Competition Law and Procurement Algorithm Transparency and Competition Law

1. Introduction

Procurement algorithms are increasingly used by public authorities and large purchasing organizations to automate or assist procurement activities such as:

identifying potential suppliers;

evaluating bids;

detecting unusually low or high prices;

scoring technical proposals;

ranking bidders;

detecting conflicts of interest;

identifying possible bid-rigging;

allocating contracts;

predicting procurement costs;

monitoring supplier performance; and

determining which tenders require human review.

Algorithmic procurement can improve efficiency and reduce administrative costs. It can also potentially strengthen competition by detecting collusion and making evaluation more consistent.

However, procurement algorithms create a distinctive competition-law problem:

If an algorithm determines or materially influences who can compete, how bids are evaluated, or which supplier wins, insufficient transparency can make discriminatory, collusive, or exclusionary effects difficult to detect and challenge.

The issue therefore lies at the intersection of competition law, public procurement law, administrative law, algorithmic governance, procedural fairness, and data governance.

2. What Is Procurement Algorithm Transparency?

Procurement algorithm transparency does not necessarily mean publicly revealing every line of source code.

It may instead require sufficient information concerning:

the purpose of the algorithm;

the data used;

the evaluation criteria;

the weighting of criteria;

the ranking methodology;

material decision rules;

potential conflicts of interest;

human involvement;

auditability;

mechanisms for correcting errors; and

reasons for decisions affecting bidders.

This distinction is important.

Source-code transparency

The actual programming code is disclosed.

Functional transparency

The affected bidder is told how the system materially operates and which factors determine the result.

Competition law will frequently be more concerned with functional and evidentiary transparency than with unrestricted publication of source code.

3. Why Procurement Algorithms Matter to Competition

Public procurement represents substantial economic activity.

A procurement authority may receive bids from:

large corporations;

SMEs;

new entrants;

foreign suppliers;

consortiums;

specialist suppliers.

The algorithm used to select among them can therefore influence market structure.

A poorly designed system can potentially:

favour incumbents;

disadvantage new entrants;

systematically favour particular suppliers;

exclude firms because of historical data;

reproduce discriminatory procurement practices;

facilitate coordination;

create artificial entry barriers;

conceal bid-rigging;

make procurement markets less contestable.

4. Pro-Competitive Uses of Procurement Algorithms

Algorithms can also strengthen competition.

A. Bid-rigging detection

An algorithm can identify suspicious patterns such as:

identical bids;

rotating winners;

unusual bid sequences;

repeated subcontracting arrangements;

geographically suspicious allocation;

synchronized withdrawals;

unusual pricing similarities.

B. Supplier discovery

Automated systems can identify SMEs that would otherwise not participate.

C. Objective evaluation

An algorithm can reduce arbitrary human decision-making where its criteria are properly designed.

D. Faster procurement

Automated evaluation can reduce transaction costs.

E. Market intelligence

Procurement authorities can identify:

supplier concentration;

price trends;

entry barriers;

repeated procurement patterns.

Thus, algorithms can become competition-enhancing enforcement tools rather than merely decision-making tools.

5. Competition Risks from Opaque Algorithms

5.1 Hidden discriminatory criteria

Suppose an algorithm evaluates suppliers partly according to previous contracts.

Established suppliers will naturally have more historical data.

New entrants may therefore receive lower scores simply because they are new.

This can create a circular effect:

incumbent wins → more procurement history → higher algorithmic score → incumbent wins again.

This is an important potential algorithmic entry barrier.

6. Algorithmic Bias and Market Foreclosure

An algorithm can unintentionally reproduce historical procurement patterns.

For example:

If previous government contracts are heavily weighted, Supplier A may continue winning regardless of whether B or C could offer a better current bid.

The problem is not necessarily intentional discrimination.

The competition concern may arise from structural foreclosure.

7. Transparency and Competition Are Closely Connected

Transparency matters because competition authorities and courts must be able to determine:

Why did Supplier A win and Supplier B lose?

Without adequate information, it may be difficult to identify:

discriminatory treatment;

exclusionary criteria;

manipulation;

conflicts of interest;

collusion;

errors;

arbitrary weighting.

Accordingly:

algorithmic transparency → auditability → contestability → stronger competitive accountability.

8. Case Law

1. Concordia Bus Finland — C-513/99

Concordia Bus Finland Oy Ab v Helsingin Kaupunki

This is a leading European procurement case concerning the use of environmental criteria in public procurement.

The Court of Justice accepted that contracting authorities could use environmental considerations when awarding contracts, provided that the criteria were connected with the subject matter and complied with the relevant procurement principles.

Competition significance

The case demonstrates that procurement evaluation criteria need not be restricted to the lowest price.

However, criteria must be:

objectively applicable;

connected with the procurement;

transparent;

compatible with equal treatment.

Algorithmic relevance

If an algorithm incorporates environmental, technical, or sustainability criteria, the criteria and their weighting should be sufficiently identifiable to ensure that bidders are competing according to known rules.

9. 2. Fabricom — Joined Cases C-21/03 and C-34/03

Fabricom SA v Belgian State

Fabricom concerned exclusion from public procurement where an undertaking had participated in preparatory work relating to the procurement.

The Court examined whether automatic exclusion could be justified without allowing the undertaking an opportunity to demonstrate that its prior involvement had not distorted competition.

Competition principle

A procurement system should not automatically exclude an undertaking without considering whether competition has actually been distorted and whether the undertaking has an opportunity to rebut the concern.

Algorithmic relevance

An algorithm that automatically flags a supplier as:

"conflicted"

or

"high risk"

should not necessarily convert that classification into automatic exclusion without an appropriate verification mechanism.

This illustrates the importance of:

explainability;

human review;

contestability;

individualized assessment.

10. 3. Assitur — C-538/07

Assitur Srl v Camera di Commercio, Industria, Artigianato e Agricoltura di Milano

The case concerned automatic exclusion of companies because of relationships between bidders.

The Court held that automatic exclusion rules could not be applied mechanically without examining whether the relationship actually affected competition or violated the relevant procurement requirements.

Algorithmic relevance

This is particularly important for procurement algorithms designed to detect:

common ownership;

related bidders;

family relationships;

corporate links;

consortium relationships.

An algorithm may identify a relationship, but the existence of a relationship does not necessarily establish collusion or anti-competitive conduct.

Principle

Detection of a risk is not necessarily proof of an infringement.

11. 4. Fastweb — C-100/12

Fastweb SpA v Azienda Sanitaria Locale di Alessandria

Fastweb concerned judicial protection in public procurement and the ability of unsuccessful bidders to challenge procurement decisions.

The judgment emphasized the importance of effective remedies in procurement proceedings.

Competition relevance

If an algorithm determines the outcome of a procurement process, affected bidders need an effective mechanism for challenging:

incorrect data;

erroneous scoring;

unlawful criteria;

procedural irregularities;

discriminatory treatment.

Algorithmic relevance

An opaque automated decision with no meaningful appeal mechanism can undermine effective procurement review.

Therefore:

algorithmic decision → reasoned outcome → review mechanism → effective remedy.

12. 5. Max Havelaar — C-368/10

Commission v Kingdom of the Netherlands

This case concerned the use of environmental and fair-trade requirements in public procurement.

The Court accepted that contracting authorities could pursue legitimate environmental and social objectives, but procurement requirements still had to comply with EU procurement principles.

Competition significance

Procurement algorithms may incorporate sustainability scores, ESG factors, labour standards, or environmental criteria.

The Max Havelaar judgment illustrates that such criteria must not be designed in a way that arbitrarily favours particular suppliers or products without an adequate connection to the procurement.

Algorithmic principle

The more significant a criterion is in determining the outcome, the more important it becomes that:

its relevance is demonstrable;

bidders can understand it;

it is applied consistently.

13. 6. Eturas — C-74/14

Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba

This is one of the most important competition cases for algorithmic procurement by analogy.

Eturas concerned a common electronic booking system used by travel agencies. The system administrator implemented a technical restriction affecting discounts available to participating businesses.

The Court considered whether the existence and communication of the automated restriction could support a finding of concerted practice.

Competition significance

The case demonstrates that a digital system can become part of a competition-law infringement.

Technology is not competition-law neutral.

Procurement relevance

Imagine several suppliers participating in a procurement ecosystem using a common algorithm that:

observes competitors' bids;

recommends prices;

restricts discounting;

automatically responds to competing bids.

The system could potentially facilitate coordination.

Principle

An algorithm can facilitate competition law violations just as a human communication mechanism can.

14. 7. T-Mobile Netherlands — C-8/08

T-Mobile Netherlands BV and Others v Raad van bestuur van de Nederlandse Mededingingsautoriteit

The Court held that a single meeting between competitors could constitute a concerted practice where the communication was capable of reducing uncertainty concerning competitors' future market behaviour.

Procurement relevance

Procurement algorithms can potentially increase transparency among competing bidders.

For example, if competing bidders obtain information about:

rivals' bids;

expected pricing;

probability of winning;

future tender participation;

they may be able to coordinate more effectively.

Principle

Competition law is particularly concerned with communications that reduce strategic uncertainty among competitors.

15. 8. Fabricom and Algorithmic Exclusion

The importance of Fabricom deserves particular emphasis.

Imagine an algorithm identifies that:

Company A previously advised the government on the design of the procurement.

The system automatically excludes Company A.

The competition-law question should not simply be:

"Did the algorithm find a conflict?"

It should also be:

"Does the prior involvement actually give Company A an unfair competitive advantage?"

The distinction between risk detection and automatic exclusion is fundamental.

16. Algorithmic Bid Evaluation

Procurement algorithms may assign scores such as:

Price — 50%

Technical quality — 25%

Environmental performance — 15%

Delivery — 10%

An algorithm can then calculate:

Supplier A — 87.4
Supplier B — 86.8
Supplier C — 81.2

The competition issue is whether the scoring model itself is:

lawful;

objective;

proportionate;

non-discriminatory;

sufficiently transparent.

If bidders do not know how the score is generated, their ability to compete effectively may be weakened.

17. Algorithmic Procurement and New Entrants

This is one of the most important competition concerns.

Suppose an algorithm gives 30% of its score to:

"Historical government performance."

An established supplier has 100 previous contracts.

A new entrant has none.

Even if the new entrant offers:

a lower price;

superior technology;

better delivery;

better environmental performance,

the historical-data criterion may systematically disadvantage it.

This can create algorithmic incumbency bias.

Competition law should therefore examine whether procurement algorithms unintentionally transform historical market success into a permanent competitive advantage.

18. Data Advantage and Information Asymmetry

Procurement authorities possess large amounts of data.

An algorithm can process:

historical bids;

supplier prices;

contract performance;

procurement failures;

delivery records;

bid withdrawals;

subcontracting patterns.

This can improve competition enforcement.

But bidders may not know:

what data is being used;

whether it is accurate;

how old it is;

how heavily it is weighted;

whether it has been corrected.

Incorrect data can therefore produce systematic exclusion.

19. Algorithmic Collusion in Procurement

Public procurement is particularly vulnerable to bid rigging.

Traditional bid-rigging mechanisms include:

bid rotation;

cover bids;

market allocation;

subcontracting arrangements;

customer allocation.

Algorithms can potentially make such coordination easier.

For example:

Algorithm A recommends ₹10.2 million.
Algorithm B observes the market price.
Algorithm C adjusts its bid to remain just above the expected winner.

The result could be highly predictable tender outcomes.

Competition authorities must therefore distinguish between:

Legitimate algorithmic pricing

Each bidder independently uses its own algorithm.

and

Coordinated algorithmic behaviour

Algorithms are deliberately configured or communicated so that competing firms coordinate their future conduct.

20. Procurement Algorithms as Competition-Enforcement Tools

The relationship also works in the opposite direction.

A procurement authority can use AI to detect bid rigging.

Potential indicators include:

Bid similarity

Unusually similar bids.

Bid rotation

Different firms repeatedly win in predictable sequences.

Price clustering

Prices repeatedly fall within an unusually narrow range.

Withdrawal patterns

Certain firms systematically withdraw.

Geographic allocation

Different suppliers consistently win particular regions.

Subcontracting

Losing bidders repeatedly become subcontractors of winners.

These indicators do not necessarily prove collusion.

They should instead generate investigative leads.

21. False Positives

A major competition problem arises when algorithms confuse legitimate competition with collusion.

For example, several firms may submit nearly identical bids because:

costs are similar;

government pricing formulas are known;

materials have standardized prices;

contract specifications are highly detailed.

An algorithm might classify this as suspicious.

Therefore:

algorithmic suspicion ≠ proof of cartelization.

This principle is consistent with the reasoning underlying cases such as Assitur and the broader competition-law requirement for evidence of prohibited coordination.

22. Transparency Versus Confidentiality

Complete transparency is not always possible.

Procurement systems may contain:

trade secrets;

cybersecurity information;

confidential algorithms;

commercially sensitive data;

personal information.

Consequently, the appropriate model is often controlled transparency.

For example:

Publicly disclosed

evaluation criteria;

weightings;

procedural rules;

basic methodology.

Available to bidders

relevant scoring methodology;

reasons for individual outcomes;

material factors affecting their bid.

Available to regulators/auditors

source code;

training data;

logs;

model documentation;

audit records.

Protected

confidential business information;

security-sensitive material.

23. Human Oversight

Human oversight is particularly important when an algorithm:

excludes a bidder;

determines a significant score;

identifies suspected collusion;

changes the competitive conditions;

determines contract allocation.

The human reviewer should be capable of:

understanding the relevant output;

examining underlying evidence;

correcting errors;

considering explanations from bidders;

overriding an erroneous result.

A nominal human reviewer who simply accepts the algorithm's recommendation does not necessarily provide meaningful oversight.

24. Indian Competition-Law Framework

The Indian Competition Act, 2002 is relevant where algorithmic procurement affects competition among suppliers.

Section 3

Section 3 addresses anti-competitive agreements.

Public procurement algorithms may become relevant where suppliers use technology to facilitate:

bid rigging;

price coordination;

market allocation;

information exchange.

Section 3(3) is particularly important for agreements involving:

price fixing;

limitation of supply;

market allocation;

bid rigging or collusive bidding.

Section 4

Section 4 becomes relevant where a dominant enterprise uses procurement or technological infrastructure to:

deny market access;

impose discriminatory conditions;

leverage dominance;

exclude competing suppliers.

Section 19

The CCI can consider factors such as:

market structure;

barriers to entry;

market power;

economic advantages;

consumer interests;

competitive effects.

These considerations can be relevant to algorithmically structured procurement markets.

25. Public Procurement and Competition Law Are Complementary

Public procurement law traditionally focuses on:

transparency;

equal treatment;

non-discrimination;

procedural fairness;

value for money.

Competition law focuses on:

preserving rivalry;

preventing collusion;

preventing exclusion;

protecting market access;

maintaining competitive conditions.

Algorithmic procurement connects the two.

For example:

An opaque algorithm may violate procurement principles because bidders cannot understand the evaluation process, while the same opacity may make competition-law violations harder to detect.

26. Competition Effects of Procurement Algorithms

27. Six Major Competition-Law Principles

Principle 1 — Procurement algorithms must not become hidden barriers to entry

If algorithmic criteria systematically favour established suppliers without objective justification, competitive entry can suffer.

Principle 2 — Automated detection is not equivalent to proof

An algorithm identifying suspicious conduct should normally generate an investigation rather than automatically establish liability.

Principle 3 — Algorithmic transparency promotes contestability

Bidders need enough information to understand and challenge material aspects of procurement decisions.

Principle 4 — Algorithms can facilitate collusion

The Eturas and T-Mobile Netherlands cases illustrate the broader principle that digital information systems and communications can reduce uncertainty among competitors.

Principle 5 — Procurement criteria can pursue legitimate policy objectives

Concordia Bus Finland and Max Havelaar demonstrate that procurement can legitimately incorporate environmental and social considerations, provided the applicable legal conditions are respected.

Principle 6 — Automated exclusion requires safeguards

Fabricom, Assitur, and Fastweb illustrate the importance of individualized assessment, equal treatment, and effective remedies in procurement decision-making.

28. Emerging Competition Issues

A. AI-generated bids

Generative AI could allow suppliers to produce sophisticated bids rapidly.

This may reduce barriers to participation but also make bid similarity more common.

B. AI-assisted bid coordination

Competing firms could potentially use algorithms that automatically respond to market information.

Competition authorities may therefore need to investigate not merely communications between executives but also:

algorithm design;

pricing instructions;

shared software;

data feeds;

configuration settings.

C. Procurement-platform dominance

If one company provides the dominant procurement infrastructure used by public authorities, questions can arise concerning:

access;

interoperability;

data portability;

discriminatory treatment;

preferential integration.

D. Algorithmic supplier blacklisting

A predictive model might classify a supplier as high risk.

If the classification is opaque or based on inaccurate historical information, the supplier could effectively be excluded from an important market without adequate procedural safeguards.

E. Self-learning procurement systems

A machine-learning system could change its weighting or decision-making patterns over time.

This creates a difficult question:

If even the contracting authority cannot clearly explain why the algorithm produced a particular outcome, how can a bidder effectively challenge it?

This makes continuous auditing particularly important.

29. Suggested Compliance Framework

A competition-sensitive procurement algorithm should ideally incorporate:

Before deployment

competition-impact assessment;

bias testing;

market-access assessment;

conflict-of-interest review;

validation of data.

During operation

audit logs;

monitoring;

human oversight;

anomaly detection;

periodic recalibration.

After a decision

explanation of material factors;

bidder access to appropriate reasons;

correction mechanism;

independent review;

appeal or challenge procedure.

30. Conclusion

Procurement algorithms can fundamentally change how competition takes place in public markets.

They can increase competition by:

identifying more suppliers;

reducing procurement costs;

detecting bid rigging;

improving objective evaluation;

lowering administrative barriers.

But they can also weaken competition when they:

favour incumbents;

conceal discriminatory criteria;

rely excessively on historical data;

exclude new entrants;

facilitate collusion;

generate unexplained rankings;

misuse commercially sensitive information;

create barriers to challenging procurement decisions.

The leading case law provides a useful framework:

Concordia Bus Finland — procurement criteria can pursue legitimate objectives while remaining subject to competition and transparency principles.

Fabricom — automatic exclusion requires safeguards against unjustified competitive distortion.

Assitur — relationships between bidders should not automatically be treated as proof of prohibited conduct.

Fastweb — effective remedies are essential to procurement legality.

Max Havelaar — sustainability and social criteria may be legitimate but must comply with procurement principles.

Eturas — digital systems can facilitate competition-law infringements.

T-Mobile Netherlands — information exchange that reduces competitive uncertainty can be legally significant.

The fundamental principle is therefore:

Procurement algorithms should make competition more objective and contestable, not transform hidden technical rules into unreviewable barriers to market participation.

For competition law, the crucial questions are who controls the algorithm, what data it uses, how it affects supplier access and ranking, whether competitors can challenge its output, and whether the system facilitates or suppresses independent competitive behaviour.