Competition Law and Forensic Analytics in Ecosystem Markets

1. Introduction

Forensic analytics in competition law refers to the systematic collection, reconstruction, testing and analysis of large datasets—such as transaction records, prices, rankings, clicks, search results, seller data, communications, algorithmic logs and internal business records—to determine whether market conduct has restricted competition.

Its importance is particularly high in ecosystem markets, where one undertaking operates across several interconnected layers—for example:

• operating system → app store → payment system;
• search engine → advertising → vertical services;
• marketplace → logistics → payments → advertising;
• cloud platform → software → data;
• digital wallet → financial services → merchant network;
• social network → advertising → data analytics.

The central competition-law problem is that conventional evidence may not reveal how an ecosystem actually functions. A platform may formally treat all businesses alike while its ranking algorithm, recommendation engine, access rules or data architecture produces systematically different outcomes.

The UK CMA has specifically recognised that increasingly sophisticated algorithms can make anti-competitive effects harder to identify, while also developing analytical methods for examining algorithmic competition issues.

2. Meaning of Forensic Analytics

Forensic analytics combines:

1. Digital evidence collection
2. Statistical analysis
3. Econometric testing
4. Algorithmic auditing
5. Network analysis
6. Text and communication analysis
7. Transaction reconstruction
8. Data lineage analysis
9. Event-study techniques
10. Counterfactual modelling

The objective is not merely to discover that a platform has a large market share.

Instead, it asks:

How did the platform's conduct affect competitive conditions, rivals, suppliers, consumers and other participants in the ecosystem?

3. Why Ecosystem Markets Require Forensic Analytics

Traditional competition analysis often concentrates on:

• market definition;
• market shares;
• prices;
• costs;
• output;
• barriers to entry.

Ecosystem markets require additional examination because competition may occur through:

• data;
• interoperability;
• default settings;
• rankings;
• recommendation systems;
• access to APIs;
• switching costs;
• network effects;
• cross-subsidisation;
• self-preferencing;
• tying;
• exclusive access;
• algorithmic discrimination.

A platform may therefore exercise market power without simply charging an excessive price.

For example, a marketplace could manipulate search rankings so that independent sellers become less visible while the platform's own products receive preferential placement. The relevant evidence may exist only within millions of ranking events and algorithmic logs.

4. Principal Areas of Forensic Analytics

A. Transaction Analytics

Investigators can reconstruct:

• purchases;
• prices;
• discounts;
• commissions;
• seller fees;
• advertising expenditure;
• delivery charges;
• consumer switching;
• transaction volumes.

This can reveal whether allegedly neutral rules systematically disadvantage particular competitors.

Example

If 10,000 sellers use a marketplace and the platform changes its ranking algorithm, investigators can compare:

Before algorithmic change → after algorithmic change

and examine:

• visibility;
• clicks;
• conversion rates;
• sales;
• commissions;
• advertising costs.

5. Ranking and Search Analytics

Search ranking is particularly important in ecosystem markets.

Forensic analysts may reconstruct:

• search queries;
• ranking positions;
• impressions;
• clicks;
• sponsored placements;
• algorithmic scores;
• product visibility;
• user engagement.

A useful test is:

Are similarly situated third-party products systematically ranked below the platform's own products?

This issue is central to self-preferencing investigations.

The European Commission's 2026 Google Search DMA decision, for example, found that Google gave preferential treatment to its own services—including shopping, hotels, transport and sports results—in Google Search.

6. Algorithmic Price Analytics

Forensic analytics can reconstruct prices over time and determine whether pricing algorithms:

• react automatically to competitors;
• rapidly match competitor prices;
• maintain supra-competitive prices;
• discriminate between consumers;
• coordinate indirectly;
• punish discounting;
• facilitate tacit coordination.

This is especially relevant in:

• e-commerce;
• airline ticketing;
• hotels;
• ride-hailing;
• food delivery;
• online advertising;
• financial platforms.

The CMA has noted that algorithms can affect market outcomes and that algorithmic/AI systems create new questions concerning coordination and competition.

7. Data Forensics

Data itself can become a source of competitive advantage.

Investigators can determine:

• what data a platform collects;
• where the data originates;
• which business units receive it;
• whether competitors receive equivalent information;
• whether seller data is used by the platform's competing business;
• whether data is combined across services;
• whether data access is discriminatory.

This is particularly significant where the platform acts simultaneously as:

intermediary + competitor.

8. Amazon Marketplace as a Major Example

The European Commission's Amazon Marketplace investigation is particularly important.

The Commission's 2020 Statement of Objections examined whether Amazon used non-public data generated by third-party sellers on its marketplace to benefit Amazon's competing retail business.

The data allegedly assisted Amazon Retail in decisions concerning:

• products to launch;
• suppliers;
• wholesale terms;
• inventories;
• prices.

The Commission therefore had to examine large-scale internal marketplace data and the relationship between Amazon's marketplace and retail operations.

This illustrates the forensic question:

Was information generated by competitors being converted into an advantage for the platform's competing business?

9. Buy Box Analytics

The Amazon Buy Box investigation provides another forensic-analytics model.

Investigators can examine:

• which seller wins the Buy Box;
• price;
• delivery speed;
• fulfilment method;
• Prime status;
• seller history;
• inventory;
• customer metrics.

The European Commission's preliminary assessment examined concerns that Amazon could favour its own retail offers and offers using Amazon's fulfilment service when determining Buy Box placement and Prime eligibility.

Forensic analysis can therefore determine whether the algorithm's outcome is explained by legitimate quality criteria or by discriminatory treatment.

10. Network Analysis

Ecosystem markets are often best understood as networks rather than isolated markets.

A forensic network model may contain:

Platform

↓

Consumers
Sellers
Advertisers
Developers
Payment providers
Logistics providers
Data suppliers
Competitors

Investigators can then identify:

• central nodes;
• bottlenecks;
• dependency relationships;
• exclusive connections;
• switching pathways;
• cross-market leverage.

This is particularly useful for assessing ecosystem foreclosure.

11. Communication Forensics

Competition authorities may examine:

• emails;
• messaging applications;
• internal presentations;
• meeting records;
• executive communications;
• pricing instructions;
• algorithm-development documents.

Text analytics can identify recurring terms such as:

• "match competitor";
• "exclude";
• "penalise";
• "bury";
• "block";
• "prefer";
• "Prime";
• "default";
• "lock-in."

Communication evidence can be combined with quantitative evidence to determine whether an observed market pattern resulted from legitimate competition or deliberate exclusion.

12. Six Important Case Laws

Case 1: Google Search (Shopping) — European Commission / General Court

Google Search (Shopping), Case AT.39740

The case concerned Google's treatment of competing comparison-shopping services within its search results.

The European Commission concluded that Google systematically favoured its own comparison-shopping service and demoted competing services.

The General Court substantially upheld the Commission's decision, while refining certain aspects of the legal reasoning.

Forensic-analytics relevance

The case demonstrates the importance of analysing:

• search-result positions;
• traffic;
• visibility;
• algorithmic treatment;
• click-through rates;
• comparison between Google's own service and competing services.

Competition principle

Algorithmic ranking can constitute an important instrument of self-preferencing and foreclosure where it disadvantages competitors.

13. Case 2: Google Android — European Commission / General Court

Google Android, Case AT.40099

The Android proceedings concerned Google's conduct relating to Android devices, including restrictions concerning search, browsers and application distribution.

The case illustrates the importance of examining an ecosystem consisting of:

OS → app distribution → search → browser → advertising

rather than examining each product entirely in isolation.

Forensic relevance

Evidence may include:

• default settings;
• pre-installation agreements;
• device activation;
• search-choice behaviour;
• application downloads;
• switching patterns;
• contractual restrictions.

The resulting analysis can establish whether conduct at one ecosystem layer strengthens market power at another.

14. Case 3: Microsoft — EU Commission

Microsoft, COMP/C-3/37.792

Microsoft's competition proceedings concerned, among other matters, interoperability and the relationship between Microsoft's operating-system position and adjacent software markets.

Forensic relevance

The case demonstrates the importance of analysing:

• interoperability;
• technical interfaces;
• documentation;
• API access;
• product architecture;
• compatibility;
• competitor access.

Forensic technical analysis is therefore not limited to financial data.

It can involve software architecture and technical evidence.

15. Case 4: Amazon Marketplace — European Commission

Amazon Marketplace, AT.40462

This is one of the clearest examples of ecosystem-oriented data analysis.

The Commission examined whether Amazon used non-public third-party seller data to compete against those sellers through Amazon Retail.

Forensic evidence potentially relevant

• seller-level sales data;
• product-level information;
• supplier information;
• inventory;
• pricing;
• demand forecasts;
• Amazon Retail product decisions;
• internal access logs.

The case illustrates a crucial forensic principle:

Data generated on one side of a platform may become strategically valuable on another side.

16. Case 5: United States v. Amazon

The FTC and 17 state attorneys general sued Amazon in 2023, alleging that Amazon used interconnected strategies to maintain monopoly power in online retail and marketplace services. The complaint included allegations concerning search visibility, seller pricing and fulfilment arrangements.

The complaint specifically alleged that Amazon could bury sellers in search results when they offered lower prices elsewhere.

Forensic relevance

Such allegations can be tested quantitatively through:

• ranking changes;
• price changes;
• seller visibility;
• sales volumes;
• Buy Box outcomes;
• advertising expenditure;
• fulfilment arrangements.

This demonstrates how algorithmic evidence can connect conduct with competitive effects.

17. Case 6: United States v. Apple

United States v. Apple Inc., 2024

The U.S. Department of Justice's antitrust case against Apple concerns alleged restrictions surrounding the iPhone ecosystem.

The broader forensic model involves examining:

• APIs;
• interoperability;
• app distribution;
• payment functionality;
• messaging;
• smart-device integration;
• switching;
• developer access.

Forensic significance

An ecosystem can produce competitive effects through technical restrictions rather than conventional pricing conduct.

Forensic investigators therefore need to analyse the architecture of the ecosystem itself.

18. Case 7: Intel — European Commission

Intel, Case COMP/C-3/37990

Intel concerned alleged exclusionary practices involving rebates offered to computer manufacturers and retailer channels.

Although it predates today's sophisticated ecosystem analytics, it is important because competition analysis increasingly depends upon reconstructing:

• transaction-level rebates;
• customer coverage;
• conditionality;
• rival access;
• foreclosure effects.

Forensic lesson

Large datasets can be used to distinguish an ordinary quantity discount from a scheme capable of excluding rivals.

19. Case 8: Booking.com — European Commission / National Competition Authorities

The Booking.com parity-practices litigation and enforcement history illustrates the relevance of platform contract and pricing data.

Forensic examination can compare:

• hotel prices across platforms;
• parity clauses;
• ranking;
• commission rates;
• consumer traffic;
• booking volumes.

This allows investigators to determine whether contractual restrictions alter competitive conditions between booking platforms.

20. Forensic Analytics and Abuse of Dominance

Under Article 102 TFEU, Section 2 Sherman Act principles, and comparable national regimes, forensic analytics can assist in establishing:

Dominance

• market share;
• user base;
• network effects;
• switching costs;
• data advantages;
• ecosystem integration.

Conduct

• self-preferencing;
• tying;
• discriminatory access;
• exclusionary rebates;
• refusal of interoperability;
• data exploitation;
• predatory conduct.

Effects

• rival foreclosure;
• reduced innovation;
• increased prices;
• reduced quality;
• reduced choice;
• higher entry barriers.

21. Forensic Analytics and Cartels

Algorithms create a different evidentiary problem.

Suppose competing firms use similar pricing software.

The investigator must distinguish:

Legitimate independent adaptation

from

unlawful coordination.

Analytics may examine:

• synchronised price movements;
• reaction times;
• price dispersion;
• communication patterns;
• algorithm changes;
• common software suppliers;
• competitor-information inputs.

The existence of parallel algorithmic pricing does not by itself establish a cartel.

Additional evidence concerning agreement, concerted practice or legally relevant coordination is necessary depending upon the jurisdiction.

The CMA's recent work specifically recognises continuing debate concerning algorithmic collusion and notes both theoretical and empirical research in the field.

22. Forensic Analytics and Merger Control

Forensic techniques can also be used in merger investigations.

Authorities can analyse:

• customer-level transactions;
• diversion ratios;
• switching;
• bidding records;
• internal documents;
• product overlap;
• innovation pipelines;
• datasets;
• platform usage.

This is particularly important in ecosystem acquisitions, where the target may have modest current revenue but strategically valuable:

• data;
• technology;
• users;
• developers;
• APIs;
• innovation capabilities.

23. Data-Driven Market Definition

Traditional market definition may not adequately capture ecosystem competition.

Forensic datasets can examine:

• actual switching;
• consumer substitution;
• multi-homing;
• user journeys;
• transaction flows;
• cross-platform activity.

For example:

Platform A → 70% exclusive users
Platform B → 80% multi-homing users

This could have significantly different implications for competitive constraints than headline market shares alone suggest.

24. Consumer-Level Analytics

Consumer data can reveal:

• switching rates;
• retention;
• churn;
• search behaviour;
• default usage;
• multi-homing;
• response to price changes;
• response to ranking changes.

A particularly useful concept is the natural experiment.

If a platform changes an algorithm for only one group of users, investigators may compare:

Treatment group

with

Control group

to estimate the causal impact of the change.

25. Algorithm Auditing

Algorithmic audits may investigate:

Input

What information does the algorithm receive?

Processing

What variables and weights are applied?

Output

What ranking, recommendation or price results?

Feedback

Does the output itself become future input?

This final point is critical.

An ecosystem can develop a feedback loop:

More users → more data → better algorithm → better service → more users

At a certain point, this may reinforce incumbent market power.

26. Feedback-Loop Analysis

Forensic analytics can measure:

Users→Data→Algorithmic Improvement→Quality→More UsersUsers \rightarrow Data \rightarrow Algorithmic\ Improvement \rightarrow Quality \rightarrow More\ Users

A second loop may be:

More Sellers→More Products→More Consumers→More SellersMore\ Sellers \rightarrow More\ Products \rightarrow More\ Consumers \rightarrow More\ Sellers

These are legitimate network effects in many circumstances.

Competition law becomes concerned where the undertaking uses the resulting position to exclude competitors rather than merely compete on the merits.

27. Digital Markets Act and Forensic Analytics

The EU Digital Markets Act increasingly complements conventional Article 102 analysis.

Recent Commission enforcement illustrates why technical and data analytics are becoming important.

In July 2026, the Commission announced fines against Google concerning Search self-preferencing and Google Play steering.

The Commission has also adopted measures concerning sharing of Google Search data with competing search services and interoperability of AI services with Android.

These developments demonstrate a move toward technical, data-oriented competition regulation.

28. Evidentiary Chain in Forensic Competition Analysis

A robust investigation should establish:

1. Data source

↓

2. Data integrity

↓

3. Data processing

↓

4. Analytical methodology

↓

5. Identified conduct

↓

6. Counterfactual

↓

7. Competitive effect

↓

8. Causal relationship

↓

9. Legal qualification

This is essential because statistical correlation alone does not establish an infringement.

29. Counterfactual Analysis

The most important forensic question is often:

What would have happened without the allegedly anti-competitive conduct?

Possible counterfactuals include:

• ranking without self-preferencing;
• prices without algorithmic restrictions;
• sales without exclusivity;
• market access without API restrictions;
• user switching without default settings;
• seller performance without discriminatory ranking.

The stronger the counterfactual, the stronger the causal inference.

30. Challenges

A. Black-box algorithms

Authorities may not have complete access to proprietary algorithms.

B. Data volume

Platforms can generate billions of observations.

C. Data quality

Incomplete or inconsistent records can distort results.

D. Dynamic markets

Algorithms may change continuously.

E. Privacy

Personal data may be subject to GDPR and other privacy laws.

F. Causation

Correlation between algorithmic conduct and competitive harm does not automatically establish causation.

G. False positives

An algorithm may produce unequal outcomes because of legitimate quality differences.

H. Reproducibility

Investigators must be able to explain and reproduce the analytical methodology.

31. Remedies Informed by Forensic Analytics

Forensic evidence can help design targeted remedies, including:

• interoperability obligations;
• API access;
• data portability;
• non-discrimination requirements;
• ranking transparency;
• separation of datasets;
• restrictions on use of competitor data;
• monitoring;
• algorithmic auditing;
• structural separation;
• behavioural commitments.

The remedy should address the identified mechanism of competitive harm rather than merely the existence of market power.

32. Forensic Analytics Framework

33. Key Legal Principles Emerging

The major lessons from the case law are:

1. Algorithmic conduct can constitute competition-relevant conduct.
2. Ranking systems can be evidence of self-preferencing.
3. Data generated by business users can become strategically significant.
4. Technical interoperability can be a competition issue.
5. Ecosystem power can extend across connected markets.
6. Statistical evidence must be connected to legally relevant conduct.
7. Correlation alone is insufficient to establish an infringement.
8. Internal documents and technical evidence can corroborate quantitative findings.
9. Counterfactual analysis is essential for effects-based theories.
10. Competition authorities increasingly require technical expertise alongside traditional economics and legal analysis.

34. Conclusion

Forensic analytics has become a central evidentiary methodology for competition law in ecosystem markets. Traditional market-share analysis can identify potential market power, but it may not explain how an ecosystem actually exercises that power.

The combination of:

transaction data + algorithmic logs + ranking data + communications + network analysis + econometrics + technical evidence

can reconstruct the competitive mechanism.

The Amazon Marketplace proceedings illustrate how platform-generated seller data can become relevant to competitive analysis. The Google Search proceedings illustrate the significance of ranking and visibility data, while the Amazon U.S. litigation demonstrates the increasing role of allegations concerning algorithmic ranking, seller visibility and platform practices.

Accordingly, in modern ecosystem markets, the decisive question is increasingly not merely "How large is the platform?", but:

"What does the platform's data, code, algorithm and ecosystem architecture actually do to the competitive process?"

That is precisely where forensic analytics connects competition economics, digital evidence, computer science and antitrust law.