Ocean Strategy, Play Bigger, Porter: support the long-horizon weight order (S and Y on top).
- The Lean Startup disagrees with Time-to-Impact's low weight (0.9). "Speed is recoverable, positioning isn't" is a defensible disagreement, not a flaw.
Deliberately left unchanged
- The eight dimensions and the weight order.
- No effort/cost dimension. RICE and WSJF cover effort downstream.
- Cagan's usability and feasibility risks (Inspired) are downstream of STARTIFY's job.
- Side-by-side alternatives (Decisive): the AI Agent for PM pipeline already scores many items.
- The duplicate T and Maslow's exclusion.
Adopted in v3
- Stage lenses replace adaptive weight shifts.
- Evidence anchors for R and S.
Several first-pass critiques graded STARTIFY against jobs it isn't meant to do, and they were withdrawn.
Question 1: what's the unit of analysis? Decided: no change
| Level | Checkout example | What it is |
|---|---|---|
| 1. Bet | "Win repeat buyers through AI-personalized commerce" | A market position you choose to take |
| 2. Opportunity | "Repeat buyers abandon checkout because re-entering preferences is tedious" | A problem worth solving (Problem Definition page) |
| 3. Solution | "One-click reorder" (a MoSCoW subfeature) | One way to solve it |
The pipeline's scoring unit is the Problem Definition Agent's output, a defined problem with MoSCoW subfeatures under it, so STARTIFY already scores at the opportunity level. The first pass overstated this as needing a restructure.
- Bets: all eight dimensions work natively.
- Opportunities: five work. I, F and Y mostly describe the company, which is where a halo can form.
- Solutions: almost everything is inherited. Leave this level to RICE, WSJF and Kano.
| Dimension | Bet | Opportunity | Solution |
|---|---|---|---|
| S: Differentiation | native | meaningful | mostly inherited |
| T: Trend Momentum | native | meaningful | inherited |
| A: Flywheel | native | meaningful | sometimes |
| R: Customer Relevance | native | native | native |
| Time-to-Impact | native | native | native |
| I: Analyst Validation | native | inherited | inherited |
| F: Force Field | native | inherited | inherited |
| Y: Yield | native | partly | inherited |
Options considered, not adopted: a contribution rule ("score the contribution, not the context", where 3 means neutral), a sizing test ("would this change how customers compare us to their alternative, or only what's in our backlog?") and a band rename to something like "Prime Category Initiative". Revisit them if opportunity scores in the pipeline bunch at the top.
Question 2: should STARTIFY favor category creators? Open
| S | T | A | R | Time | I | F | Y | Total | |
|---|---|---|---|---|---|---|---|---|---|
| Category creator | 5 | 3 | 5 | 5 | 2 | 1 | 3 | 5 | 32.7 |
| Trend follower | 3 | 5 | 3 | 4 | 4 | 5 | 3 | 3 | 32.7 |
They tie at 74% (Differentiated Play). Trend & Category Momentum and Industry & Analyst Validation are 2.2 of 8.8 weight (25%), both drawn from Gartner and Forrester. T is mostly fine, since its question says "aligned with or shaping". The problem is I, which can't see an unnamed category. This extends the known S-vs-I tension and cuts against the category-leader mindset STARTIFY came from.
- Option 1, stay neutral: keep I as a hubris check. Most category-creation claims are wrong.
- Option 2, redefine I as category evidence: "Is there evidence this category will exist?" Analysts become one source among several (adjacent spending, customer workarounds, investor activity, new job titles, search demand). The review leans this way.
- Option 3, exempt I at the startup stage: no longer works, since stage lenses don't change the math.
Who is STARTIFY for: you and teams you lead, or any team? If yours, option 2 is clearly right. If general, keeping I as is has a fair case.
Using it
Narrative first
- Write the strategic narrative
- Challenge assumptions
- Gather evidence
- Score
- Discuss disagreements, especially where score and story diverge
- Decide
The score summarizes the conversation. If it replaces the conversation, the framework has failed.
Where it sits
| STARTIFY | Whether something deserves to exist |
| RADAR | Orders what passes, and shows which teams it affects |
| PM decision framework | Making and defending a call, including the risk and exit criteria STARTIFY doesn't cover |
In practice: AI Agent for PM
- v1.5 (2026-11-18): shadow only. The STARTIFY Agent still scores every feature, but its score no longer ranks features or reaches the Synthesis Agent or Review. A per-cycle report compares its ranking with your final order and with 60-day verdicts.
- v1.2: the STARTIFY Agent proposes the eight dimension scores after the narrative, and code computes the weighted total over the fixed 44. Open: whether the other scorers feed STARTIFY's dimensions or run beside them.
- v1.2–v1.4: one of four ranking agents. A dedicated Strategic Narrative Agent drafts the strategic case first, so narrative stays before scoring.
- A standing Competitive/Market Scan Agent feeds T and I with live research through the Strategic Narrative Agent.
- The Synthesis Agent flags score-vs-narrative disagreements rather than averaging them away.
How it was built
Developed in a dedicated Claude project, which produced a framework document, a source-frameworks document and the Medium article. The work went foundation first, then layered in worked examples, references and visualized tables over successive passes. The motivation was your "category-leader-first" mindset, influenced by Play Bigger. A separate origin doc, The Story Behind STARTIFY, isn't in the wiki yet.
Known design tensions and open questions
- Duplicate "T": Trend & Category Momentum and Time-to-Impact share a letter.
- Differentiation vs. validation: in category-creation cases, high S can come with low I, because analysts haven't named the category yet (see question 2).
- Maslow: appears only as sidebar context. Making it part of the lineage would be a deliberate design decision. The frameworks doc is inconsistent: its lineage table cites "loosely, Maslow" for R, while its scope note excludes Maslow.
- Small wording slip in the frameworks doc: from Layer Two, Porter is called "two layers up", but it's Layer Five, three layers above market direction.
- The leverage / gravity / inertia terminology.
- Maslow's status.
- Whether STARTIFY should favor category creators.
- The Story Behind STARTIFY hasn't been captured into the wiki yet.