The provenance effect: 0.31 pp
With provenance recorded on every record, the index can measure whether supplier-supplied samples read better than material bought independently. They do.
Provenance became a required field in 2024-Q4. By 2025-Q1 the index held enough lots assayed from both a supplier-supplied sample and an independently purchased vial to measure the difference.
The result
Across all paired lots, the supplier-supplied result reads a mean 0.31 pp higher than the independently purchased vial, with a 95 % confidence interval of 0.19 to 0.43 pp. The effect is small on average, statistically clear, and highly variable between individual lots.
| Provenance class | Share of index | Median purity, matched compounds | Weight applied |
|---|---|---|---|
| Buyer-sourced | about 85 % | reference | Full — used in all sensitivity analyses |
| Supplier-supplied | about 12 % | +0.31 pp | Published; excluded from buyer-sourced subsets |
| Laboratory-retained | about 3 % | not significantly different | Published with storage duration stated |
Three mechanisms, none of them separable
- The sample may be from a different fill than retail stock. Several submitters have reported physically different vials, caps or fill volumes for samples versus retail purchases.
- The sample may be from a hand-picked lot.
- The retail vial may have degraded in distribution in a way the sample did not — in which case the finding is about the supply chain rather than about anyone's honesty.
The index cannot distinguish between these and does not claim to. What it does is record provenance on every record, report each supplier's buyer-sourced subset separately from its total, and flag the individual lots where the gap exceeds 0.8 pp under rule AX-R10.
Supplier-supplied fractions
The fraction varies substantially between suppliers, from under 10 % to over 30 %. Suppliers above 25 % now carry a note on their scorecard, because at that share the aggregate is materially affected by the effect measured above. The laboratory that purchases anonymously as a matter of policy has a buyer-sourced fraction of 94 % and its record set is the cleanest available subset for anyone wanting to exclude the effect entirely.