A public architecture case study and reusable Excel template for selecting an enterprise IT solution, platform or product through a structured trade study: non-compensatory mandatory gates, evidence-backed weighted shortlist scoring, TCO, risk analysis and sensitivity testing.
Role: enterprise & solution architecture, decision analysis, criteria design, architecture assessment, TCO/risk trade-offs and architecture recommendation.
Publication boundary: this is a newly authored vendor-neutral public framework. It does not publish internal corporate standards, source templates, real projects, vendors, commercial figures, scores or procurement results.
A single giant scoring matrix can hide critical defects. A mandatory failure may be compensated by unrelated high scores; subjective numbers can look falsely precise; cost, risk and uncertainty may disappear inside one opaque total.
This framework therefore separates the decision into distinct layers:
- decision framing;
- non-compensatory mandatory gates;
- long-list screening;
- weighted evidence-backed evaluation of the viable shortlist;
- TCO on one common horizon and cost perimeter;
- initial and residual risks;
- sensitivity analysis;
- an explicit recommendation with trade-offs, conditions and evidence gaps.
flowchart LR
A[Decision framing] --> B[Mandatory gates]
B --> C[Long list]
C --> D{Gate screening}
D -->|Fail| X[Eliminate]
D -->|Pass| E[Shortlist]
E --> F[Weighted evidence-backed scoring]
F --> G[TCO]
G --> H[Risk]
H --> I[Sensitivity]
I --> J[Recommendation]
| Artifact | Purpose |
|---|---|
| Excel template | clean vendor-neutral workbook |
| Completed demo | fully populated synthetic example |
| Methodology | step-by-step decision process |
| Scoring model | gates, weights, scoring anchors, TCO, risk and sensitivity |
| Anonymization | public-sanitization boundary |
assets/screenshots/ |
screenshots from the completed demo |
Mandatory gates are non-compensatory. A failed mandatory requirement eliminates an alternative; unknown evidence is incomplete rather than an implicit pass.
For viable alternatives, the default scored scale is 0…5:
Suitability % = Σ(weight_i × score_i) / 5
Weights sum to 100%. Material scores should be backed by verifiable evidence such as documentation, a PoC, benchmark, test result, contract term or reference architecture.
TCO and risk are deliberately kept outside the suitability score. This prevents a cheap solution from being presented as architecturally strong, or a technically strong solution from hiding a material cost/risk penalty.
The demo selects a fictional enterprise API management platform.
| Alternative | Gates | Suitability | 3-year TCO | Residual risk | Result |
|---|---|---|---|---|---|
| Atlas Gateway | PASS | 87.4% | 3225 | 10 | recommended with conditions |
| Boreal API Platform | PASS | 87.0% | 3175 | 8 | strong alternative |
| Cedar Integration Edge | FAIL | — | — | — | eliminated |
| Delta API Fabric | FAIL | — | — | — | eliminated |
The example intentionally keeps the top two alternatives close. Atlas leads the base architecture score; Boreal has lower TCO and residual risk and becomes preferable in operations-first and cost-conscious sensitivity scenarios. The point is to expose the trade-off rather than manufacture a universal winner.
The public framework uses general systems-engineering decision-analysis practices. NASA guidance explicitly discusses measurable criteria and their importance, mandatory vs enhancing criteria, total ownership/life-cycle cost, risk, evidence, uncertainty and sensitivity.
- NASA Systems Engineering Handbook — Decision Analysis: https://www.nasa.gov/reference/6-8-decision-analysis/
- NASA NPR 7123.1 — Systems Engineering Processes and Requirements: https://nodis3.gsfc.nasa.gov/displayDir.cfm?t=NPR&c=7123&s=1
This repository is not a copy of an internal corporate method. The public method was reconstructed and generalized from professional decision-analysis practice. Real organizations, projects, products, internal standards, prices, vendor assessments and evidence are not included.
See ORIGIN.md and docs/anonymization.md.
MIT License. See LICENSE.