1# Evaluation Matrix
2 
3Score options on explicit criteria. Never decide by vibe. The matrix exists so the user can see *which assumption* would have to flip for the recommendation to flip.
4 
5## Process
6 
71. Pick criteria (defaults below; tailor with the user  -  ask one at a time).
82. Weight criteria 1-5 (5 = decisive). Default: all 3.
93. Score each option per criterion 1-5.
104. Compute weighted total. Show top-2 plus the dark-horse.
115. Run a sensitivity check: which single weight or score, if flipped, changes the winner?
12 
13## Default criteria  -  software
14 
15| Criterion | What you're scoring |
16|---|---|
17| Outcome value | How much of the basic function is delivered. |
18| Implementation effort | Inverse of cost. 5 = trivial. |
19| Reversibility | How fast can we undo if wrong (5 = config flip). |
20| Risk surface | Inverse of risk; data-loss/permission/migration weigh heaviest. |
21| Operational burden | On-call, runbooks, dashboards, manual ops. |
22| User comprehension | Will a typical user grok this without training. |
23| Testability | Can we prove correctness before ship. |
24| Architecture fit | Conformance with existing patterns; deviation costs more. |
25| Time to feedback | How fast we'll learn if the bet is right. |
26 
27Drop any criterion the user explicitly waives. Add domain-specific ones (compliance, latency budget, infra cost) when relevant.
28 
29## Default criteria  -  non-software
30 
31| Criterion | Notes |
32|---|---|
33| Outcome value | Movement toward the stated outcome. |
34| Cost | Money + opportunity cost. |
35| Time to value | Calendar weeks until benefit lands. |
36| Stakeholder fit | Alignment with the people whose buy-in matters. |
37| Reversibility | Cost to undo. |
38| Risk | Severity x likelihood of negative outcomes. |
39| Learning value | What we'll know we didn't know before. |
40| Energy / morale | Will the team actually do it. |
41 
42## Template
43 
44```
45Options: A=<no-build>, B=<config-only>, C=<lean build>, D=<full build>, E=<dark-horse>
46 
47Criterion           Weight  A   B   C   D   E
48Outcome value           5   1   2   4   5   3
49Implementation effort   3   5   5   3   1   4
50Reversibility           4   5   5   3   1   4
51Risk surface            4   5   4   3   1   3
52Operational burden      3   5   4   3   1   4
53User comprehension      2   3   4   4   3   2
54Testability             3   5   4   3   2   3
55Architecture fit        3   5   3   4   3   2
56Time to feedback        4   5   5   3   1   4
57 
58Weighted totals:    (compute)  -  recommend top-2 + dark-horse.
59```
60 
61## Sensitivity check
62 
63After the table, state:
64 
65- The winner.
66- The single change that would unseat it (e.g. "if Reversibility drops to weight 2, D wins").
67- Which assumption the user should challenge first.
68 
69## Pitfalls
70 
71- Don't pad with criteria that don't differentiate options.
72- Don't let the sponsor weight; weight first, then reveal the winner.
73- If two options tie, run a sacrifice test (cut the most expensive function in each) before adding more criteria.
74