Mandatory Logging of Ranking and Recommendation Decisions

1. Introduction

Mandatory logging of ranking and recommendation decisions refers to a regulatory or competition-law requirement that digital platforms, search engines, app stores, marketplaces, social-media services, recommender systems and AI-driven intermediaries maintain an auditable record of how their algorithms ranked, recommended, demoted, filtered or otherwise prioritized content, products, sellers, applications or services.

The objective is not necessarily to require disclosure of the entire algorithm or source code. Rather, the obligation is to preserve sufficient information to determine what decision was made, when it was made, which inputs materially influenced it, what rules or model version were operative, and whether the outcome was subsequently altered.

This becomes particularly important where ranking itself constitutes a source of market power. A dominant platform may influence consumer demand by determining:

• which products appear first;
• which sellers receive visibility;
• which applications are prominently displayed;
• which search results are recommended;
• which advertisements receive exposure;
• which competitors are demoted;
• which content is amplified;
• which users are shown particular offers; and
• how algorithmic recommendations affect switching and multi-homing.

Logging therefore converts an otherwise opaque algorithmic process into an ex post verifiable decision trail.

2. Meaning of Ranking and Recommendation Decisions

A ranking decision determines the relative position of items in an interface.

For example:

Product A → position 1
Product B → position 2
Product C → position 3

A recommendation decision goes further by determining whether an item should be suggested to a particular user or audience at all.

Examples include:

• “Recommended for you”;
• “Top result”;
• “Customers also bought”;
• “Suggested application”;
• “Featured seller”;
• “Most relevant”;
• “Trending”;
• “People you may know”; and
• AI-generated commercial recommendations.

The legal significance arises because ranking and recommendation may determine access to demand.

3. What Mandatory Logging Would Require

A sophisticated logging regime could require platforms to preserve at least the following information.

A. Timestamp

The precise date and time of the ranking or recommendation.

B. Algorithm or model version

The platform should record which:

• ranking model;
• recommendation model;
• rules engine;
• experiment;
• policy;
• configuration; or
• algorithmic version

was operational.

C. Material input categories

The system should record relevant categories of inputs, such as:

• relevance;
• price;
• quality;
• consumer history;
• seller performance;
• advertising status;
• commissions;
• conversion rates;
• engagement;
• availability;
• contractual status; and
• platform-specific commercial incentives.

This need not necessarily disclose commercially sensitive raw data.

D. Ranking outcome

The system should preserve the resulting:

• rank;
• recommendation;
• demotion;
• exclusion;
• visibility score; or
• placement.

E. Reason codes

Platforms could be required to maintain machine-readable explanations such as:

“rank increased because of relevance score”

or

“seller demoted because minimum service-level threshold was not satisfied.”

F. Human intervention

Where employees override an automated ranking decision, the intervention should be recorded.

G. Commercial influence

Particular importance attaches to whether ranking was affected by:

• advertising;
• preferential contracts;
• platform commissions;
• payments;
• exclusivity;
• self-preferencing;
• rebates; or
• strategic commercial relationships.

4. Why Logging Matters in Competition Law

Algorithmic conduct presents a major evidentiary problem.

Traditional antitrust investigations frequently rely upon:

• emails;
• meeting records;
• contracts;
• pricing documents;
• internal memoranda; and
• communications between executives.

Algorithmic decision-making may leave none of these conventional forms of evidence.

The relevant evidence may instead be contained in:

• model versions;
• feature-weight changes;
• API calls;
• experiment logs;
• recommendation outputs;
• model configuration files;
• audit trails;
• ranking scores; and
• automated decision records.

Mandatory logging therefore creates an algorithmic equivalent of documentary evidence.

5. Relationship With Abuse of Dominance

Mandatory logging is particularly relevant where a dominant platform is alleged to engage in:

Self-preferencing

The platform ranks its own service above competing services.

Exclusionary demotion

Competitors are systematically pushed down in search or recommendation results.

Discriminatory ranking

Comparable businesses receive materially different rankings without objective justification.

Leveraging

A dominant position in one market is used to influence competition in another.

Predatory algorithmic conduct

Ranking rules disadvantage competitors in order to reduce competitive pressure.

Refusal or restriction of visibility

A platform technically permits market participation but makes competitors practically invisible through ranking mechanisms.

The evidentiary question becomes:

Why did the platform's algorithm produce this particular ranking?

Logging enables a competition authority to answer that question.

6. Relationship With Transparency

Logging should be distinguished from public transparency.

A platform may have to maintain detailed logs without publishing them to the public.

A three-level system is therefore possible:

This is important because unrestricted disclosure could expose:

• trade secrets;
• cybersecurity information;
• anti-fraud mechanisms;
• commercially sensitive data; and
• proprietary algorithms.

The regulatory objective is therefore generally auditability rather than complete public disclosure.

7. Six Major Case Laws

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

Google Search (Shopping) is one of the most important precedents for algorithmic ranking and competition law.

The European Commission found that Google systematically gave prominent placement to its own comparison-shopping service while demoting competing comparison-shopping services.

The case demonstrates that ranking is capable of producing exclusionary effects.

Relevance to mandatory logging

If ranking decisions are potentially capable of excluding rivals, regulators need evidence concerning:

• ranking criteria;
• algorithmic changes;
• treatment of Google's own services;
• treatment of competing services;
• traffic effects; and
• internal experimentation.

Mandatory logging would make such evidence systematically available rather than dependent upon retrospective reconstruction.

Principle

Algorithmic visibility can constitute an important competitive parameter.

2. Google Android — European Commission

In Google Android, the European Commission examined Google's contractual and technical practices concerning Android devices and competing search and browser services.

The case demonstrates the importance of examining how technical design and contractual arrangements can influence competitive access.

Relevance

A logging regime could capture whether:

• competing applications were disadvantaged;
• default arrangements affected rankings;
• recommendation systems altered visibility;
• technical configurations changed competitive exposure.

The broader principle is that digital architecture can influence market competition even without an explicit refusal to deal.

3. Google AdSense — European Commission

The Google AdSense proceedings concerned Google's contractual restrictions affecting competing search advertising providers.

Although not principally a ranking case, it is highly relevant to algorithmic intermediation because digital advertising systems determine which commercial opportunities are presented to users.

Relevance

Mandatory logging could preserve:

• auction parameters;
• ranking criteria;
• eligibility rules;
• changes to advertising algorithms;
• differential treatment of advertisers; and
• commercial incentives affecting placement.

The case supports the proposition that intermediary control over visibility can become a competition-law concern.

4. Amazon Marketplace — European Commission

The European Commission's investigations into Amazon's marketplace practices examined the use of non-public marketplace seller data and Amazon's role as both marketplace operator and retailer.

The competition concern arises from the possibility that a platform possessing extensive information about independent sellers can use that information in competition with those sellers.

Relevance to ranking logs

Where marketplace data feeds ranking or recommendation systems, logs could help determine whether:

• third-party seller information influenced ranking;
• Amazon's own products received preferential treatment;
• ranking criteria changed after the platform acquired competitive information;
• seller performance data was selectively used; or
• platform incentives affected visibility.

Principle

A platform's informational advantage can have competitive significance when combined with control over market visibility.

5. Microsoft — Tying / Internet Explorer

The Microsoft Internet Explorer proceedings are an important historical precedent concerning the ability of a powerful platform to influence downstream competition through its control over a technological ecosystem.

The case demonstrates that technical architecture and default positioning can have competitive effects.

Relevance to ranking systems

Modern recommendation systems can perform a function analogous to digital defaults.

Instead of technically forcing a product onto a user, a platform can effectively prioritize it by placing it:

• first;
• prominently;
• automatically;
• persistently; or
• through default recommendations.

Mandatory logs allow authorities to determine whether such prominence resulted from genuine consumer-relevance criteria or strategic exclusion.

6. United States v. Google — Search and Search Advertising

The U.S. Google search litigation provides another important body of precedent concerning the competitive significance of Google's search ecosystem and distribution arrangements.

The litigation demonstrates the importance of examining how control over search and distribution can reinforce market power.

Relevance to logging

A competition authority investigating algorithmic search conduct may need to reconstruct:

1. the ranking algorithm in operation;
2. relevant changes;
3. affected search queries;
4. treatment of competing services;
5. commercial incentives;
6. distribution arrangements; and
7. resulting traffic patterns.

Mandatory logging makes that reconstruction substantially easier.

7. Additional Case Law: FTC v. Qualcomm

FTC v. Qualcomm is not primarily a ranking case, but it is useful for understanding the evidentiary significance of technologically complex competitive conduct.

The litigation involved licensing practices, modem-chip markets and competition between technology suppliers.

Relevance

As markets become technologically complex, courts and regulators increasingly require evidence showing how a technical practice actually affects competitive conditions.

For algorithmic platforms, ranking logs can provide that evidentiary bridge.

8. The Evidentiary Function of Logging

Mandatory logging has three major evidentiary functions.

Function 1 — Attribution

It identifies who or what caused the ranking decision.

Was it:

• an automated model;
• a human;
• a commercial rule;
• an advertiser;
• a contractual arrangement; or
• a combination?

Function 2 — Reconstruction

Investigators can reconstruct historical algorithmic behaviour.

This is essential because algorithms may change thousands of times.

Function 3 — Causation

Logs can help determine whether a suspected practice actually caused:

• traffic diversion;
• competitor demotion;
• consumer foreclosure;
• reduced visibility;
• market-share changes; or
• increased platform self-preferencing.

9. Logging and Self-Preferencing

Self-preferencing presents one of the strongest arguments for mandatory ranking logs.

Consider a marketplace where the platform sells its own product and hosts independent sellers.

Suppose:

Platform product = Rank 1
Rival product = Rank 17

The regulator must determine whether the difference resulted from legitimate factors such as:

• price;
• quality;
• delivery;
• availability;

or from an undisclosed preference for the platform's own product.

A log could reveal:

Model version 7.3 → self-preferencing feature enabled → platform product receives ranking adjustment +20.

That evidence could fundamentally change the competition-law analysis.

10. Logging and Algorithmic Discrimination

A platform may argue that its ranking system is neutral.

However, two competitors might receive systematically different treatment.

Mandatory logs enable regulators to test whether apparently neutral criteria produce discriminatory outcomes.

For example:

The log could reveal whether the ranking discrepancy is explained by legitimate criteria or by an undisclosed platform preference.

11. Logging and Algorithmic Collusion

Logging also has relevance beyond unilateral conduct.

Suppose competing firms use algorithmic pricing systems that repeatedly converge on parallel prices.

The key questions become:

• What information did the algorithms receive?
• How frequently did they update?
• Did they observe competitors' prices?
• Did they react automatically?
• Were common optimization parameters used?
• Did firms intentionally design systems to achieve coordination?

Logging can therefore provide evidence in cases involving:

• algorithmic collusion;
• hub-and-spoke coordination;
• tacit coordination;
• automated price matching; and
• machine-to-machine interaction.

12. Mandatory Logging as a Remedy

A competition authority could impose logging as a behavioural remedy.

For example:

“The undertaking shall retain for five years all records necessary to reconstruct material ranking and recommendation decisions affecting competing products.”

The remedy could require:

• model-version preservation;
• input-category records;
• ranking outputs;
• rule changes;
• human overrides;
• testing records;
• A/B experiments;
• commercial adjustments; and
• audit access.

13. Possible Regulatory Standard

A practical regulatory framework could establish the following requirements:

Tier 1 — Basic logs

Every ranking event records:

• timestamp;
• item;
• position;
• model version.

Tier 2 — Decision logs

Add:

• material ranking factors;
• reason codes;
• rule changes;
• human interventions.

Tier 3 — Competition-sensitive logs

For dominant platforms, additionally preserve:

• self-preferencing adjustments;
• commercial incentives;
• competitor treatment;
• exclusion decisions;
• marketplace-access decisions.

Tier 4 — Regulatory audit

Authorities receive secure access to historical logs and technical documentation.

14. Proportionality Problems

Mandatory logging is not without risks.

Compliance costs

High-volume platforms may process billions of ranking events.

Privacy

Individual-level recommendation logs may contain sensitive personal information.

Cybersecurity

Detailed logs could reveal system architecture.

Trade secrets

Complete disclosure could expose proprietary algorithms.

Storage costs

Long-term preservation can be expensive.

Therefore, the obligation should ordinarily be based on materiality and proportionality.

15. Data Minimisation

A sophisticated system should avoid indiscriminate collection.

Instead of retaining every personal attribute, platforms could record:

“Personalisation factor P7 materially affected ranking.”

rather than retaining the entire underlying personal profile.

This creates an important distinction:

auditability ≠ unlimited surveillance.

16. Interaction With GDPR and Data Protection

Ranking logs may contain personal data where recommendations are individualized.

Accordingly, a regulatory framework must reconcile:

• competition law;
• data protection;
• confidentiality;
• cybersecurity; and
• trade-secret protection.

The competition authority could therefore receive pseudonymised or aggregated logs, while obtaining identifiable information only where necessary.

17. Logging and Ex Ante Digital Regulation

Mandatory logging is particularly suited to ex ante regulation of systemic platforms.

Instead of waiting until harm occurs, a regulator can continuously examine:

Input → algorithmic decision → ranking → market effect

This transforms competition enforcement from a purely retrospective model into a combination of:

monitoring + investigation + intervention.

18. Relationship With the Digital Markets Act

The EU Digital Markets Act is particularly significant because it imposes obligations on designated gatekeepers concerning matters such as self-preferencing, data use and interoperability.

Mandatory logging can operate as the evidentiary infrastructure necessary to test compliance.

The conceptual relationship is:

Substantive obligation → logging requirement → audit → evidence → enforcement

Without reliable logs, a prohibition against discriminatory ranking may be difficult to enforce.

19. UK Competition-Law Significance

In the United Kingdom, mandatory logging could support enforcement under the Competition Act 1998, particularly where algorithmic ranking contributes to:

• abuse of dominance;
• exclusionary conduct;
• discriminatory treatment;
• leveraging;
• self-preferencing;
• refusal of access; or
• anti-competitive platform design.

It can also complement the UK's newer digital-markets framework by giving the regulator a reliable evidentiary record concerning compliance with conduct requirements imposed on firms with substantial and entrenched market power.

20. Key Legal Principle

The central principle can be expressed as follows:

A platform that possesses substantial control over market visibility should not be permitted to make competition-significant ranking decisions without retaining sufficient evidence to reconstruct those decisions.

The obligation is therefore not necessarily:

“Disclose your algorithm.”

It is instead:

“Preserve enough evidence to establish how your algorithm affected competitive visibility.”

21. Advantages

Mandatory logging can:

1. improve regulatory investigations;
2. reduce evidentiary uncertainty;
3. detect self-preferencing;
4. expose discriminatory ranking;
5. identify algorithmic exclusion;
6. preserve evidence after algorithms change;
7. facilitate independent audits;
8. improve accountability;
9. discourage manipulation; and
10. make behavioural remedies enforceable.

22. Limitations

However, logging alone cannot establish an infringement.

A log may show:

“Platform-owned product ranked first.”

It does not automatically prove:

“The platform unlawfully abused dominance.”

Authorities must still establish the relevant legal elements, including where applicable:

• dominance;
• conduct;
• foreclosure;
• competitive harm;
• causation;
• objective justification; and
• proportionality.

Thus, logging is an evidentiary and governance mechanism, not a substitute for substantive competition-law analysis.

23. Overall Assessment

Mandatory logging of ranking and recommendation decisions represents an important development in digital competition enforcement because visibility itself has become an economic resource.

Traditional markets often allowed investigators to examine contracts and prices directly. Digital markets increasingly require investigators to examine the decision architecture that determines who receives attention.

The Google Shopping, Google Android, Google AdSense, Amazon Marketplace, Microsoft and U.S. Google cases demonstrate different dimensions of this problem: search ranking, technical defaults, advertising intermediation, marketplace information, ecosystem control and distribution.

The strongest regulatory model would therefore combine:

mandatory logging + confidentiality protections + independent audit + regulator access + proportional retention + privacy safeguards.

The ultimate objective is not to eliminate algorithmic ranking. It is to ensure that algorithmic control over market visibility remains legally auditable and competitively accountable.