1. Introduction

Autonomous benchmarking systems are software or artificial-intelligence systems that independently collect, compare, analyse, and evaluate market information such as prices, costs, discounts, output, quality, inventory, wages, procurement terms, or other commercial indicators.

Unlike traditional benchmarking, where employees periodically compare competitors' performance, autonomous systems can operate continuously:

Data collection → automated comparison → benchmarking → prediction → recommendation → commercial decision → new data → further benchmarking

These systems can generate substantial efficiencies. Businesses can identify inefficient costs, improve supply chains, detect market trends, optimise procurement, and respond more quickly to consumer demand.

However, from a competition-law perspective, autonomous benchmarking can create risks where the system:

• facilitates coordination between competitors;
• uses competitively sensitive information;
• makes future prices more predictable;
• reduces uncertainty about competitors' strategies;
• facilitates price alignment;
• enables information exchange;
• excludes competing suppliers;
• creates discriminatory access;
• reinforces a dominant firm's market position; or
• becomes an intermediary through which competitors coordinate without communicating directly.

The central legal issue is therefore not whether benchmarking is automated, but whether the autonomous system changes the competitive conditions in a manner prohibited by competition law.

2. Meaning of Autonomous Benchmarking Systems

An autonomous benchmarking system may perform several functions.

A. Price benchmarking

The system automatically compares:

• competitors' prices;
• historical prices;
• promotional discounts;
• wholesale prices;
• freight charges; and
• customer-specific pricing.

B. Cost benchmarking

The system compares:

• input costs;
• labour costs;
• energy expenditure;
• logistics costs;
• production efficiency; and
• margins.

C. Performance benchmarking

The system can compare:

• sales;
• market shares;
• conversion rates;
• delivery times;
• customer retention;
• product quality; and
• inventory turnover.

D. Competitive benchmarking

The system may go further by predicting:

• likely competitor price changes;
• likely capacity expansion;
• likely promotional campaigns;
• expected output;
• market-entry strategies; and
• reactions to a firm's own commercial decisions.

This final category presents the greatest competition-law concern because it may reduce strategic uncertainty between competitors.

3. Why Autonomous Benchmarking Is Different

Traditional benchmarking may involve an annual industry report.

An autonomous system may instead perform benchmarking:

• every minute;
• every transaction;
• across thousands of competitors;
• using real-time information; and
• with automatic commercial recommendations.

For example:

Competitor A reduces price by 5% → system detects change → predicts competitor B's response → recommends price adjustment → company automatically changes price.

If many competing firms use interconnected systems, the market can become highly transparent.

That may have two very different consequences.

Procompetitive effect

Greater transparency may:

• reduce search costs;
• increase price competition;
• identify inefficient suppliers;
• improve consumer information; and
• encourage firms to reduce costs.

Anticompetitive effect

Excessive transparency may:

• facilitate coordination;
• make deviations from an understanding easier to detect;
• reduce incentives to compete aggressively;
• facilitate punishment of discounting;
• enable price alignment; and
• increase the stability of supracompetitive prices.

4. Autonomous Benchmarking and Information Exchange

Information exchange is one of the most important competition-law issues.

Competitors may legitimately use publicly available information.

The problem becomes more serious where benchmarking systems exchange or aggregate strategically sensitive information, such as:

• future prices;
• planned discounts;
• production volumes;
• capacity;
• costs;
• customer allocation;
• future investment;
• inventory;
• business strategies; or
• individualised sales data.

The distinction can be expressed as:

Public historical information

Generally less problematic.

Recent individualised information

Potentially more problematic.

Confidential future strategic information

Potentially highly problematic.

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

5. Benchmarking as a Hub-and-Spoke Mechanism

An autonomous benchmarking provider can function as a hub connecting competing firms.

For example:

Firm A →
Benchmarking Platform ← Firm B
↑
Firm C

The platform receives information from competing firms and produces benchmark recommendations.

The legal issue is whether the system merely provides legitimate aggregated market information or facilitates coordinated behaviour.

A particularly important question is whether participating firms know that the system receives competitors' information and uses it to influence their own commercial decisions.

The fact that the system is operated by software does not by itself eliminate the possibility of competition-law liability.

6. Reduction of Strategic Uncertainty

Competition depends partly upon uncertainty.

A competitor normally does not know with certainty:

• what another firm will charge tomorrow;
• how much it will produce;
• whether it will discount;
• whether it will enter a market;
• how much capacity it will add; or
• how aggressively it will compete.

Autonomous benchmarking can reduce that uncertainty.

This produces a potential transparency paradox:

Greater market transparency can improve consumer comparison while simultaneously making coordination easier.

Competition authorities therefore need to examine whether transparency promotes independent competition or facilitates coordination.

7. Benchmarking and Tacit Coordination

The most difficult situation arises where there is no explicit agreement.

Suppose three competitors use autonomous systems.

Each system:

1. monitors competitors;
2. detects price movements;
3. predicts reactions;
4. adjusts prices;
5. learns from market responses.

Eventually all three systems converge on similar prices.

There may be no email saying:

"We agree to maintain prices."

This raises the distinction between:

Lawful parallel adaptation

Each undertaking independently responds to market conditions.

and

Unlawful coordination

The systems operate as part of a mechanism that facilitates a common understanding or concerted practice.

Competition law generally does not prohibit merely intelligent or rational adaptation to competitors' publicly observable behaviour.

8. Autonomous Benchmarking and Price Fixing

Benchmarking systems can facilitate price fixing in several ways.

Direct price recommendation

The system recommends a particular price based on competitor information.

Price corridor

The system recommends maintaining prices within a narrow range.

Deviation detection

The system identifies competitors that discount below the benchmark.

Automatic retaliation

The system recommends or implements a response to competitors that reduce prices.

Common pricing infrastructure

Several competitors rely on the same pricing or benchmarking infrastructure.

These features may significantly increase the risk that benchmarking becomes a mechanism for coordination rather than an independent analytical tool.

9. Autonomous Benchmarking and Dominant Firms

Section 4-type abuse-of-dominance concerns may arise where the benchmarking system is controlled by a dominant undertaking.

For example, a dominant platform could benchmark sellers on its marketplace and then use the information to:

• determine rankings;
• alter commissions;
• identify rival products;
• disadvantage competing sellers;
• favour its own products;
• condition access;
• impose discriminatory terms; or
• use sellers' commercially sensitive data.

The system may therefore become both:

benchmarking infrastructure + competitive intelligence infrastructure.

This creates potential leveraging and exclusion concerns.

10. Benchmarking and Self-Preferencing

A dominant digital platform might permit independent merchants to sell through its marketplace while simultaneously operating its own competing retail business.

The platform's autonomous benchmarking system could observe:

• merchant prices;
• sales;
• conversion rates;
• inventory;
• product demand;
• customer preferences.

It could then use that information to improve the platform's own competing products.

This creates potential concerns concerning:

• self-preferencing;
• exploitation of commercially sensitive information;
• leveraging;
• discriminatory access; and
• foreclosure.

11. Benchmarking and Labour Markets

Autonomous benchmarking is not limited to product markets.

Employers increasingly use data systems to benchmark:

• salaries;
• bonuses;
• benefits;
• working hours;
• recruitment costs; and
• labour turnover.

If competing employers receive highly specific information about each other's current or future compensation policies, competition concerns may arise.

The relevant theory is employer-side coordination.

Instead of competitors coordinating the price paid to consumers, employers may coordinate or converge concerning the price paid for labour.

This can potentially involve:

• wage fixing;
• no-poach arrangements;
• restrictions on hiring;
• coordinated benefit reductions; or
• suppression of competition for employees.

12. Benchmarking and Procurement

Autonomous benchmarking can also affect procurement.

A procurement system may continuously compare:

• supplier bids;
• historical tender prices;
• production costs;
• delivery terms;
• capacity; and
• winning bids.

This can improve procurement efficiency.

But if competing bidders obtain information allowing them to predict:

• competitors' bids;
• reserve prices;
• likely winning bids; or
• procurement strategies,

the system could facilitate bid coordination.

The problem becomes particularly serious where the benchmarking provider serves multiple competing bidders and provides individualised information to each.

13. Autonomous Benchmarking and Bid Rigging

Bid-rigging arrangements traditionally involve:

• bid rotation;
• cover bids;
• market allocation;
• customer allocation;
• suppression of bids.

Autonomous systems could potentially automate aspects of such behaviour.

For example:

System identifies competitor's likely tender price → recommends non-aggressive bid → competitor wins → system updates benchmark.

The use of software would not fundamentally change the underlying antitrust concern.

14. Six Major Case Laws

1. United States v. Topkins

Court: U.S. District Court for the Northern District of California
Law: Sherman Act §1

Topkins involved online sellers accused of using pricing algorithms to implement an agreement concerning prices for posters and related products.

Importance for autonomous benchmarking

The case demonstrates that software can be the mechanism through which competitors implement coordinated pricing.

The relevant lesson is:

An algorithm does not provide immunity from traditional cartel rules.

Where an autonomous benchmarking system is used pursuant to an agreement to coordinate prices, competition authorities can analyse the underlying arrangement rather than treating the software as legally neutral.