Competition Law and Procurement Algorithm Transparency

1. Introduction

Modern public and private procurement increasingly relies on electronic procurement platforms, automated tender evaluation, bid-screening systems, algorithmic scoring, dynamic pricing tools, fraud-detection systems, and AI-assisted procurement software. These technologies can improve speed, accuracy and detection of suspicious bids, but they also create new competition-law risks.

Procurement algorithm transparency refers to the degree to which the procuring authority, bidders, competition authorities and courts can understand and scrutinise how an algorithm:

• ranks or scores bids;
• determines eligibility;
• detects similarities between bids;
• recommends winners;
• calculates prices or benchmarks;
• identifies suspicious bidding patterns;
• allocates procurement opportunities;
• uses historical bid data;
• shares or processes competitively sensitive information; and
• modifies procurement outcomes through automated decision-making.

The central competition-law issue is that algorithmic opacity must not become a mechanism through which bid rigging, discriminatory access, information exchange, exclusion or coordinated conduct is concealed.

The OECD's current procurement guidance specifically recognises the importance of electronic procurement systems, while also warning that transparency must be balanced because excessive disclosure of bidder information can itself facilitate collusion.

2. Meaning of Procurement Algorithm Transparency

Procurement algorithm transparency does not necessarily mean disclosure of the entire source code.

A competition-law-oriented transparency framework may instead require disclosure or auditability of:

1. Purpose of the algorithm
   What procurement decision is being automated?
2. Input variables
   What information is used to evaluate bids?
3. Weighting methodology
   How are price, quality, delivery, experience and other factors weighted?
4. Treatment of bidders
   Are all similarly situated bidders processed under the same rules?
5. Data provenance
   Where does the algorithm obtain its data?
6. Sensitive-information safeguards
   Can one bidder's confidential information influence another bidder's evaluation?
7. Human oversight
   Can procurement officials review or override an algorithmic decision?
8. Audit logs
   Can investigators reconstruct what the system did and when?
9. Version control
   Was the algorithm changed during the tender?
10. Detection mechanisms
    Does the system identify cover bidding, bid rotation, identical pricing or suspicious bid patterns?

Thus, transparency is fundamentally connected with contestability, equal treatment, accountability and effective competition.

3. Competition-Law Risks Created by Procurement Algorithms

A. Algorithmic Bid Rigging

Competitors may use algorithms to implement:

• bid rotation;
• cover bidding;
• bid suppression;
• market allocation;
• coordinated pricing;
• coordinated withdrawal; or
• artificial price convergence.

The OECD identifies cover bidding, bid suppression, bid rotation and market allocation as principal forms of bid rigging.

Under Indian law, bid rigging or collusive bidding between competitors falls within Section 3(3)(d) of the Competition Act 2002, subject to the statutory framework concerning appreciable adverse effect on competition.

B. Algorithm as a Communication Mechanism

An algorithm can effectively become a digital communication channel.

For example:

Bidder A uploads its pricing information → procurement algorithm processes it → information is indirectly made available to Bidder B → Bidder B adjusts its bid.

Even where the competitors do not communicate through ordinary emails or meetings, the competitive significance of the information flow must be examined.

This creates an important distinction between:

legitimate automated procurement
and
algorithmically facilitated coordination.

4. Transparency Versus Collusion

An unusual feature of procurement law is that more transparency is not always better for competition.

Suppose a procurement authority publicly reveals:

• identity of every bidder;
• precise bid amounts;
• bid histories;
• rejected bids;
• individual cost information;
• future procurement schedules.

This may improve administrative transparency but can simultaneously give competitors information enabling them to coordinate future bids.

The OECD's 2025 guidelines therefore recommend a balanced approach: procurement should be transparent, but authorities should avoid unnecessarily revealing bidder identities and competitively sensitive information to rivals.

This produces an important principle:

Transparency to the regulator is not necessarily the same thing as transparency to competitors.

5. Algorithmic Discrimination and Exclusion

A procurement algorithm may favour particular suppliers through:

• discriminatory scoring;
• hidden eligibility filters;
• supplier-specific thresholds;
• preferential historical-data weighting;
• automated rejection;
• exclusion based on allegedly poor performance;
• biased risk classifications.

Where the procuring authority possesses substantial market power or controls an essential procurement platform, algorithmic exclusion can potentially raise abuse-of-dominance concerns.

The competition-law inquiry should therefore ask:

1. Is there discriminatory treatment?
2. Is there an objective justification?
3. Does the algorithm disadvantage equally efficient competitors?
4. Does it restrict market access?
5. Is the effect temporary or structural?
6. Does it foreclose competitors?
7. Can affected suppliers challenge the decision?

6. Algorithmic Bid Evaluation and Due Process

An automated scoring system may determine:

Final Score=w1(Price)+w2(Quality)+w3(Experience)+w4(Delivery)+w5(Other Factors)Final\ Score = w_1(Price)+w_2(Quality)+w_3(Experience)+w_4(Delivery)+w_5(Other\ Factors)

Competition concerns arise if bidders do not know:

• what the variables mean;
• how they are weighted;
• whether the weights change;
• whether qualitative information is interpreted consistently;
• whether the algorithm contains supplier-specific adjustments.

For competition purposes, predictability of the rules is particularly important.

A bidder should not face an undisclosed algorithmic rule that materially changes competitive conditions after bids have been submitted.

7. Procurement Algorithms as Evidence

Algorithms can also strengthen competition-law enforcement.

A procurement authority can analyse:

• identical bids;
• suspiciously similar bid sequences;
• repeated winning patterns;
• geographic allocation;
• bid withdrawal;
• abnormal price differences;
• timing of submissions;
• common metadata;
• repeated supplier combinations.

The resulting algorithmic evidence should, however, be treated as evidence of possible coordination rather than automatic proof of an agreement.

This distinction is important because similar bids can have legitimate explanations, such as:

• identical input costs;
• regulated prices;
• common market conditions;
• standardised tender specifications;
• small numbers of suppliers.

The CCI has similarly recognised the evidentiary importance of contextual "plus factors" in bid-rigging cases.

8. Important Case Laws

Case 1 — A Foundation for Common Cause & People Awareness v. PES Installations Pvt. Ltd. & Ors., Case No. 43 of 2010 — India

This was a public-procurement bid-rigging matter involving a tender for the supply, installation, testing and commissioning of modular operation theatres and medical-gas systems.

The CCI identified common mistakes in tender documents as evidence supporting an inference of collusion.

The case demonstrates that competition authorities need not depend exclusively upon direct evidence such as an explicit cartel agreement. Patterns in electronically or physically submitted tender documents can constitute important circumstantial evidence.

Relevance to algorithm transparency

A modern procurement algorithm should therefore preserve:

• submission metadata;
• document histories;
• timestamps;
• common formatting;
• common errors;
• IP/device information where lawfully collected; and
• document-generation characteristics.

An algorithm capable of detecting such patterns could substantially improve cartel detection.

The CCI's published procurement material records the original penalty and its subsequent reduction by COMPAT.

Case 2 — Aluminium Phosphide Tablets Manufacturers, Suo Motu Case No. 02 of 2011 — India

The case concerned procurement by the Food Corporation of India.

The CCI considered identical bid prices, combined with other circumstances including simultaneous presence of bidders at FCI premises, as evidence supporting a finding of collusion.

The important principle is that:

Identical prices alone should not mechanically be treated as proof of collusion; the surrounding circumstances matter.

Relevance to algorithms

An algorithm should therefore not simply flag:

BidA=BidBBid_A = Bid_B

as "cartel".

Instead, it should evaluate multiple variables:

Risk=f(price similarity, timing, bid history, withdrawal, rotation, communications, market conditions)Risk = f(price\ similarity,\ timing,\ bid\ history,\ withdrawal,\ rotation,\ communications,\ market\ conditions)

This is a much more defensible competition-law architecture.

The CCI's procurement guidance identifies this case as an important Indian bid-rigging decision.

Case 3 — LPG Cylinder Manufacturers, Suo Motu Case — India

The CCI examined allegations concerning procurement of LPG cylinders by Indian Oil Corporation.

The competitors submitted identical price quotations, and the Commission treated the conduct as evidence of coordinated bidding when considered with the surrounding circumstances.

Competition-law significance

This demonstrates the danger of algorithmically coordinated tender pricing.

If competing firms use automated systems that continuously monitor one another and automatically adjust bids, the technology does not immunise the conduct from competition law.

The relevant legal question remains:

Was there an agreement, understanding or concerted action restricting competition?

The CCI's public-procurement material records the finding and sanctions in this case.

Case 4 — In Re: Alleged Bid-Rigging in E-Tenders invited by the Department of Agriculture, Government of Uttar Pradesh for Soil Sample Testing, Case No. 01/2020 — India

This case is particularly important for the present topic because it concerned e-tenders.

The CCI opened proceedings concerning alleged bid rigging in electronic tenders issued by the Uttar Pradesh Department of Agriculture for soil sample testing.

Importance for procurement algorithms

Electronic procurement creates extensive digital evidence:

• submission times;
• bid revisions;
• bidder behaviour;
• tender participation patterns;
• pricing sequences;
• system-generated records.

Consequently, competition authorities can potentially use automated analytics to identify unusual patterns.

But the same digital infrastructure must be designed so that the evidence remains:

• accurate;
• reproducible;
• auditable;
• securely preserved; and
• capable of independent verification.

Case 5 — In Re: Alleged Bid-Rigging in Tenders invited by Department of Printing, Case No. 03/2019 — India

The CCI investigated alleged bid rigging concerning tenders issued by the Department of Printing for printing, packing and dispatch of confidential documents. The case was decided on 12 February 2021.

Relevance

The case illustrates the importance of examining procurement conduct through patterns rather than isolated bid outcomes.

For algorithmic procurement systems, this supports the development of detection models examining:

• repeated bidder combinations;
• recurring winners;
• unusual price gaps;
• bid withdrawals;
• bid rotations;
• supplier participation patterns.

However, algorithmic flags should trigger human investigation rather than automatic punishment.

Case 6 — Eastern Railway Axle Bearings Bid-Rigging Case — India

In the Eastern Railway axle-bearing matter, the CCI found evidence of an arrangement among cartel members involving mutually decided prices in railway tenders during 2015–2019.

The evidence included statements and call-detail records, and the Commission concluded that the parties had engaged in bid rigging/collusive bidding.

Relevance to algorithmic procurement

The case demonstrates why procurement systems should be capable of integrating multiple evidence sources.

An advanced detection system could combine:

Tender Data+Bid Data+Communication Evidence+Historical PatternsTender\ Data + Bid\ Data + Communication\ Evidence + Historical\ Patterns

rather than relying solely on price similarity.

It also demonstrates an important limitation of algorithmic transparency: the algorithm should assist investigators, not replace legal evaluation of evidence.

Case 7 — United States v. David Topkins, No. 3:15-cr-00201 — United States

Although this was not a public-procurement case, it is highly relevant to the algorithmic dimension.

The defendant participated in a price-fixing arrangement involving online sales of posters. The competitors agreed to use pricing algorithms to implement coordinated pricing.

Topkins pleaded guilty to a horizontal price-fixing offence.

Principle

An algorithm does not break the causal chain between the human agreement and the antitrust violation.

In simplified form:

Human agreement → algorithmic implementation → coordinated prices

remains potentially unlawful.

Procurement application

If competing tenderers agree:

"We will use the same algorithmic rules so that Company A wins Tender 1 and Company B wins Tender 2"

the fact that the algorithm executes the arrangement automatically does not make the arrangement lawful.

Case 8 — Eturas and Others v Lietuvos Respublikos konkurencijos taryba, Case C-74/14 — CJEU

This is one of the most important European cases concerning competition law and a digital platform.

The E-TURAS system was an online travel-booking platform used by travel agencies. A system message announced a restriction on discounts and technical modifications were introduced to facilitate implementation of the restriction.

The CJEU held that awareness of the platform communication could support an inference of participation in a concerted practice in appropriate circumstances, although the evidentiary requirements and rights of defence remained important.

Procurement significance

The case illustrates the platform-facilitated coordination problem.

A procurement platform or algorithm provider could potentially become the technological mechanism through which competitors receive common instructions or competitively sensitive information.

Therefore, procurement algorithms should have:

• access controls;
• information segregation;
• audit logs;
• identifiable system messages;
• records of algorithmic changes;
• clear responsibility for platform administrators.

9. Central Legal Principle from the Cases

The cases collectively demonstrate three different situations.

Model 1 — Human cartel + algorithm

Topkins

Competitors agree to coordinate → algorithm implements agreement.

This is the clearest form of algorithmic collusion.

Model 2 — Common digital platform

Eturas

Platform communicates a common restriction → participants may face competition-law consequences depending on knowledge and conduct.

Model 3 — Algorithmic detection of suspicious bidding

Indian procurement cases

Patterns such as identical prices, common mistakes, bid rotation and repeated behaviour may provide evidence of collusion.

Therefore:

Algorithmic evidence can help establish competition-law violations, but an algorithm should not itself become the legal substitute for proving the elements of the violation.

10. Transparency Requirements for Procurement Algorithms

A competition-compliant procurement algorithm should ideally contain the following safeguards.

11. Source-Code Transparency

A difficult question is:

Must the procurement authority disclose the source code?

Not necessarily.

Complete disclosure may itself create risks:

• cybersecurity vulnerabilities;
• gaming of procurement rules;
• manipulation of scoring;
• disclosure of trade secrets;
• reverse engineering;
• strategic adaptation by bidders.

A better model may be controlled transparency.

Layer 1 — Public transparency

Publish:

• purpose;
• major criteria;
• material weights;
• procurement rules;
• appeal mechanisms.

Layer 2 — Bidder transparency

Provide:

• bidder-specific reasons;
• relevant scoring information;
• material factors affecting the result.

Layer 3 — Regulatory transparency

Give competition authorities and auditors:

• source code where necessary;
• datasets;
• audit logs;
• algorithm versions;
• model documentation.

Layer 4 — Judicial transparency

Courts may require deeper technical disclosure where necessary to determine legality.

Thus:

The correct objective is not maximum transparency, but sufficient transparency for effective competition, accountability and review.

12. Algorithmic Transparency and Confidentiality

Procurement law must reconcile two competing interests:

Interest 1 — Transparency

The bidder should understand why it lost.

Interest 2 — Confidentiality

Competitors should not obtain commercially sensitive information.

For example, disclosure of:

"Bidder A submitted ₹100 million and Bidder B submitted ₹102 million"

may be acceptable in some contexts, but systematic publication of detailed bid histories could allow firms to learn each other's strategies and coordinate future tenders.

The OECD specifically warns that transparency must be balanced so that information disclosure does not facilitate collusion.

13. AI-Based Procurement and Competition

Artificial intelligence introduces additional risks because an AI procurement system can:

• learn from historical bids;
• predict competitor behaviour;
• recommend bid prices;
• identify likely winners;
• adjust evaluation parameters;
• automatically reject suppliers;
• optimise procurement outcomes.

The central competition question becomes:

Who controls the algorithm and whose interests are embedded in its design?

An algorithm operated by an independent public authority is different from one controlled by a dominant private procurement platform.

14. Algorithmic Collusion in Procurement

There are several possible models.

A. Messenger model

Competitors expressly agree to collude and use an algorithm to execute their agreement.

Topkins provides the clearest example of this conceptual model.

B. Hub-and-spoke model

A common platform or software provider facilitates coordination among competing suppliers.

Eturas demonstrates the importance of analysing platform-mediated coordination.

C. Autonomous coordination

Competitors independently deploy algorithms that learn from market behaviour and converge on similar outcomes.

This presents more difficult questions concerning the traditional requirement of an agreement or concerted practice. The OECD has recognised that algorithms raise questions concerning traditional concepts of agreement and tacit collusion.

15. Procurement Algorithm Red Flags

A competition authority should investigate circumstances such as:

1. Repeated identical bids.
2. Repeated winner rotation.
3. Competitors alternating successful tenders.
4. Identical bid submission timestamps.
5. Suspicious bid withdrawals.
6. Common algorithmic configurations.
7. Shared software providers.
8. Unusual parallel price movements.
9. Competitors submitting technically identical documents.
10. Identical errors or metadata.
11. Sudden algorithmic changes before tenders.
12. Unexplained exclusion of particular suppliers.
13. Different treatment of similarly situated bidders.
14. Algorithmic access to competitors' confidential information.
15. Procurement officials overriding algorithmic decisions without documented reasons.

These should be treated as investigative indicators rather than automatic findings of infringement.

16. Role of Competition Authorities

Competition authorities should develop technical capacity to examine:

• machine-learning models;
• procurement databases;
• source code;
• APIs;
• system logs;
• metadata;
• automated decision systems;
• data-sharing arrangements.

The need for cooperation between procurement bodies and competition authorities is increasingly recognised internationally. The OECD's 2026 work emphasises institutional cooperation, information exchange and training to identify bid-rigging risks.

17. Indian Competition-Law Framework

For India, the principal statutory provisions are:

Section 3(1)

Prohibits agreements causing or likely to cause an appreciable adverse effect on competition.

Section 3(3)

Deals with horizontal agreements between competitors.

Relevant forms include:

• price fixing;
• limitation of production/supply;
• market allocation;
• bid rigging or collusive bidding.

The CCI expressly identifies bid rigging and collusive bidding within this framework.

Section 4

Potentially becomes relevant where algorithmic procurement involves conduct by a dominant enterprise that results in:

• discriminatory conditions;
• denial of market access;
• unfair conditions;
• exclusionary conduct; or
• leveraging of dominance.

18. Competition-Law Test for a Procurement Algorithm

A useful analytical framework is:

Step 1 — Identify the algorithm

Who designed and operates it?

Step 2 — Identify its function

Is it:

• evaluating bids;
• recommending prices;
• detecting fraud;
• allocating tenders;
• sharing information?

Step 3 — Identify the relevant market

Determine the relevant product/service and geographic market.

Step 4 — Examine market structure

Consider:

• number of suppliers;
• concentration;
• entry barriers;
• repeat interactions;
• switching costs.

Step 5 — Examine data flows

Determine whether competitors obtain:

• prices;
• bid histories;
• capacity information;
• costs;
• future strategies.

Step 6 — Examine algorithmic conduct

Was the algorithm:

• independently designed?
• jointly configured?
• supplied by a common intermediary?
• instructed by competitors?

Step 7 — Examine outcome

Did the system produce:

• exclusion;
• coordination;
• price increases;
• reduced participation;
• market allocation?

Step 8 — Establish legal causation

Distinguish:

algorithmic correlation

from

legally actionable coordination.

Step 9 — Consider justification

Determine whether the conduct has legitimate efficiency or security explanations.

Step 10 — Determine remedy

Possible measures include:

• algorithm modification;
• information firewalls;
• independent audits;
• transparency requirements;
• access remedies;
• procurement redesign;
• compliance programmes;
• penalties where infringement is established.

19. Remedies

Competition authorities and procurement regulators may consider:

Structural remedies

• separating procurement-platform functions;
• preventing a dominant platform from controlling both procurement infrastructure and competing suppliers.

Behavioural remedies

• non-discrimination requirements;
• equal access;
• information firewalls;
• transparency obligations.

Technical remedies

• independent algorithm audits;
• immutable audit logs;
• model documentation;
• version control;
• explainability requirements.

Procedural remedies

• bidder appeals;
• human review;
• independent procurement review;
• competition-authority access to technical records.

20. Relationship Between Procurement Transparency and Competition

The fundamental balance can be expressed as:

Effective Transparency=Accountability+Contestability+Auditability−Unnecessary Competitor Information SharingEffective\ Transparency = Accountability + Contestability + Auditability - Unnecessary\ Competitor\ Information\ Sharing

Too little transparency creates:

• hidden discrimination;
• arbitrary exclusion;
• unverifiable decisions;
• unaccountable algorithms.

Too much transparency can create:

• information symmetry among cartel members;
• strategic adaptation;
• price coordination;
• bid manipulation.

Therefore, competition law should pursue targeted transparency rather than indiscriminate disclosure.

21. Conclusion

Procurement algorithm transparency is becoming an important component of modern competition law because algorithms increasingly determine who participates, how bids are evaluated and how suspicious conduct is detected.

The Indian cases involving public procurement demonstrate that competition authorities already rely heavily on patterns, circumstantial evidence, identical bids, common mistakes, communications and bidding behaviour to investigate bid rigging.

Topkins demonstrates that competitors cannot avoid antitrust liability merely because an algorithm executes their agreement, while Eturas demonstrates the additional evidentiary and legal difficulties created when digital platforms facilitate coordinated conduct.

The emerging legal principle can therefore be stated as follows:

Algorithms are not outside competition law; they are increasingly becoming the infrastructure through which competition is organised, distorted and investigated.

For procurement, the appropriate model is auditable algorithms + controlled transparency + confidentiality protection + human oversight + competition-law monitoring.