Build Buy Partner

Build / Buy / Partner

Three options, one scoring matrix, then a verdict. The matrix comes first because a recommendation written before the scoring is a preference wearing a framework.

Verification gates (defined first; output is blocked until all pass)

• G1 — Same axes, all three, matrix first: build, buy, and partner are each scored on the same named axes, and the full matrix appears before the recommendation. A skipped option, an option scored on fewer axes, or a verdict-first draft fails the gate.
• G2 — Scores carry bases: every cell is a comparative label (strong/moderate/weak) with a one-line basis. Hard numbers appear only from the input or as labeled estimates with derivations — no invented vendor pricing or build-cost figures.
• G3 — Decision cites its differentials: the recommendation names which axis differences decided it, and a kill-the-recommendation line states what fact would flip it.

Steps

1. Define the capability and its role. Core differentiation or adjacent plumbing? This classification drives the differentiation axis and must be argued, not assumed — customer-visible ≠ core.
2. Fix the axes. Default five: time to capability · total cost class (build + maintain vs per-unit/licence vs partner rev-share) · differentiation value · risk (compliance, vendor lock-in, partner dependency — name which) · reversibility. Add a context axis if the input demands it (e.g. data residency); every axis applies to all three options.
3. Score the matrix. All three options, every axis, label + basis. Use input numbers where they exist (volume × unit price = run-rate, tagged to input). Where an option is degenerate (no partner motion exists), score it honestly weak with that basis — don't silently drop it.
4. Read the matrix. The verdict comes from axis differentials, stated: which columns win where, and which differences carry the decision weight given the context (a 12-engineer team weights time-to-capability heavier than unit cost).
5. Write the reversal. What fact flips the call — volume crossover (shown as labeled estimate), the capability becoming core, a partner motion materializing. Include the second-best option and the trigger for revisiting.
6. Gate pass. Matrix complete and first (G1), every cell based (G2), differentials + kill line present (G3). Fix and re-run; maximum 2 repair loops, then report the failure instead of the output.

Output format

CAPABILITY: e-signature in proposal flow — classified: adjacent (customer-asked, not differentiating)
MATRIX (before verdict)
| Axis                | BUILD           | BUY                    | PARTNER           |
| time to capability  | weak (quarters) | strong (weeks, API)    | moderate (bizdev) |
| cost class          | weak (eng + maintain + compliance) | strong ($1,000/mo run-rate [input: $0.50 × 2,000]) | moderate (rev-share) |
| differentiation     | weak (adjacent) | neutral                | neutral           |
| risk                | compliance burden ours (ESIGN/eIDAS) | vendor lock-in, low | dependency, no existing motion |
| reversibility       | weak (sunk)     | strong (swap vendors)  | weak (contractual)|
RECOMMENDATION: BUY — decided by time-to-capability and compliance-risk differentials;
build's only win is unit cost at volumes we don't have.
WOULD FLIP IT: e-sign becomes core differentiation, or volume grows past the build
crossover [ESTIMATE: order-of-magnitude above 2,000/mo — derive before acting].
SECOND BEST: partner, if a motion materializes. Revisit trigger: volume 10x or strategy shift.
GATE CHECK: G1 pass (3 options × n axes, matrix first) · G2 pass · G3 pass

Hard rules

1. No recommendation before the complete matrix. If the answer feels obvious, the matrix is cheap; if the matrix changes the answer, it just paid for itself.
2. Never drop an option. A degenerate option (no partner exists) is scored weak with its basis, visibly — absence of a column is a gate failure, absence of a partner is just a weak score.
3. Never invent pricing, build estimates, or vendor terms. Input numbers and labeled estimates only; a decision resting on an invented number is worse than no decision.
4. The differentiation axis must be argued from the product's wedge, not from whether customers ask for the capability — customers ask for plumbing too.

Limitations

• The matrix structures judgment; axis weights come from stated context and are re-weighable by the reader — two honest readers can land differently on close calls, and the differential line shows exactly where.
• Vendor-market specifics (which vendor, actual contract terms) are out of scope — this decides the mode, vendor selection comes after a BUY.
• Build-cost scoring uses cost classes, not engineering estimates; a real build candidate deserves an engineering spike before commitment.
• Partner scoring assumes arm's-length partnership; M&A-grade options are beyond this skill.