ChileCompra buys on behalf of the state using framework agreements, where a central agency auctions an assortment of products and agencies then buy from the winners. During 2018-19 these accounted for 23% of the agency's procurement spend.
But a look at the transaction data showed some framework auctions drew little competition. The catalogue described the same items inconsistently, so bidders could not compete directly and suppliers had less reason to cut prices.
Working with researchers, ChileCompra standardised the catalogue with natural-language processing and used it to induce more competition in the auction stage. In a redesigned food framework, competition cut transaction prices by 8%; the design was then rolled out to all framework agreements and written into procurement regulation.
Standardize items to spark competition
Inconsistent descriptions split demand and dulled competition
Standardising the catalogue made items directly comparable across bidders
More competition in the auction stage reduced transaction prices by 8%
The redesign was scaled to all frameworks and codified in new regulation
Procurement fails silently when no one competes on the same thing; standardise the units in a catalogue and the auction stage becomes a genuine contest, cutting spend without touching the budget.
ChileCompra adopted the standardisation-and-competition design across every framework agreement, and several improvements entered the new government-procurement regulation. Extrapolating from the pilot, the savings reached about $74 million in 2022, without any change to the budget.
FOLLOW THE EVIDENCE
The sources
- Saving Millions in Government Procurement Through Data Science and Market Design Stanford Graduate School of Business · gsb.stanford.edu
- Saving Millions in Government Procurement Through Data Science and Market Design INFORMS Journal on Applied Analytics · pubsonline.informs.org ·