R048: Heterogeneous supplier

Detects suppliers with unusually diverse product portfolios across awards

The variety of items supplied by a tenderer is a high outlier.

Methodology

This indicator requires awarded items to be classified using a hierarchy of numeric codes. For example, UNSPSC and CPV have four primary levels.

UNSPSC Classification

Digits	Level
XX000000	Segment
XXXX0000	Family
XXXXXX00	Class
XXXXXXXX	Commodity

CPV Classification

Digits	Level
XX000000	Division
XXX00000	Group
XXXX0000	Class
XXXXX000	Category

For each tenderer, the variety is calculated as the number of distinct first-level classifications across all items awarded to the tenderer, across all contracting processes.

A tenderer is flagged if its variety is greater than or equal to the upper fence of:

$$Q_3 + 1.5(IQR)$$

where $Q_3$ is the third quartile and $IQR$ is the interquartile range for the set of varieties among tenderers awarded in at least 20 contracting processes.

Example

25% ($Q_1$) of bidders in Atlantis supply items in 1 product category. 75% ($Q_3$) of bidders supply items in at most 4 product categories. This yields an upper fence of 8.5 (9 in effect).

RentSeekers Inc. supplies items in 20 product categories. This is higher than the upper fence of 8.5. Therefore, RentSeekers Inc. is flagged.

Why is this a red flag?

A heterogeneous supplier is more likely to be unsuitable, increasing the risk of low value for money, low quality of delivery, and/or rent extraction.

_{Based on “Component C1B” in Design and measurement of a corruption risk index from a supplier perspective in the context of COVID-19 emergency.}

Output

The indicator’s value is the variety, as a decimal (the count of distinct classification codes).

Configuration

All configuration is optional.

R048.threshold · integer

Override the calculated upper fence with a fixed threshold for the number of distinct product categories.

R048.digits · integer · default: 2

Number of digits to use from the classification code for grouping. Use 2 for first-level classification (e.g., XX000000). Increase if the first level is too broad.

R048.minimum_contracting_processes · integer · default: 20

Minimum number of contracting processes a tenderer must be awarded in to be included when calculating the upper fence. This minimizes correlation between award count and variety.

[R048]
threshold = 10  # optional: override calculated upper fence
digits = 2  # default: use first-level classification
minimum_contracting_processes = 20  # default

Implementation Details

The indicator (from src/indicators/r048.rs):

• Collects classification codes from awarded items for each supplier
• Takes only the first N digits of each classification code (configurable)
• Counts distinct classification codes per supplier
• Tracks the number of contracting processes each supplier participates in
• Uses statistical outlier detection (quartiles and IQR) to identify heterogeneous suppliers
• Only includes suppliers meeting the minimum contracting process threshold in the calculation

The indicator only processes awards with exactly one supplier. Awards to multiple suppliers are excluded to avoid overcounting.

Exclusions

An active award is excluded if it is made to multiple suppliers.

Demonstration

Input

// Example OCDS releases showing supplier winning awards across many diverse product categories

Output

ocdscardinal indicators --settings docs/examples/R/048.ini --no-meta docs/examples/R/048.jsonl

{"Tenderer":{"F":{"R048":2.0}}}