Competition Law and Predictive Market Allocation Systems

1. Introduction

Predictive market allocation systems are technological systems that use historical and real-time data, algorithms, artificial intelligence, machine learning, forecasting models, or automated decision-making to predict and influence the allocation of customers, territories, suppliers, products, capacity, resources, or market opportunities.

They may be used for:

customer allocation;

geographic territory allocation;

supplier selection;

procurement;

sales forecasting;

advertising allocation;

logistics;

inventory allocation;

financial markets;

digital marketplaces;

healthcare resource allocation; and

platform-based distribution.

From a competition-law perspective, the important distinction is between a system that independently optimizes a firm's own allocation decisions and one that facilitates coordination between competitors concerning who receives which customers, territories, suppliers, or market opportunities.

The latter can closely resemble traditional market-sharing or customer-allocation arrangements, even where the coordination is implemented through sophisticated technology rather than an explicit human agreement.

2. Meaning of Predictive Market Allocation Systems

A predictive market allocation system can be represented as:

Data → Prediction → Allocation Recommendation → Automated Decision → Market Outcome → New Data

For example, an algorithm could predict that:

Customer A is most likely to purchase from Firm X;

Customer B is likely to purchase from Firm Y;

Region 1 should be served by Supplier A;

Region 2 should be served by Supplier B.

When used internally by one company, this may simply be legitimate optimization.

Competition concerns arise when competing firms use the same system, exchange commercially sensitive information through it, or allow the system to determine which competitors receive particular customers or territories.

3. Market Allocation Under Competition Law

Market allocation is one of the most established forms of anticompetitive conduct.

Traditional allocation arrangements may divide:

geographic territories;

customers;

products;

suppliers;

distribution channels;

contracts;

public procurement opportunities.

For example:

Competitor A agrees to serve northern India while Competitor B serves southern India.

Such an arrangement can eliminate rivalry between the participants.

A predictive system can potentially reproduce this outcome through algorithms.

4. Predictive Allocation Versus Traditional Market Sharing

There is an important distinction.

Traditional arrangement

Company A explicitly agrees with Company B:

“You take these customers; we will take those customers.”

Predictive system

Companies supply commercially sensitive information to a common algorithm that predicts:

“Company A should serve Customer 1; Company B should serve Customer 2.”

More serious scenario

The competing companies knowingly design or use the system to achieve coordinated customer or territorial allocation.

The technology does not necessarily change the underlying economic substance of the conduct.

5. Section 3 of the Indian Competition Act

Under Section 3 of the Competition Act, 2002, agreements that cause or are likely to cause an appreciable adverse effect on competition can be prohibited.

Section 3(3) is particularly important for:

determining prices;

limiting production or supply;

sharing or allocating markets or customers;

bid rigging or collusive bidding.

A predictive market allocation system could therefore become relevant where its use facilitates an arrangement involving:

allocation of markets or customers among competitors.

The existence of sophisticated AI would not by itself immunize the conduct from Section 3.

6. Article 101 TFEU

Under Article 101 TFEU, agreements between undertakings, decisions by associations, and concerted practices that restrict competition can be prohibited.

Market sharing and customer allocation are traditionally treated as particularly serious restrictions.

A technologically mediated allocation system may therefore be examined according to the underlying coordination rather than its technological form.

7. United States Antitrust Law

In the United States, market allocation can constitute a per se unlawful horizontal restraint when competitors agree to divide markets.

The central question is whether there is an agreement or concerted action.

The use of an algorithm does not necessarily eliminate the need to establish the required antitrust agreement.

This makes evidence concerning:

communications;

contractual arrangements;

system design;

data inputs;

implementation;

knowledge;

intent;

particularly important.

8. Case Law 1: United States v. Topco Associates

United States v. Topco Associates, Inc., 405 U.S. 596 (1972)

Topco involved agreements among competitors concerning territorial restrictions.

The Supreme Court treated horizontal market allocation as a particularly serious restraint.

Relevance to predictive systems

A predictive allocation system could theoretically produce an equivalent outcome by algorithmically dividing:

territories;

customers;

sales opportunities.

The important principle is that competition law focuses on the competitive substance of the arrangement, not simply whether the allocation is performed manually or technologically.

9. Case Law 2: Palmer v. BRG of Georgia

Palmer v. BRG of Georgia, Inc., 498 U.S. 46 (1990)

In Palmer, competitors divided geographic territories between themselves.

The Supreme Court treated the arrangement as unlawful horizontal market allocation.

Relevance

A predictive system that enables competitors to systematically avoid competing in particular territories could raise similar concerns if supported by an agreement or concerted practice.

For example:

Algorithm assigns Territory A predominantly to Firm 1 and Territory B predominantly to Firm 2 pursuant to an understanding between them.

The technological mechanism would not necessarily alter the fundamental nature of the arrangement.

10. Case Law 3: United States v. Sealy

United States v. Sealy, Inc., 388 U.S. 350 (1967)

Sealy concerned territorial restrictions involving competitors.

The Supreme Court examined arrangements that limited competition among manufacturers through territorial allocation.

Importance for predictive allocation

Modern predictive systems could potentially implement territorial restrictions more dynamically.

Instead of a written territorial agreement:

“Firm A cannot sell in Territory B,”

the system might automatically redirect customers in Territory B toward Firm A.

Competition analysis would need to determine whether the arrangement produces a similar exclusionary allocation and whether competitors participated in creating or implementing it.

11. Case Law 4: United States v. Container Corporation

United States v. Container Corporation of America, 393 U.S. 333 (1969)

Container Corporation concerned exchanges of competitively sensitive pricing information among competitors.

The case is important because competition concerns can arise not only from explicit agreements on prices but also from mechanisms that reduce uncertainty between competitors.

Relevance

Predictive allocation systems may process:

customer information;

demand forecasts;

capacity information;

future pricing intentions;

supply information.

If competing firms share sensitive information through a common system, the system may potentially facilitate coordination.

The critical issue is whether the information exchange contributes to an unlawful concerted practice.

12. Case Law 5: Interstate Circuit v. United States

Interstate Circuit, Inc. v. United States, 306 U.S. 208 (1939)

Interstate Circuit is an important early case concerning concerted action inferred from coordinated conduct and communications.

It demonstrates that antitrust law does not always require a traditional signed contract between competitors.

Relevance

Predictive systems can create complex evidentiary questions.

For example:

Did competitors know what the algorithm was doing?

Did they communicate with one another?

Did they deliberately configure the system to produce coordinated outcomes?

Did they knowingly participate in the same allocation mechanism?

The answers may be more important than the fact that the allocation was technically automated.

13. Case Law 6: United States v. Apple Inc.

United States v. Apple Inc., 791 F.3d 290 (2d Cir. 2015)

The Apple e-books litigation concerned coordination between Apple and publishers.

The case is important for demonstrating that technological or contractual structures can be scrutinized where they facilitate coordination among market participants.

Relevance

A predictive market allocation platform could potentially act as a coordination mechanism where multiple competitors use it to structure their commercial behaviour.

The legal question would remain whether the evidence establishes the requisite concerted conduct and competitive harm.

14. Case Law 7: Eturas

Case C-74/14, Eturas UAB v Lietuvos Respublikos Konkurencijos Taryba

Eturas is particularly relevant to algorithmically mediated competition.

A common online booking platform transmitted a system message concerning limits on discounts offered by participating travel agencies.

The European Court of Justice addressed the circumstances in which participants using a common platform could potentially be regarded as engaging in a concerted practice.

Importance

Eturas demonstrates a crucial principle for predictive allocation systems:

A digital platform can become a mechanism through which commercially significant coordination is facilitated.

The relevant issue is not simply that multiple businesses use the same software, but whether they knew or could be regarded as participating in a coordinated mechanism.

15. Case Law 8: AC-Treuhand

AC-Treuhand AG v Commission, Joined Cases C-194/14 P

AC-Treuhand concerned the liability of an entity that facilitated cartel activity despite not itself operating as a competitor in the relevant product market.

Relevance to predictive allocation systems

This becomes important where a technology provider:

develops the allocation mechanism;

manages information;

facilitates coordination;

administers the system.

A technology intermediary may therefore attract competition-law scrutiny depending upon its participation in the anticompetitive arrangement.

The case does not mean that software providers are automatically liable for customers' antitrust violations.

The degree of participation remains crucial.

16. Case Law 9: T-Mobile Netherlands

T-Mobile Netherlands BV v Raad van bestuur van de Nederlandse Mededingingsautoriteit, Case C-8/08

The European Court of Justice addressed concerted practices and the exchange of competitively sensitive information.

The case is significant for understanding how information exchange can reduce strategic uncertainty among competitors.

Relevance

A predictive allocation system may make competitor behaviour more predictable.

If competitors receive information about:

customer allocation;

capacity;

future commercial strategy;

territory allocation,

the system could potentially reduce strategic uncertainty.

That can become competition-sensitive where the underlying arrangement satisfies the requirements for a concerted practice.

17. Case Law 10: Wood Pulp

A. Ahlström Osakeyhtiö v Commission, Joined Cases 89/85 etc.

The Wood Pulp litigation is important for understanding concerted practices and parallel behaviour in concentrated markets.

It demonstrates that parallel conduct cannot simply be equated with unlawful coordination.

Importance for predictive systems

This distinction is critical.

Suppose several companies independently use predictive algorithms and reach similar allocation decisions.

That fact alone does not necessarily establish a cartel.

Competition authorities would need evidence concerning:

communication;

common arrangements;

information exchange;

algorithmic design;

contractual relationships;

participation in coordinated mechanisms.

18. Algorithmic Market Allocation

Predictive systems can create several types of allocation.

Customer allocation

The algorithm decides which firm receives which customers.

Geographic allocation

The system determines which competitor serves which geographical region.

Supplier allocation

Suppliers are divided among competing purchasers.

Product allocation

Different competitors specialize in different products.

Capacity allocation

Competitors divide production or service capacity.

Procurement allocation

Competitors effectively divide tenders or contracts.

Each can generate different competition-law questions.

19. Customer Allocation and Digital Platforms

Consider a platform used by competing service providers.

A customer searches for a service.

The platform's algorithm predicts:

Provider A should receive Customer X.

If the algorithm independently optimizes matching based on legitimate factors such as:

location;

availability;

quality;

customer preference;

there may be no antitrust problem.

But suppose competing providers agree:

Provider A receives premium customers while Provider B receives price-sensitive customers.

That may raise market-allocation concerns.

20. Geographic Allocation

Predictive systems can allocate customers based on geographic boundaries.

For example:

Firm A receives Delhi;

Firm B receives Mumbai;

Firm C receives Bengaluru.

If this occurs independently because of logistics efficiencies, it may be legitimate.

If competing firms coordinate their geographic allocation to avoid competing against one another, the legal analysis changes substantially.

21. Predictive Procurement Allocation

Public procurement presents particularly significant risks.

A system could predict:

which competitor is likely to win;

which competitor should submit the lowest bid;

which competitor should refrain from bidding;

which contracts should be allocated to each participant.

Such conduct could potentially amount to bid rigging.

Under Indian competition law, bid rigging and collusive bidding fall within the serious restrictions addressed by Section 3.

22. Algorithmic Bid Rotation

Consider four contractors:

A;

B;

C;

D.

An algorithm predicts the procurement cycle and allocates:

Tender 1 → A;

Tender 2 → B;

Tender 3 → C;

Tender 4 → D.

If this results from coordinated conduct between the firms, it can resemble traditional bid rotation.

The fact that the allocation is generated automatically would not by itself eliminate the underlying competition concern.

23. Predictive Allocation and Information Exchange

A predictive system can function as an information intermediary.

Competitors might submit:

expected capacity;

expected demand;

inventory;

intended territories;

customer priorities.

The system then produces an allocation.

This raises questions about whether competitors are receiving commercially sensitive information about each other's future conduct.

The more granular, current, and strategically significant the information, the greater the potential competition concern.

24. Common Algorithm Problem

Suppose ten competing companies use the same third-party allocation algorithm.

The algorithm observes:

their prices;

inventory;

customer demand;

capacity;

market shares.

It then recommends allocation decisions.

Three legal scenarios must be distinguished.

Scenario 1 — Independent use

Each company independently uses the software.

This does not automatically constitute collusion.

Scenario 2 — Sensitive information sharing

Companies knowingly permit competitors' commercially sensitive information to be incorporated into the system.

Competition concerns become stronger.

Scenario 3 — Deliberate coordinated allocation

Companies use the system to divide customers or territories.

This can potentially resemble traditional market allocation.

25. Hub-and-Spoke Theory

A predictive platform can potentially function as the hub in a hub-and-spoke arrangement.

Conceptually:

Competitor A ↔ Platform ↔ Competitor B ↔ Competitor C

The platform may facilitate:

information exchange;

common pricing;

customer allocation;

territory allocation;

output decisions.

The central legal question is whether there is sufficient evidence of coordination between the participants.

A platform's mere existence between competitors does not automatically establish a hub-and-spoke cartel.

26. Indian Competition-Law Application

For India, the following provisions are particularly relevant.

Section 3(1)

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

Section 3(3)

Particularly relevant where competitors:

allocate markets;

allocate customers;

manipulate bids;

coordinate prices;

limit supply.

Section 4

Could apply where a dominant platform abuses its position through:

discriminatory access;

denial of market access;

leveraging;

tying;

exclusionary allocation;

discriminatory algorithmic treatment.

27. Dominant Predictive Allocation Platforms

Suppose Platform X controls the largest market-allocation infrastructure.

It could potentially:

prioritize affiliated companies;

exclude competitors;

allocate premium customers to its subsidiary;

disadvantage rival suppliers;

restrict access to prediction data.

These practices could raise Section 4 concerns if dominance and abuse are established.

The analysis would require examination of the relevant market and actual competitive effects.

28. Self-Preferencing Through Allocation

An especially important scenario is:

Platform owns a marketplace + operates a predictive allocation engine + competes with marketplace participants.

The platform could potentially configure the allocation engine to favour its own subsidiary.

For example:

more profitable customers;

high-value territories;

priority delivery slots;

better advertising placements.

This resembles the broader self-preferencing concern encountered in digital-platform cases.

29. Discriminatory Algorithms

A dominant allocation platform may apply apparently neutral rules differently to:

independent businesses;

affiliated businesses;

large customers;

small customers.

Competition law may therefore need to examine whether algorithmic discrimination has a legitimate business justification or produces exclusionary effects.

Relevant factors may include:

technical efficiency;

quality;

capacity;

cost;

consumer preferences;

objective performance criteria.

30. Predictive Allocation and Network Effects

Allocation platforms can become more powerful as participation increases.

More participants produce:

More transactions → more data → better predictions → more users → more transactions.

This can produce significant entry barriers.

Competitors may struggle because they lack:

historical datasets;

network information;

customer behaviour;

prediction accuracy.

Thus, predictive allocation can potentially create data-driven market power.

31. Switching Costs

Businesses may become dependent on a predictive allocation platform because it integrates with:

CRM systems;

ERP systems;

payment systems;

logistics;

advertising;

customer databases.

Switching can therefore require:

technical migration;

retraining;

data conversion;

API redesign;

customer reconfiguration.

High switching costs can reinforce incumbent market power.

32. Exclusivity

An allocation platform may impose:

“Participants cannot use another allocation system.”

Exclusive-use clauses can raise competition concerns where they substantially foreclose competing platforms.

The analysis should examine:

duration;

market coverage;

switching costs;

alternatives;

market share;

efficiency justifications.

Exclusivity is not automatically unlawful.

33. Tying

A dominant allocation platform might require customers to purchase:

allocation software + cloud services

or

allocation software + payment services.

Potential tying concerns arise where the platform uses power in one product to restrict competition in another.

34. Predatory Allocation

Another possible theory concerns deliberate allocation of opportunities away from rivals.

For example:

A dominant platform systematically directs low-value customers to independent competitors while retaining high-value customers for its own affiliate.

Alternatively, it might allocate scarce resources in a manner that makes competing businesses commercially unviable.

The competition analysis would need to establish dominance, discriminatory conduct, foreclosure, and the applicable legal requirements.

35. Merger Control

Predictive allocation technology may also influence merger analysis.

A major platform might acquire:

an AI allocation startup;

a procurement algorithm provider;

a customer-matching platform;

a predictive logistics company.

Even a small target could possess important:

algorithms;

datasets;

engineers;

customer relationships;

intellectual property.

Authorities may therefore examine potential competition and innovation effects.

36. Legitimate Uses

Competition law should distinguish anticompetitive coordination from legitimate predictive optimization.

Legitimate applications may include:

Logistics

Predicting delivery routes.

Inventory

Forecasting demand.

Healthcare

Allocating hospital resources.

Agriculture

Allocating irrigation or supply resources.

Energy

Forecasting electricity demand.

Retail

Matching inventory with consumer demand.

Manufacturing

Allocating production capacity.

The use of AI or prediction does not itself create an antitrust violation.

37. Compliance Measures

Businesses using predictive allocation systems should consider:

Avoiding competitor-specific commercially sensitive information where unnecessary.

Establishing clear data-access controls.

Separating competitor information.

Auditing algorithmic recommendations.

Preventing customer or territory allocation among competitors.

Documenting independent decision-making.

Reviewing third-party algorithm contracts.

Monitoring platform self-preferencing.

Establishing competition-law compliance procedures.

Conducting periodic algorithmic competition audits.

38. Competition Audit Questions

A useful compliance framework could ask:

Data

What competitor information enters the system?

Is the information commercially sensitive?

Is it aggregated or individualized?

Algorithm

Who designed the allocation rules?

Can users influence the rules?

Does the system intentionally coordinate competitors?

Governance

Who can access outputs?

Are competitors able to see each other's information?

Allocation

Does the system divide customers?

Does it divide territories?

Does it determine tender participation?

Platform power

Does the platform compete with its users?

Does it favour affiliated businesses?

39. Key Distinction: Prediction Versus Coordination

The most important doctrinal distinction is:

Prediction itself is not market allocation.

A company may legitimately predict that Customer A is likely to purchase Product X.

Competition concerns arise when the predictive mechanism becomes part of an arrangement that coordinates competing firms or forecloses competition.

Thus:

Prediction → generally neutral technology

but potentially:

Prediction + competitor information exchange + coordinated allocation → competition concern

and:

Dominant platform + discriminatory allocation + foreclosure → potential abuse of dominance

40. Overall Legal Framework

41. Conclusion

Predictive market allocation systems create a technologically sophisticated version of several traditional competition-law problems.

Their most significant risks arise where technology is used to:

divide customers;

allocate territories;

rotate procurement opportunities;

exchange competitively sensitive information;

coordinate competitors;

favour affiliated businesses;

foreclose rival platforms; or

leverage control over allocation infrastructure.

The leading authorities—including Topco, Palmer v. BRG, Sealy, Container Corporation, Interstate Circuit, Apple, Eturas, AC-Treuhand, T-Mobile Netherlands, together with Indian competition jurisprudence—demonstrate that competition law generally examines the economic substance and coordination mechanism, rather than allowing technological automation to determine legality.

For India, Sections 3 and 4 of the Competition Act, 2002 provide the principal framework. Section 3 is particularly important for coordinated customer/market allocation, bid rigging and information exchange, while Section 4 becomes relevant where a dominant predictive allocation platform uses its position to discriminate, deny market access, tie services, or leverage power into neighbouring markets.

The central regulatory challenge is therefore to distinguish legitimate predictive optimization from algorithmically facilitated market division and exclusion, while preserving the substantial efficiency benefits that automated allocation can provide.