Sensitivity study
10brands · prompt3 × 3
Stability by metric
Spread across trials, in score points. A pair counts as unstable when its standard deviation across trials exceeds 5 points.
| Metric | Pairs | Mean SD | Median SD | Max SD | Within variant | Between variants | Unstable |
|---|---|---|---|---|---|---|---|
| Cluny Affinity | 45 | 1.03 | 0.996 | 2.3 | 1.03 | — | 0 |
| Brand affinity | 45 | 1.222 | 1.018 | 2.851 | 1.222 | — | 0 |
| Market proximity | 45 | 1.152 | 0.983 | 2.837 | 1.152 | — | 0 |
Against the production matrix
Compared with run 4d388d02 (published, universe brands, 100 brands). Only pairs present in both are counted.
| Metric | Pairs | Mean delta | MAE | RMSE | Max delta | Pearson r | Spearman ρ |
|---|---|---|---|---|---|---|---|
| Cluny Affinity | 45 | 2.332 | 3.103 | 3.618 | 6.892 | 0.98 | 0.983 |
| Brand affinity | 45 | -0.457 | 2.366 | 2.862 | 6.833 | 0.986 | 0.984 |
| Market proximity | 45 | 4.041 | 4.275 | 5.003 | 9.806 | 0.974 | 0.961 |
Trials
Each trial is preserved as its own affinity run and can be exported on its own with brand-map:export.
| Run | Variant | Replicate | Status | Complete pairs |
|---|---|---|---|---|
| 9020637b | prompt3 9fb2cfab52fb | 1 | completed | 45 |
| 33fb4c3f | prompt3 9fb2cfab52fb | 2 | completed | 45 |
| 1c66535d | prompt3 9fb2cfab52fb | 3 | completed | 45 |
Least stable pairs
Ranked by spread across trials, worst first, for each metric.
| Pair | Mean | SD | Range | Within variant | Between variants | Production | Delta |
|---|---|---|---|---|---|---|---|
| Aesop · Flamingo Estate | 78.575 | 2.3 | 4.5 | 2.3 | — | 79.65 | -1.075 |
| Aesop · EltaMD | 46.171 | 2.155 | 4.288 | 2.155 | — | 47.275 | -1.104 |
| MALIN+GOETZ · Merit | 69.45 | 1.889 | 3.662 | 1.889 | — | 64.438 | 5.012 |
| Flamingo Estate · Supergoop! | 51.042 | 1.86 | 3.288 | 1.86 | — | 51.95 | -0.908 |
| EltaMD · Flamingo Estate | 35.009 | 1.804 | 3.363 | 1.804 | — | 34.163 | 0.846 |
| Aesop · Merit | 65.213 | 1.674 | 3.325 | 1.674 | — | 62 | 3.213 |
| Flamingo Estate · Merit | 60.408 | 1.574 | 2.775 | 1.574 | — | 57.825 | 2.583 |
| Merit · Rare Beauty | 79.021 | 1.533 | 2.888 | 1.533 | — | 80.263 | -1.242 |
| EltaMD · Merit | 57.137 | 1.363 | 2.712 | 1.363 | — | 53.575 | 3.562 |
| Aesop · Rare Beauty | 41.104 | 1.312 | 2.413 | 1.312 | — | 42.125 | -1.021 |
| EltaMD · Supergoop! | 73.867 | 1.284 | 2.312 | 1.284 | — | 75.888 | -2.021 |
| Le Labo · Rare Beauty | 42.129 | 1.281 | 2.562 | 1.281 | — | 38.625 | 3.504 |
| Rare Beauty · Supergoop! | 75.059 | 1.227 | 2.137 | 1.227 | — | 71.725 | 3.334 |
| MALIN+GOETZ · Rare Beauty | 50.109 | 1.21 | 2.15 | 1.21 | — | 49.275 | 0.834 |
| EltaMD · Oribe | 45.108 | 1.164 | 2.075 | 1.164 | — | 42.125 | 2.983 |
| La Mer · Merit | 54.917 | 1.119 | 2.238 | 1.119 | — | 50.5 | 4.417 |
| Aesop · MALIN+GOETZ | 85.4 | 1.108 | 2.213 | 1.108 | — | 84.7 | 0.7 |
| Flamingo Estate · Le Labo | 77.879 | 1.07 | 2.137 | 1.07 | — | 78.2 | -0.321 |
| Flamingo Estate · Rare Beauty | 36.305 | 1.063 | 2 | 1.063 | — | 38.613 | -2.308 |
| La Mer · Oribe | 73.28 | 1.041 | 2 | 1.041 | — | 67.45 | 5.83 |
| La Mer · Rare Beauty | 39.909 | 1.032 | 1.812 | 1.032 | — | 35.588 | 4.321 |
| EltaMD · MALIN+GOETZ | 54.087 | 1 | 1.937 | 1 | — | 53.363 | 0.724 |
| Oribe · Supergoop! | 58.542 | 0.996 | 1.8 | 0.996 | — | 52.738 | 5.804 |
| Oribe · Rare Beauty | 48.134 | 0.963 | 1.912 | 0.963 | — | 44.9 | 3.234 |
| EltaMD · Le Labo | 37.4 | 0.953 | 1.675 | 0.953 | — | 33.863 | 3.537 |