Competition Law and Competition Implications of Decision Engine Concentration

1. Introduction

Decision engine concentration refers to a market situation in which a small number of undertakings control technologies, platforms, algorithms, artificial-intelligence systems, or data infrastructures that significantly influence economic decisions made by consumers, businesses, governments, or other market participants.

A "decision engine" may include:

Search and ranking algorithms;

Recommendation systems;

AI decision-support systems;

Credit-scoring systems;

Pricing algorithms;

Advertising-selection systems;

Hiring and matching algorithms;

Insurance-risk engines;

Fraud-detection systems;

Navigation and routing systems;

Automated procurement systems.

Competition concerns arise when control over these systems becomes concentrated because the decision engine may become a competitive bottleneck. The undertaking controlling it may influence which products are visible, which suppliers are selected, what prices are offered, or which competitors obtain access to customers.

Importantly, concentration itself is not necessarily unlawful. Competition law generally focuses on market power, anticompetitive agreements, abuse of dominance, mergers, and their effects under the applicable legal regime.

2. Meaning of Decision Engine Concentration

A decision engine can be described as a system that transforms:

Data + rules/algorithms + computing power → recommendation, ranking, selection, or decision.

For example:

Consumer data → algorithm → product recommendations → consumer purchase.

If one undertaking controls the technology through which millions of consumers receive recommendations, that undertaking may exercise influence beyond the market for the underlying software.

Thus, decision-engine concentration concerns control over the mechanism that determines economic choices.

3. Examples of Decision Engines

A. Search engines

Determine which websites or products users encounter first.

B. Recommendation engines

Determine which:

Products;

Videos;

Restaurants;

Services

are recommended to users.

C. Advertising engines

Determine which advertisements consumers see.

D. Pricing engines

Use algorithms to determine or recommend prices.

E. Credit decision engines

Evaluate creditworthiness.

F. Matching engines

Connect:

Drivers and passengers;

Workers and employers;

Buyers and sellers.

G. AI decision systems

Generate recommendations or decisions based upon large datasets.

4. Why Concentration Matters

When decision engines become concentrated, competition can be affected because the system may control access to consumers.

A dominant decision engine could potentially influence:

Market visibility;

Consumer choice;

Supplier selection;

Advertising;

Pricing;

Market entry;

Innovation.

Therefore, the competitive significance of the engine may exceed its apparent software market.

5. Sources of Decision-Engine Concentration

Concentration can develop because of:

Network effects;

Large datasets;

Economies of scale;

High computing costs;

Intellectual property;

Proprietary algorithms;

Cloud infrastructure;

Consumer switching costs;

Vertical integration;

Acquisition of innovative competitors.

6. Data as a Barrier to Entry

Modern decision engines often improve as they receive more data.

This can produce a feedback loop:

More users → more data → better algorithm → better decisions → more users.

A smaller competitor may therefore face difficulty matching the incumbent's performance even if it possesses a technically capable algorithm.

The competitive concern becomes stronger where data is:

Difficult to obtain;

Proprietary;

Frequently updated;

Essential for algorithmic performance.

7. Network Effects

Decision engines can benefit from direct and indirect network effects.

For example:

More consumers → more sellers → more data → better recommendations → more consumers.

This can create substantial advantages for established platforms.

Network effects can therefore reinforce concentration.

8. Economies of Scale

Large decision engines may spread their costs across millions of transactions.

For example, an AI recommendation system may require:

Expensive computing;

Data centres;

Engineers;

Training datasets;

Cybersecurity;

Continuous model development.

A large platform can distribute these costs across a huge user base.

This can create economies of scale that smaller competitors cannot easily replicate.

9. Algorithmic Gatekeeping

A decision engine can function as a gatekeeper.

For example:

A search algorithm determines which businesses consumers see.

or:

An app-store recommendation system determines which applications receive visibility.

The company controlling the decision engine may therefore control access to customers without directly controlling the underlying product market.

10. Self-Preferencing

One important concern is self-preferencing.

A platform may operate:

A decision engine; and

A competing downstream service.

It could potentially favour its own service in:

Search rankings;

Recommendations;

Default settings;

Advertising placement;

Product comparisons.

This can disadvantage competitors that depend upon the platform.

Self-preferencing is not automatically unlawful; the legal assessment depends upon the relevant competition regime and evidence of anticompetitive effects.

11. Decision Engines and Vertical Integration

A decision-engine operator may be vertically integrated.

For example:

Search engine → advertising → marketplace → payment → logistics

or:

Operating system → app store → payment → recommendation engine

Vertical integration can generate efficiencies but may also create opportunities for foreclosure.

12. Refusal of Access

Suppose a decision engine becomes an important distribution mechanism for businesses.

A dominant operator might restrict competitors from accessing:

APIs;

Ranking systems;

Data;

App stores;

Search interfaces;

Payment infrastructure.

Such conduct can potentially raise refusal-to-deal or essential-facility issues under the applicable legal framework.

13. Exclusive Agreements

A decision-engine provider may require customers or suppliers to use its system exclusively.

For example:

A dominant recommendation platform requires merchants not to use competing recommendation engines.

The competition analysis could consider:

Market coverage;

Duration;

Market power;

Switching possibilities;

Entry barriers;

Competitive effects.

14. Tying and Bundling

A dominant company might tie its decision engine to another product.

For example:

Access to an operating system is conditioned upon using the provider's search or recommendation service.

Such arrangements may raise concerns where dominance in one market is used to strengthen another market.

15. Pricing Algorithms and Collusion

Decision engines can create a different competition problem.

Competitors may independently use similar algorithms to set prices.

If algorithms merely respond independently to market conditions, their use does not automatically create a cartel.

However, competition concerns may arise where competitors:

Agree to use a common pricing system;

Exchange competitively sensitive information through the system;

Coordinate prices;

Design algorithms specifically to facilitate collusion.

Thus, competition law must distinguish algorithmic interdependence from unlawful coordination.

16. Tacit Coordination

Highly transparent algorithmic markets can potentially make it easier for firms to observe and respond rapidly to competitors.

For example:

Firm A changes price → algorithm detects change → Firm B immediately responds.

Repeated interactions may reduce incentives to compete aggressively.

However, the mere existence of parallel algorithmic pricing does not by itself establish an unlawful agreement.

17. Consumer Choice

Decision engines can affect consumer choice without increasing prices.

They may influence:

Which products consumers see;

Which sellers receive attention;

Which advertisements appear;

Which services are recommended.

Therefore, competition authorities may consider quality, choice, innovation and visibility, in addition to price.

18. Important Case Law

The following cases do not all concern "decision engines" by name. They provide established competition-law principles that are particularly relevant to concentrated search, platform, algorithmic and decision-making systems.

Case 1: United States v. Microsoft Corp. (2001)

Facts

Microsoft held a dominant position in PC operating systems and engaged in conduct relating to Internet Explorer and competing browser technologies.

Principle

A dominant undertaking can face antitrust liability where it uses exclusionary conduct to protect or extend its dominant position.

Relevance to decision engines

An operating system can function as a technological environment through which users access other services.

Similarly, a dominant decision engine could potentially use control over its platform to disadvantage competing decision technologies.

19. Case 2: Google Shopping — European Commission

The European Commission's Google Shopping decision concerned Google's treatment of its comparison-shopping service in search results.

Principle

The case illustrates how a dominant search platform can affect competition through the organisation and presentation of information.

Decision-engine relevance

Search ranking is itself a decision engine.

The case therefore demonstrates why control over algorithmically generated visibility can have competitive significance.

20. Case 3: Google Android — European Commission

The Google Android proceedings examined Google's conduct involving the Android ecosystem and related services.

Principle

A dominant platform may face competition concerns where arrangements involving one technological layer reinforce its position in connected markets.

Decision-engine relevance

A decision engine embedded within an operating-system or platform ecosystem may influence competition in:

Search;

Applications;

Advertising;

Payment;

Other digital services.

21. Case 4: United Brands v Commission (1978)

Principle

The European Court of Justice's decision is a foundational authority concerning dominance and abuse of dominance.

A dominant undertaking possesses a position of economic strength that can enable it to behave to an appreciable extent independently of competitors, customers and consumers.

Relevance

A highly concentrated decision-engine market may potentially create comparable market power if users and businesses lack effective alternatives.

However, concentration alone does not establish abuse.

22. Case 5: Bronner v Mediaprint (1998)

Principle

The case established important conditions concerning refusal by a dominant undertaking to provide access to infrastructure.

Relevance to decision engines

If a decision engine becomes an indispensable technological gateway for market participation, competitors may seek access to it.

The case demonstrates why competition law generally approaches compulsory-access claims cautiously.

23. Case 6: Intel v Commission

The Intel litigation concerned rebates offered by a dominant undertaking.

Principle

Conditional commercial arrangements by dominant firms can raise concerns where they are capable of restricting competitors' access to the market.

Relevance

A decision-engine operator could potentially use discounts, preferential access or commercial incentives to make customers dependent upon its ecosystem.

The analysis must focus on the applicable legal test and competitive effects.

24. Case 7: Google Search / Google Advertising Proceedings

Competition authorities in different jurisdictions have examined Google's conduct concerning search and online advertising.

Relevance

These proceedings demonstrate the competition significance of controlling the technological systems that determine:

Search visibility;

Advertising placement;

Matching of advertisers and consumers;

Data used for advertising.

The decision engine may therefore become an important competitive bottleneck.

25. Case 8: Amazon Marketplace Investigation

European competition authorities examined Amazon's use of non-public seller information obtained through its marketplace.

Principle

A platform may possess strategically valuable information concerning businesses that depend upon its marketplace.

Decision-engine relevance

Where the same undertaking controls:

data → algorithm → marketplace → competing products

it may obtain a substantial informational advantage.

This is particularly significant where algorithms use marketplace data to influence product selection or ranking.

26. Indian Competition-Law Framework

In India, decision-engine concentration may potentially be examined under the Competition Act, 2002.

The principal areas are:

Section 3

Anti-competitive agreements.

Section 4

Abuse of dominant position.

Sections 5 and 6

Combinations and merger control.

The Competition Commission of India (CCI) has examined several digital-platform issues involving data, platform access and market power.

27. CCI v. Steel Authority of India Ltd. (2010)

The Supreme Court's decision is important concerning the statutory framework governing CCI proceedings.

Relevance

Decision-engine disputes may involve other regulatory regimes, but competition intervention must remain within the authority granted by the Competition Act.

28. Excel Crop Care Ltd. v. CCI (2017)

The Supreme Court considered cartel conduct and the approach to penalties.

Relevance

The case demonstrates that technological mechanisms do not remove traditional prohibitions on collusion.

If competing businesses use a common decision system to coordinate prices pursuant to an agreement, ordinary cartel principles may become relevant.

29. CCI v. Bharti Airtel Ltd. (2019)

The Supreme Court examined the relationship between competition law and sector-specific regulation.

Relevance

Decision engines may operate in highly regulated sectors such as:

Finance;

Telecommunications;

Insurance;

Healthcare;

Transportation.

Competition authorities may therefore need to coordinate their analysis with specialised regulatory frameworks.

30. Market Definition in Decision-Engine Markets

Market definition may be particularly difficult.

A decision engine may be used across multiple sectors.

For example:

AI engine → finance + insurance + healthcare + retail

The relevant question is whether these uses constitute:

One market;

Several separate markets;

Complementary markets;

A platform ecosystem.

Traditional market-definition tools may need to be supplemented by analysis of:

Data;

Functionality;

User behaviour;

Switching;

Multi-homing;

Innovation.

31. Multi-Sided Decision Engines

Many decision engines connect different groups.

For example:

Consumers ↔ recommendation engine ↔ sellers

or:

Advertisers ↔ advertising algorithm ↔ consumers

The platform may provide one side with free services while charging the other side.

Competition analysis must therefore consider the interaction between the different sides.

32. Data Advantages

A concentrated decision engine can benefit from enormous quantities of:

Search data;

Consumer data;

Transaction data;

Behavioural data;

Location data;

Advertising data.

This can create an important competitive advantage.

A new entrant may have a good algorithm but insufficient data to achieve comparable performance.

33. Artificial Intelligence and Decision-Engine Concentration

AI could intensify concentration because sophisticated AI systems require:

Large datasets;

Computing resources;

Advanced chips;

Cloud infrastructure;

Skilled personnel;

Capital.

Consequently, competition may become concentrated at several levels:

Data → Compute → Models → Applications → Distribution.

Control over multiple layers can create significant ecosystem power.

34. Vertical Integration in AI

Suppose Company A controls:

Cloud infrastructure;

AI model;

Recommendation engine;

Consumer platform.

It could potentially favour its own AI applications over competing systems.

Competition authorities may therefore need to examine vertical foreclosure.

35. Merger Control

Acquisitions can strengthen decision-engine concentration.

A dominant company might acquire:

An AI start-up;

A recommendation engine;

A data analytics company;

A competing search system;

A specialised decision technology.

Authorities may examine whether the transaction removes an important potential competitor.

36. Killer Acquisitions

A "killer acquisition" generally refers to an acquisition of an emerging competitor that might otherwise develop into a significant competitive threat.

In decision-engine markets, this could be particularly important because an innovative start-up may initially have:

Few customers;

Limited revenue;

A small workforce;

but possess valuable technology.

Traditional market-share analysis may therefore underestimate the competitive significance of the acquisition.

37. Interoperability

Interoperability can reduce concentration.

Possible mechanisms include:

Standardised APIs;

Data portability;

Technical compatibility;

Open interfaces;

Cross-platform functionality.

However, compulsory interoperability must also account for:

Security;

Privacy;

Intellectual property;

Technical feasibility.

38. Transparency and Explainability

A decision engine may be commercially valuable precisely because its algorithm is proprietary.

Competition law does not ordinarily require every algorithm to be publicly disclosed.

Nevertheless, where algorithmic conduct is central to an abuse investigation, authorities may need access to sufficient information to understand:

How rankings operate;

How recommendations are generated;

Whether discrimination occurs;

Whether rivals are systematically disadvantaged.

39. Algorithmic Discrimination

A dominant decision engine could potentially rank equivalent businesses differently.

Competition concerns may arise where discriminatory treatment:

Favors the platform's own services;

Excludes rivals;

Raises competitors' costs;

Reduces their ability to reach customers.

Objective and technically justified differences may nevertheless be legitimate.

40. Decision Engines and Essential Facilities

A decision engine may potentially become an important gateway where competitors cannot effectively reach customers without it.

An essential-facility analysis may consider:

Whether the facility is genuinely indispensable;

Whether duplication is feasible;

Whether access is technically possible;

Whether refusal eliminates effective competition;

Whether objective justification exists.

These requirements are applied cautiously in competition law.

41. Privacy and Competition

Privacy may become a competitive parameter.

For example:

Platform A: extensive data collection
Platform B: stronger privacy protections

If consumers cannot realistically switch because of lock-in, competition on privacy quality may weaken.

Competition authorities may therefore consider privacy as a non-price competitive parameter, where legally relevant.

42. Consumer Autonomy

Concentrated decision engines can influence consumer autonomy by controlling:

Defaults;

Recommendations;

Rankings;

Search results;

Personalisation.

The competition issue arises when such control is used in a manner that restricts competitive alternatives.

Consumer-protection law may also independently apply.

43. Potential Competitive Benefits

Decision-engine concentration can also produce legitimate efficiencies.

Large-scale systems may provide:

Better recommendations;

Lower transaction costs;

Faster decisions;

Fraud detection;

Improved matching;

Lower prices;

Greater convenience.

Competition law should therefore distinguish between efficiency-producing scale and exclusionary use of market power.

44. Challenges for Competition Authorities

1. Rapid technological development

Algorithms change continuously.

2. Lack of transparency

Proprietary systems may be difficult to investigate.

3. Zero-price services

Traditional price analysis may be inadequate.

4. Data complexity

Competitive advantages may depend on datasets rather than physical assets.

5. Network effects

Markets can tip quickly.

6. Innovation competition

Future competitors may be more important than current market shares suggest.

45. Regulatory Approach

Competition authorities may consider:

Merger review;

Abuse-of-dominance investigations;

Monitoring exclusionary agreements;

Data-access issues;

Interoperability;

Platform neutrality;

Algorithmic conduct;

Self-preferencing;

Exclusive arrangements;

Potential competition.

The appropriate legal response depends upon the jurisdiction and evidence.

46. Examination-Oriented Analysis

A useful framework is:

Step 1 — Identify the decision engine

What algorithm, platform or AI system makes the relevant decisions?

Step 2 — Identify users

Who depends upon it?

Step 3 — Define the relevant market

Is it a product market, platform market, data market or ecosystem?

Step 4 — Measure market power

Consider:

Market share;

Network effects;

Data;

Switching costs;

Entry barriers.

Step 5 — Examine conduct

Look for:

Self-preferencing;

Exclusivity;

Tying;

Refusal of access;

Discriminatory access;

Predatory conduct;

Collusion.

Step 6 — Examine effects

Assess:

Prices;

Choice;

Quality;

Innovation;

Market entry;

Consumer welfare.

Step 7 — Consider efficiencies

Determine whether the conduct creates legitimate benefits that are recognised under the applicable competition regime.

47. Quick Revision Table

48. Conclusion

Decision-engine concentration represents an emerging competition-law issue because economic power increasingly lies not only in controlling products, infrastructure or distribution, but also in controlling the systems that determine which economic choices are presented to market participants.

The major competition implications include:

Concentration of data;

Algorithmic gatekeeping;

Self-preferencing;

Ecosystem leverage;

Network effects;

Switching costs;

Exclusionary access restrictions;

Algorithmic coordination;

Vertical foreclosure;

Acquisition of potential competitors;

Reduced innovation.

The principles developed in Microsoft, Google Shopping, Google Android, United Brands, Bronner, Intel, Amazon-related proceedings, SAIL, Excel Crop Care and Bharti Airtel provide useful foundations for analysing these markets.

The key legal distinction is between efficient scale and unlawful exclusion. A large or technologically superior decision engine is not, merely because of its size, an infringement of competition law. The relevant question is whether its market power and conduct satisfy the legal requirements for an anticompetitive agreement, abuse of dominance, or problematic combination under the applicable jurisdiction.