Yield decomposition of existing data¶
Part of Inverse Innovation with the Encyclopedia of Abstractions · Yield decomposition of existing data · Last revised August 2026
Status: complete, explicitly post hoc
Unit of inference: only the tested protocol strata; not the Encyclopedia's full archetype × domain universe
Result¶
Observed yield was neither confined to one archetype nor uniformly broad. The strongest defensible reading is mixed, protocol-dependent breadth. Strict outcomes appeared under multiple archetypes in both portfolio experiments, while the empirical-partner lane in Experiment 6 was broader still. Sparse event counts make rank ordering unstable.
| Stratum | Level | Outcome | Events / denominator | Archetypes with ≥1 | Top-two share | HHI | Frozen diagnostic |
|---|---|---|---|---|---|---|---|
| EXP03_CLOSED_BOOK_320 | cell | STAGE1_SUCCESS | 297/320 | 5/5 | 42.4% | 0.201 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP03_CLOSED_BOOK_320 | cell | CLOSED_BOOK_ADOPTER_PIPELINE | 33/320 | 5/5 | 57.6% | 0.2305 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP03_EXTERNALLY_ASSESSED_SELECTED_47 | proposal | EXTERNALLY_VERIFIED_PIPELINE | 21/47 | 5/5 | 66.7% | 0.2834 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP03_EXTERNALLY_ASSESSED_SELECTED_47 | proposal | ADJACENT_OR_NO_CLOSE | 6/47 | ⅗ | 83.3% | 0.3889 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP03_EXTERNALLY_ASSESSED_SELECTED_47 | proposal | ESTABLISHED_OR_SUBSTANTIAL_COLLISION | 40/47 | 5/5 | 57.5% | 0.24 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP04_PROPOSAL_FIRST_20 | cell | STRICT_SUCCESS | 5/20 | ⅘ | 60.0% | 0.28 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP04_RETRIEVAL_FIRST_20 | cell | STRICT_SUCCESS | 1/20 | ⅕ | 100.0% | 1.0 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP05_FIVE_PROPOSAL_20 | proposal | STRICT_SUCCESS | 4/100 | ⅖ | 100.0% | 0.625 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP05_FIVE_PROPOSAL_20 | cell | AT_LEAST_ONE_STRICT | 4/20 | ⅖ | 100.0% | 0.625 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP06_FOUR_PROPOSAL_60 | proposal | STRICT_SUCCESS | 15/240 | ⅘ | 60.0% | 0.2622 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP06_FOUR_PROPOSAL_60 | cell | AT_LEAST_ONE_STRICT_SUCCESS | 11/60 | ⅘ | 63.6% | 0.2727 | CONCENTRATED_IN_TESTED_ARCHETYPES |
| EXP06_FOUR_PROPOSAL_60 | proposal | EMPIRICAL_PARTNER_CANDIDATE | 32/240 | 5/5 | 53.1% | 0.2383 | BROAD_ACROSS_TESTED_ARCHETYPES |
| EXP06_FOUR_PROPOSAL_60 | cell | AT_LEAST_ONE_EMPIRICAL_PARTNER_CANDIDATE | 24/60 | 5/5 | 54.2% | 0.2361 | BROAD_ACROSS_TESTED_ARCHETYPES |
Interpretation¶
The full closed-book Experiment 3 matrix unsurprisingly shows broad Stage-1 and adopter-pipeline coverage because every archetype was crossed with all 64 domains. The selected external subset is not a comparable denominator. In Experiments 5 and 6, strict survivors were distributed across several archetypes rather than produced by a single universal winner. Experiment 6's partner candidates reached all five tested archetypes, indicating that the pipeline's unresolved-real-world-evidence lane is not an isolated archetype effect.
The frozen breadth labels are diagnostics for designing a future random breadth probe. They are not tests of whether one archetype is intrinsically more innovative, and domain rows must not be read as domain effects.