1 Introduction

The effectiveness of competition law litigation has traditionally been evaluated through the prisms of case volume before national courts (Centre for European Policy Studies et al. 2007, 11, 15-16; European Commission 2013, 27; Lande and Davis 2007, 882) and compensation awarded by the court. The prevailing assumption is that a greater number of cases indicates a more effective enforcement mechanism and access to justice (Centre for European Policy Studies et al. 2007, 12). This study challenges that assumption by examining whether the Competition Appeal Tribunal—a specialised UK court with one of the highest volumes of judicial activity in Europe—truly reflects an effective enforcement regime. This article details the methodological approach adopted for the research and the challenges encountered in obtaining answers to the research questions.

While this study does not aim to provide an exhaustive definition of “effective remedy” or “access to justice” within the context of competition damages actions, it adopts a working definition informed by the jurisprudence of the European Court of Justice and the UK Supreme Court. The UK legal systems² recognise the right of access to the court as a fundamental principle of common law.³ This right comprises not only the right to open the court door by commencing proceedings but also the right to have the claim adjudicated by the court.⁴ Additionally, the Woolf report highlighted that effective access to the justice system involves ensuring an equitable distribution of the court’s resources among all litigants (Woolf 1996, 24; Woolf 1995). Furthermore, the European Convention on Human Rights (ECHR) forms an integral part of the broader framework of judicial protection within the UK as a ratifying State. The right to access to justice and the right to compensation are therefore interpreted in light of Articles 6 and 13 ECHR, which collectively establish a robust standard for access to effective judicial remedies.

Accordingly, this research poses the following central question: To what extent does the competition law regime provide meaningful access to effective remedies, particularly for small and medium-sized enterprises (SMEs) and consumers? It further asks: What barriers to justice persist within this framework, and to what are they attributable?

To address these questions, the study examines over 400 cases of the Competition Appeal Tribunal (CAT), with the aim of generating insights that are objective, falsifiable, and reproducible. While there is extensive discourse on the objectives of competition damages actions and various proposals for reform, meaningful recommendations must be grounded in a clear and evidence-based understanding of the existing framework’s strengths and limitations. Absent such clarity, reform efforts risk being constructed on unstable foundations, potentially undermining their legitimacy and effectiveness.

The study undertakes a systematic analysis of judicial decisions of the CAT to assess the effectiveness of the UK’s compensatory legal framework, with particular reference to the right to compensation under Sections 47A and 47B of the Competition Act 1998. This article in particular focuses on the methodological dimensions of the research—its design, process, and limitations—and reflects on the practical challenges encountered in extracting and analysing data from the CAT’s case database. To that end, Section 2 outlines the scope and objectives of the underlying empirical study. Section 3 sets out the background and context of the data sources. Section 4 details the initial methodology for data acquisition and structuring, followed by Section 5, which addresses the specific challenges and computational methods employed to identify SMEs. Section 6 explains the systematic content analysis applied to unstructured judicial records. Section 7 introduces the Cost Estimator Tool (CET), detailing its heuristic calibration, phase-weighting benchmarks, and its application in quantifying the 'recovery gap' facing litigants, alongside a discussion of its inherent limitations. Section 8 presents the provisional outcomes of the study regarding party participation, costs, and case durations. Finally, Section 9 concludes by discussing the broader challenges of this empirical approach and outlining avenues for future research.

2 The Underlying Study

The study underlying this project empirically investigates key aspects of the competition damages actions system in the UK, focusing on access to an effective remedy. Competition law infringements primarily affect vulnerable industry stakeholders such as SMEs, microbusinesses, and consumers (European Commission 2013, para. 6; European Commission 2024). One would consequently expect a significant share of claims to originate from these groups. To test this hypothesis, the study tracked the number of competition damages actions over time and identified emerging trends. In the process, it compiled an extensive dataset, encompassing multiple aspects of competition law litigation, including party and claim types, duration, costs, and numerous other variables.

As a starting point, the analysis examined the extent to which various stakeholder groups (particularly SMEs, microbusinesses, and consumers) have engaged with the competition damages regime, particularly by raising a claim for damages under Section 47A or 47B of the Competition Act 1998. This inquiry was conducted in several stages. In particular, SME participation was estimated through three complementary approaches. The first involved matching claimant names against entries in the UK public registry of SMEs (Companies House 2022) using RStudio for data analysis. The second employed a proxy indicator—the use of fast-track procedures specifically designed to enhance access for smaller entities.⁵ Thirdly, and most effectively, data were extracted from the UK Companies House, categorising claimants and defendants by size based on their filed accounts.⁶ Furthermore, the research categorised claims by party type and procedural form, and then cross-referenced these dimensions to identify patterns—for example, assessing how many SMEs pursued fast-track proceedings as follow-on actions, or comparing the prevalence of follow-on versus stand-alone claims since the CAT began hearing stand-alone cases in 2015. This categorisation also helped find the claims initiated by consumers and private individuals.

This research also investigates the costs in both individual damage claims and collective actions, identifying them as one of the main barriers to access to justice. To this end, all cost judgments issued by the Competition Appeal Tribunal (CAT) in Section 47A and 47B cases were analysed by extracting data directly from the CAT website. As a result, the study developed a system to estimate litigation costs based on the procedural phase and the complexity of each case.⁷

The study also seeks to uncover the relationship between actual and expected monetary outcomes and the costs of litigation. This analysis is crucial for evaluating the cost-benefit ratio of bringing competition damages actions and for assessing the likelihood of securing third-party funding. An additional analysis was conducted to compare the duration of legal proceedings between cases that have related (connected) cases and those that are independent.⁸

Finally, there is a general agreement in both literature and court decisions that proving causation and quantifying damages are the primary hurdles for claimants in damages actions (Lombardi 2020, Lianos and Lombardi 2023).⁹ As this study demonstrates, while causation and quantification are indeed complex, this research examines whether, in practice, they have been significant obstacles in competition damages actions thus far.

3 Background: The Data Source and its Context

This empirical study draws upon a comprehensive database of all damages actions adjudicated by the Competition Appeal Tribunal (CAT) between 2004 and December 2025.¹⁰ The dataset comprises 340 individual claims brought under Section 47A of the Competition Act 1998 and 60 collective proceedings initiated under Section 47B. These cases form the evidentiary foundation for assessing the practical accessibility and effectiveness of the UK’s compensatory legal framework in competition law. Section 47A establishes the right to make claims for damages (and injunctions) caused by competition law infringements (Chapter I and II Competition Act 1998). Such actions have been available since the establishment of the CAT in 2003, following the implementation of the Enterprise Act 2002. In contrast, Section 47B, establishing specific procedural rules for collective proceedings, became available only from 1 October 2015, pursuant to reforms introduced by the Consumer Rights Act 2015. This temporal distinction is analytically significant, as it reflects the evolution of the legal framework and, correspondingly, the composition of the datasets examined in this study.

The database includes both structured and unstructured data extracted from the CAT’s public records.¹¹ Structured elements encompass case identifiers (such as case number, claimant and defendant names, registration and last document dates), procedural markers (e.g. fast-track designation, outcome status), and coded variables for analytical categorisation. These include claimant type, claim type, legal basis, microbusiness status, and representation format (e.g. physical person, litigant in person, pro bono representation). The extraction process required careful parsing of judicial documents to ensure consistency and analytical utility.

The research culminated in the construction of five distinct datasets: one covering all Section 47A claims, another detailing Section 47B collective actions, a third isolating cases that reached final judgment, a fourth compiling cost-related decisions, and a fifth examining the size and locations of litigants in individual and collective litigation. These datasets serve as the basis for the systematic content analysis presented in the following sections. The first two datasets focus on all cases, with the data reflecting the total number of cases analysed. In contrast, the other datasets focus on specific decisions, which in some instances may pertain to the same case.

4 Methodology: Data Acquisition and Initial Structuring

The initial dataset of damages actions was compiled by extracting all cases listed under “Cases / Case Type / 47A” on the Competition Appeal Tribunal (CAT) website, yielding a total of 340 cases. This figure accounts for all cases filed initially separately, even if they were later grouped together under a common umbrella of related cases.¹² Similarly, all collective actions were identified by selecting “Cases / Case Type / 47B.” Structured information—including party names, case numbers, registration dates, and case summaries—was extracted using Microsoft Power Automate. This automation process proved both reliable and efficient, owing to the CAT website’s public accessibility and its well-structured format, which lends itself to parsing and crawling by tools such as Power Automate. Subsequently, structured data concerning claim type, claimant type, and legal basis, among other factors, were extracted and incorporated into the broader analysis of damages actions. Moreover, the analysis also collected information on archived and stayed proceedings, as these then became useful in determining the actual case duration.¹³ Then, the remaining two datasets on final decisions and cost decisions were included. In this case, the CAT’s “judgments” database was used, as it contained all relevant decisions. The dataset of final decisions was compiled manually by parsing the “Judgments” sections of both Section 47A and Section 47B cases. Finally, the cost decisions dataset was assembled by selecting “Judgments / Case Type / 47A” and “Judgments / Case Type / 47B,” respectively.

To ensure consistency across all datasets, a core set of primary variables was collected uniformly. This information details structured data, such as case names or registration dates, which are essential to obtain a consistent classification, also for cross-referencing of cases across the four datasets. Each dataset was then supplemented with additional variables specific to its focus. Coding followed a standardised pattern, with structured data consistently categorised first, followed by the dataset-specific variables. Annex 1 provides a comprehensive overview of all variables used in the first dataset, including all shared variables, while the subsequent Annexes describe only those variables not included in Annex 1 and unstructured variables.

Similarly, the study gathered all relevant data from proceedings initiated under Section 47B (collective actions) of the Competition Act 1998 that were available at the time of research. Its main aim is to analyse the nature of the claims and the characteristics of the parties involved, while also facilitating a cost-benefit assessment of collective redress mechanisms. Particular focus is given to the connection between the type of claim and the procedural and financial aspects of litigation, including the costs incurred and the strategic decisions made by claimants.

To facilitate this analysis, the dataset includes a set of coded variables that capture both legal and procedural attributes of each case. These variables are designed to support comparative and inferential analysis across different types of collective actions, and to assess the practical viability of opt-in and opt-out mechanisms for various claimant groups, including consumers, competitors, and others.¹⁴

The “Decided Cases” dataset comprises all final judgments issued by the CAT in standalone and collective damages actions, totalling ten decisions as of March 2025. This subset of cases was selected to enable a focused analysis of judicial outcomes and their relationship to the nature and structure of the claims brought. The primary objective of this dataset is not only to determine the outcome of each case, but, more importantly, to assess the alignment between the claims advanced and the relief ultimately granted—thereby facilitating a risk-benefit analysis of litigation strategies in competition law enforcement.

To achieve this, the study extracted and coded a substantial volume of information from each judgment, resulting in a rich set of variables designed to capture both procedural and substantive dimensions of the litigation. These variables include claimant characteristics (e.g., direct or indirect purchaser, competitor), legal basis (Section 47A or 47B), procedural format (opt-in or opt-out), and financial dimensions such as claimed and awarded damages, litigation costs, and the presence of third-party funding.¹⁵

The fourth dataset captures the analysis of cost decisions and is based on the twenty cost decisions publicly available on the Competition Appeal Tribunal (CAT) website at the time of the research.¹⁶ This component of the study seeks to infer patterns in cost allocation by examining the relationship between litigation phase, claim type, party characteristics, and financial outcomes. Given the limited number of cases, the project adopted a comprehensive approach to data extraction, aiming to maximise analytical value by coding each decision using a set of targeted variables.

These variables capture both procedural context and financial details, allowing for a nuanced assessment of cost exposure and recovery. Specifically, the dataset includes information on the stage of litigation to which the cost ruling pertains, the amounts claimed and awarded to both claimants and defendants, the total value of the underlying claim, and the application of cost-capping measures.¹⁷

This dataset also lays the foundation for the development of a cost estimator tool, a key component of this project. By systematically capturing and analysing cost-related variables, the dataset enables the construction of a model capable of estimating potential cost exposure to legal fees in competition damages actions. The methodology and application of the cost estimator tool are examined in greater detail in Section 7 below. Finally, the fifth dataset includes information about the size and location of all claimants and defendants in Section 47a and 47B cases, which is detailed in Section 5 below.

5 The Problem of Identifying SMEs

Since one of the primary research questions concerned the participation of SMEs in competition law litigation, the project initially faced a significant obstacle: determining the true nature of the parties involved.

Initially, the project attempted to match party names against the UK public registry of SMEs (Companies House 2022) using RStudio for data analysis. This approach was quickly tempered by the recognition that the registry is voluntary and therefore incomplete. As a practical alternative, we turned to fast-track applications as a proxy for SME involvement in litigation: if an SME has access to a cheaper, more streamlined procedure, it seems reasonable to expect that it would make use of it. Yet, anecdotal evidence suggested that some SMEs joined collective actions or opted for standard procedures, particularly in ‘umbrella proceedings’ such as those against Visa and Mastercard. This made reliable categorisation challenging: the available data and analytical tools did not always permit straightforward classification of parties.

To address this, we shifted to a different method, leveraging the “accounts_type” variable in the UK Companies House API. This variable, which reflects the type of account a company files (a legal requirement), serves as a robust proxy for firm size. The Companies House API classifies companies based on turnover, assets, and employee numbers: micro-entities have turnover below £632,000 or assets below £316,000; small enterprises have turnover below £10.2 million or fewer than 50 employees; medium enterprises have turnover below £36 million or fewer than 250 employees. Companies exceeding these thresholds cannot file simpler accounts and were deemed large enterprises or groups thereof, while foreign companies were manually examined and classified. The project relied on this information to classify all claimants and defendants in the dataset.

Using Python, we matched parties in litigation, both claimant and defendants, by name. This process required an initial ‘cleaning’ of the party names to ensure compatibility with the Companies House database. This is because our records included only legal citations, such as “Mastercard Incorporated and Others”), whereas Companies House lists only the formal corporate name, for example, “Mastercard Incorporated”. If left unadjusted, the search algorithm would interpret terms like “and Others” as part of the legal name, resulting in mismatches. To reduce this risk, we systematically removed generic legal suffixes (including “and others,” “& ors” and “claimants”) prior to querying the API.

Often, litigation involved multiple parties on both the claimant and defendant sides. For example, cases might list claimants such as “(1) BCL Old Co Limited (2) DFL Old Co Limited (3) PFF Old Co Limited” and defendants such as “(1) Aventis SA (2) Rhodia Limited (3) F Hoffman-La Roche AG (4) Roche Products Limited”. Because the search and data extraction process was automated via Python, we configured the script to query the API for each individual claimant and defendant. This required the dataframe to be ‘exploded’, generating a new row for every party so that each entity could be analysed independently.

Initially, the script returned only the company’s size at the moment of the API query. However, for research purposes, we needed the company’s size (for both claimants and defendants) as of the time of the claim. To address this, we implemented a ’time machine’ logic in the code: instead of checking only the current status, the script retrieved the company’s full filing history. It then identified the relevant PDF document filed in the year of interest (such as group or annual accounts) and determined the company size based on the specific document’s description. The script was instructed to collect data on corporate accounts at the time of litigation, not at the time of data extraction. However, several undertakings were classified as dormant at the time of opening of litigation. For that reason, all dormant undertakings were manually checked to determine their nature immediately before filing the first dormant accounts.

In parallel, we extracted postcodes for all parties to determine their geographic locations. The script used the same Excel file containing party names, sending each to the Companies House API. The principal variable of interest was the UK postal code; parties without one were initially classified as foreign entities. Upon further review, however, we found that some defendants had UK registration status but no postcode. To account for this, we added a logic check: if the ”company_type” field included “overseas” or “foreign”, the entity was flagged as “Foreign (Registered in UK)”. We also employed inference techniques to classify foreign entities and determine their nationalities. The script primarily focused on suffixes like Ltd (UK), Inc (US), GmbH (Germany), among others. In this process, manual review and adjustments were also necessary. Overall, around 25 per cent of the results required manual verification or override, as the code encountered errors. Ultimately, the results were visualised using the folium and pgeocode libraries, which converted postcodes into latitude and longitude coordinates and plotted them on an interactive map—blue for claimants, red for defendants.

Regarding the ”sector/industry” variable, the script initially scanned for keywords; for example, a label of the public sector was automatically assigned to “Council”, while “hotel” was categorised under hospitality. Additionally, simple heuristics were used for other cases—for instance, the script was set to label financial services if “Mastercard” or “Visa” appeared in the name, and retail if companies like Sainsbury, Asda, Tesco, Morrisons, Next, or TK Maxx were detected.

We applied the same research process to parties listed in the Section 47B class action file. In addition to previous steps, the script was modified to identify and tag natural persons. Without this adjustment, individuals such as “Mr Walter Merricks” could be misidentified as companies (e.g., “Merricks Consulting Ltd”). To prevent this, the script scanned for honorifics (Mr, Ms, Dr, Prof, Lord, Lady); if found, the party was labelled “Class Representative (Individual)”, and size and ownership checks were marked as not applicable (N/A). A manual verification step followed to ensure accuracy. This step took a substantial amount of time, because, as in the case of the Section 47A file, there were several instances in which the code had misidentified parties or simply was not able to find information in the Companies API. In that case, the information about the class representatives and the companies involved was extracted manually.

Furthermore, many class actions are led by special-purpose companies that file “micro” accounts but do not operate as typical SMEs. To avoid misclassifying these entities, we designated them as “legal vehicles”. However, this was probably the least successful part of the script as many of the Special Purpose Vehicles (SPVs) had to be manually updated. Finally, defendants in class actions are frequently foreign corporations not listed in the UK database. To address this, we used a ”manual_overrides” list to force-label these known entities (such as Google, Apple, and Qualcomm).

6 The Analysis of Unstructured Data

The empirical approach adopted in this study was inherently iterative, shaped by the nature of the data available through the Competition Appeal Tribunal and the evolving demands of the research questions. While certain structured data—such as case numbers, registration dates, and party names—could be reliably extracted using automated tools, the substantive dimensions of the cases required a more nuanced and manual process of interpretation.

At the core of this methodology lies a systematic content analysis of unstructured judicial texts. Variables such as claimant type, claim type, and legal basis are not explicitly tagged in the CAT’s decisions and thus necessitated a close reading of the factual background and legal reasoning in each case. For example, identifying the claimant type involved examining the allegations and market relationships described in the judgment. Claimants were categorised according to their relationship with the defendant:

• Direct purchasers (coded as 1) were those who transacted directly with the infringing undertaking;

• Indirect purchasers (2) were those affected by downstream harm and were further subdivided into indirect loss and secondary harm;

• Competitors (3) were claimants operating in the same market as the defendant and alleging exclusionary conduct;

• A combined category (4) was used in multi-defendant cases where the claimant’s relationship varied across parties.

The classification of claim type followed a similarly rigorous process. Claims were coded as follow-on (1) where they relied on a prior infringement decision, or standalone (2) where the claimant initiated proceedings independently. This distinction was not always straightforward. The presence of phrases such as “followed” or “based on” a decision of the authority often required deeper scrutiny, as they could refer merely to the factual trigger for litigation rather than its legal foundation. In such cases, the claim was classified as standalone if the alleged infringement diverged materially from the authority’s findings.

Legal basis was coded according to the substantive provisions invoked:

• Chapter I CA98 / Article 101 TFEU (1) for anti-competitive agreements;

• Chapter II CA98 / Article 102 TFEU (2) for abuse of dominance;

• Both Chapter I and II / Articles 101 and 102 TFEU (3) where claims engaged multiple forms of conduct.

The Truck Cartel and Merchant Interchange Fee (MIF) litigation, however, contain numerous non-independent cases, which could lead to sampling bias and overrepresentation if not handled separately. Indeed, these cases, while voluminous, presented uniform characteristics across claims and would have disproportionately influenced the statistical model. The exclusion of umbrella cases applies only to selected variables—specifically those relating to claimant type, claim type, legal basis, and case duration—where their inclusion would disproportionately distort averages due to their exceptional size and complexity. The remaining sample of 88 cases was sufficiently diverse and representative to support meaningful analysis. Nevertheless, a separate analysis was conducted by including these two umbrella cases to demonstrate their impact on the model.

As the research progressed, particularly in the domain of costs and damages, additional variables were incorporated to capture the financial dimensions of litigation. This prompted the creation of a dataset on cost decisions and the inclusion of specific variables in the final judgment datasets. In the latter, we decided to include both costs and damages. The amounts claimed and awarded—both in terms of costs and damages—were manually extracted from the judgments, as they are not available in structured formats. This process revealed that cost awards generally pertained to distinct procedural phases (e.g., interim hearings, final judgment), while damages were awarded across different categories of harm (e.g., actual loss, loss of profit, loss of opportunity).

To address this complexity, costs were coded according to the phase of proceedings to which they related, both in the costs and final judgment datasets, allowing for a more granular analysis of financial exposure over time. Damages were similarly categorised by type, where the judgment provided sufficient detail. This iterative refinement of variables enabled the construction of robust datasets capable of supporting both descriptive and inferential analysis. Ultimately, this methodological framework not only facilitated the empirical mapping of competition damages actions before the CAT but also laid the groundwork for the development of the cost estimator tool, discussed in Section 6. That tool draws directly on the variables and coding logic established here, offering a practical application of the research for prospective claimants and policy evaluation.

7 The Cost Estimator Tool

The Cost Estimator Tool (CET) is designed to provide an evidence-based estimate of solicitor time costs across the main phases of competition law proceedings before the CAT. It combines two key elements: the principle of proportionality under Rule 104 of the CAT Rules 2015 and empirical insights drawn from recent cost rulings and prevailing market rates. In doing so, it aims to calculate the average claimed legal fees in different phases of competition litigation and the average compensated legal costs under Rule 104. This approach ensures that the model reflects both normative requirements and practical realities, and should enable fair comparison between actual costs and compensated costs in the event of a win.

To establish the estimated claimed solicitor costs, the estimator relies on a combination of observed CAT rulings, a dedicated dataset on cost judgments, and extrapolated market rates across all four fee-earner grades (A through D). Published CAT cost rulings, such as Ryder,¹⁸ CICC,¹⁹ and Roberts,²⁰ provide explicit empirical data for approved Grade A (senior) and Grade B (mid-level) rates, averaging £854.50 and £678.42, respectively. Because junior fee-earners typically conduct the bulk of document-heavy work in competition litigation, the model applies a weighted average rather than a simple mean. However, judgments generally discuss the rates of Grade A and B fee-earners because those are the rates the losing party usually contests. The rates for junior associates and paralegals (Grades C and D) are frequently agreed upon by the parties behind closed doors or waived through by the judge without being explicitly written into the final judgment text. However, the ratio between Grade A/B and C/D rates can be assumed to be consistent with the UK Master of the Rolls' Guideline Hourly Rates (GHR). To establish the baseline ratio, we used the GHR for “London 1”, which covers heavy commercial and corporate work; and then applied that ratio to the CAT-approved rates.

Based on the 2025 London 1 GHR:

• Grade A Guideline: £546 (2024) / £566 (2025)

• Grade B Guideline: £371 (2024) / £385 (2025)

• Grade C Guideline: £288 (2024) / £299 (2025)

• Grade D Guideline: £198 (2024) / £205 (2025)

If we look at the historical ratios in the GHR, according to Practice Direction 3E²¹ Grade C is consistently priced at roughly 53 per cent of Grade A, and Grade D is priced at roughly 36 per cent of Grade A. Even considering the fact that generally competition law litigation involves Magic Circle or US law firms, the absolute numbers in Roberts and confirmed in our dataset are much higher than the standard GHR. However, we can assume that the internal firm leverage ratio (the difference between Grade A/B and C/D) remains the same. Therefore, we calculated our missing variables (Grades C and D) based on the ratio above:

• Estimated Grade C: £854.50 (Grade A) × 0.53 = £452.88

• Estimated Grade D: £854.50 (Grade A) × 0.36 = £307.62

We then divided the action into different phases following the UK Civil Procedure Rules (CPR) Practice Direction 3, which includes: pre-action, issue/statements of case, Case Management Conferences (CMCs), disclosure, witness evidence, expert reports, PTR, trial, and ADR/settlement.²² The UK courts enforce a strict rule of proportionality based on CPR Part 44.3(5) whereby low complexity work should be delegated to junior lawyers or paralegals. We henceforth applied heuristics to determine the distribution of work for each phase of the proceeding between higher and lower grade lawyers based on the type of work required in the specific phase. Standard commercial litigation is quite different in this sense from competition litigation. The CAT rules 2015 emphasise front-loaded case management and economic evidence,²³ where senior staff (Grade A and B) have a key role in drafting or working together with experts to provide the evidence needed. Considered this, we have allocated the weights for each phase as presented in Table 1.

Based on the distribution in Table 1, the model determines a phase-specific blended hourly rate (R_phase) using the following formula:

\[R_{phase}\ = \ (RA\ \times \ WA)\ + \ (RB\ \times \ WB)\ + \ (RC\ \times \ WC)\ + \ (RD\ \times \ WD)\]

Where RA, RB, RC and RD²⁴ represent the hourly rates for Grade A to Grade D fee-earners, and WA, WB, WC and WD are the weighted contributions of each grade. For the purposes of this estimator, and to ensure the tool remains accessible and practical for a generalist audience, the model relies on the simplifying assumption that all four grades contribute to every procedural phase in the exact proportions outlined in Table 1.

7.1 Phase Weighting

Once the blended rate for each phase is established, the next step is to determine the total volume of work and how it is distributed. The action follows the same phases used in Table 1, which were derived from the UK CPR Practice Direction 3E. Since cost decisions often present overall financial figures instead of detailed hourly phase breakdowns, we initially set benchmarks for phase allocation as a percentage of the total procedure. These benchmarks are derived from authoritative guidance and widely used training materials on complex commercial and competition litigation budgets (Civil Justice Council 2023; Commercial Litigation Funding Association 2025; Competition Appeal Tribunal 2015; European Commission, Directorate-General for Competition 2025; HM Courts & Tribunals Service 2019; Judiciary of England and Wales 2022).

The model uses heuristics to adapt commercial litigation phasing to the procedural realities and formal requirements of competition law. For example, proceedings before the CAT are relatively front-loaded; claimants are precluded from issuing skeletal claim forms and the initial claim form must be fully particularised.²⁵ Furthermore, the Tribunal's active case management mandates that expert econometric methodologies be conceptualised early and for collective actions, claimants must pass the rigorous Collective Proceedings Order (CPO) certification stage. Consequently, the model mathematically shifts a higher proportion of baseline billable hours toward the pre-action, CMC, and expert phases than standard commercial budgets would traditionally dictate. Moreover, because procedural intensity varies significantly depending on the nature of the claim, the model requires the user to self-assess the complexity of their case, categorising it into a designated intensity tier.²⁶

7.2 Estimated Billable Hours

Once the weight of each phase was established, it was necessary to estimate the total hours per case in order to infer the overall costs and the costs attributable to each phase.

Since public CAT judgments rarely disclose explicit hour counts, the total baseline hour ranges assigned to each intensity tier (e.g., 2,000 to 5,500 hours for a medium bilateral claim) were initially established via heuristic triangulation. These baselines were synthesised by applying the 'antitrust premium' identified in the European Commission's TPLF report(European Commission 2021) to historical CPR Precedent H commercial budgets (Civil Justice Council 2023). Crucially, these heuristic ranges are corroborated by the empirical dataset itself through reverse calculation. By dividing the total gross costs claimed in standard CAT bilateral rulings by the Tribunal-approved blended hourly rates, the inferred billable hours consistently align with the model's estimated baseline ranges, validating the tool's foundational inputs. The same intensity tiers in Table 2 were utilised to dictate the baseline range of total overall solicitor hours (H_total) required to litigate the case.

By applying the percentages of Phase Weighting (PW_i) from Table 2 to a representative total hour baseline (H_total), the model mechanically derives the estimated absolute hours (H_i) for each specific phase using the formula:

\[H_{i}\ = \ H_{total}\ \times \ {PW}_{i}\]

7.3 Calculating Final Claimed Costs and Disbursements

With both the estimated hours (H_i) and the phase-specific rates (R_phase) established, the model calculates the corresponding estimated claimed costs for each phase. The total claimed costs (C_claimed) for a full case is the sum of the costs across all 7 phases:

\[C_{claimed}\ = \ \Sigma(i = 1)\hat{}7\ (H_{i}\ \times \ R_{phase})\]

Furthermore, the calculator incorporates an independent variable that can be added to the total cost calculation (D_factor). This includes additional non-solicitor expenses such as expert witness fees, counsel fees (including barristers, King’s Counsel (KC), and junior counsel), as well as other disbursements like electronic disclosure provider costs and court fees. In collective actions, these weights are approximately 2.0, reflecting data from sources such as CICC and Merricks, which demonstrate that expert and counsel fees can significantly increase overall costs. Conversely, for bilateral actions, the D_factor ranges from 0.6 to 1.2, depending on complexity, with fast-track cases set at 0.3.²⁷ These estimates are, of course, based on heuristics and are therefore provided as an addition to the cost estimator.

Therefore, the final estimated total cost of the litigation, including both solicitor fees and external disbursements, is calculated as:

\[Total\ = \ C_{claimed}\ \times (1\ + \ D_{factor})\]

In essence, the estimator offers a structured and transparent way to approximate likely solicitor costs based on observed practice and published rulings. While it does not predict exact figures, it provides a practical benchmark for understanding the financial implications of competition litigation before the CAT.

7.4 Claimed vs. Compensated Costs

Calculating the gross legal spend (Total_Cost) only answers half of the access-to-justice equation. Under Rule 104 of the CAT Rules 2015, the Tribunal will only order the losing party to reimburse costs that are deemed proportionate and reasonably incurred. A core finding of this empirical study is that successful parties rarely recover their gross spend; instead, they are subjected to a strict Rule 104 ’haircut,’ whereby the CAT determines the compensability solely of proportional costs. Based on the dataset of published CAT cost rulings, the average recovery rate (Rec_rate) for successful claimants stands at 56.6 per cent (compared to an even lower 29.0 per cent recovery rate for successful defendants). To reflect the actual financial reality of litigation, the CET applies this empirical claimant multiplier to determine the total compensated costs (C_comp)—the amount a victorious claimant can realistically expect to recover from the losing side:

\[C_{comp} = \ Total\ \times 0.566\]

The distinction between what is billed and what is recovered is vital for prospective claimants. The delta between C_claimed and C_comp represents an unrecoverable ’sunk cost’. Crucially, this procedural sunk cost is compounded by the substantive risk of the litigation itself. Analysis of the substantive damages judgments within the dataset reveals that the total damages actually awarded by the Tribunal represent merely 0.86 per cent of the total damages initially claimed (a figure heavily skewed by multi-billion-pound collective actions resulting in much smaller awards). Therefore, claimants not only face guaranteed losses on their legal fees, but they also face a statistical improbability of recovering the full economic value of their alleged harm.

Table 4 summarises this financial modelling across the complexity tiers and highlights the ’recovery gap’ in private enforcement.

7.5 Limitations and Assumptions in the CET

The CET is designed to provide a transparent and evidence-based benchmark for litigation costs, but its utility depends on an understanding of its underlying assumptions and inherent limitations.

Although the initial CET design was based on a flat-rate model, we recognised that it relied on simplifying assumptions that could lead to overly approximate results. Therefore, we introduced a phase-specific weighted rate (Rphase) to account for fluctuations in seniority composition, which vary depending on the procedural phase and directly influence costs. While this method improves realism, the seniority weights (W) were mainly derived from heuristics. Moreover, hourly rates for Grades A and B are based on recent CAT rulings but are limited in number. Rates for Grades C and D are instead extrapolated from historical GHR ratios, which we assumed to stay relevant in practice, although they may differ depending on firm-specific structures.

Furthermore, we incorporated a ”disbursement factor" (D_factor) to account for non-solicitor costs, including King’s Counsel and economic experts. However, these costs can vary significantly between proceedings, and there was no available data to estimate them accurately. Consequently, we decided to estimate these costs as a multiplier of solicitor fees, as they are generally comparable to the overall case expenses. Nevertheless, these disbursements remain highly volatile; for example, the fees for a specialised econometrician in a standalone "mega-case" may surpass the heuristic multipliers provided. Additionally, the model currently does not include after-the-event insurance premiums or third-party funding success fees, which are critical components of the total financial risk in collective proceedings. For these reasons, the ”D_factor” was left as an optional addition that the CET user can choose to include in the overall calculation.

The user also needs to self-assess the complexity of the case and other elements. In other words, the model has shifted from deterministic mathematical multipliers to a self-assessment framework. This shifts the responsibility to the user to evaluate qualitative factors, such as the nature of the claim, the burden of proof, and the evidentiary asymmetry between parties, to select the appropriate intensity tier. While this approach more accurately reflects judicial discretion in costs assessment, it introduces a level of subjectivity that may cause variation between different users' estimations for the same case.

The CET is calibrated against an original dataset of CAT cost rulings, with the 56.6 per cent recovery rate serving as its core empirical anchor. However, this dataset remains focused on published costs judgments, which represent only a fraction of CAT activity. The model does not yet account for out-of-court settlements, which are confidential and often structured differently than tribunal awards.

Finally, the estimator is a deterministic guide intended for scenario planning and policy analysis; it is not a precise predictor of specific cost outcomes. Future iterations will benefit from broader empirical validation as more collective and bilateral proceedings reach the costs-ruling stage, as well as the potential integration of sensitivity analysis to account for fluctuating market rates for KCs and experts.

8 Provisional Outcomes

This study has revealed a number of unexpected details about competition law litigation and access to effective remedies, but it has also confirmed some of the previous assumptions.

The data analysed show a very low level of use of Section 47A actions by SMEs and microbusinesses, with only over 10 per cent of the claimants classified as SMEs, 14 fast-track applications submitted by self-assessed small or microbusinesses and two registered cases involving SMEs in collective litigation. On the consumer side, the findings indicate a growing number of collective actions. However, a key obstacle in these cases is the limited availability of third-party funding, which is essential given the substantial costs associated with collective proceedings.²⁸

When the study focused more specifically on the participation of SMEs in damages actions before the Competition Appeal Tribunal, the data revealed a significant gap between theory and practice: less than 15 per cent of cases featured SMEs as claimants, and approximately half of those were concentrated within just two ‘umbrella proceedings’. Despite the common understanding of SMEs as the primary victims of competition law infringements—and thus the intended beneficiaries of private enforcement—their direct engagement in litigation remains strikingly low. From this perspective, the analysis of collective actions proved less illuminating. Our methodology focused on the nature of class representatives rather than the specific composition of each class. For instance, while it is safe to assume that numerous SMEs are represented in BIRA Trading Limited v Amazon.com, Inc., a granular analysis of every individual class member would have stretched this empirical study far beyond its intended scope.

Another aspect that is usually considered to help claimants prove their claims is to rely on a competition authority’s decision with a follow-on claim. If one excludes the two ‘umbrella cases’ (but not those with related proceedings, such as Google and Cardiff City Transport Services Limited), the number of follow-on actions stands at 53, compared to 38 stand-alone actions. However, if we consider that the CAT began hearing stand-alone proceedings only from 2015, the adjusted figures are 29 follow-on and 38 stand-alone cases. This suggests that stand-alone actions have, in fact, outpaced follow-on claims. Even more striking is that among the relatively few fast-track applications, the greatest majority are stand-alone actions. This trend indicates that the procedural tools introduced to facilitate follow-on claims have not significantly altered the litigation landscape—particularly for SMEs.

With regard to claimant types, the study reveals a marked underrepresentation of indirect purchasers and parties only indirectly affected by the anticompetitive conduct. Litigation appears to be predominantly pursued by direct purchasers and competitors. This pattern suggests that the current legal framework may not sufficiently incentivise or enable claims from more diffuse categories of harm.

Furthermore, the findings suggest that the cost-benefit balance is generally weak, offering limited justification for claimant involvement or third-party investment if the aim is compensation alone.²⁹ This is largely due to the high litigation costs and the relatively modest compensation awards—particularly when compared to the amounts claimed in both costs and damages. Of course, a comprehensive analysis would require access to settlement data, as the vast majority of cases are resolved privately. However, since settlements are typically concluded behind closed doors, a subsequent study will conduct interviews with solicitors, barristers, and other key stakeholders to validate the data and refine the estimates. Current empirical research, however, seems to confirm this point (Danov 2024). While it is true that settlements—likely the majority of case resolutions—are confidential and therefore excluded from this study, the analysis is explicitly focused on the cost-benefit dynamics of litigation that proceeds to a final decision. This scope is deliberate: the aim is to understand the structural and economic barriers inherent in the litigation pathway itself, rather than to capture the broader settlement landscape, which would require a different methodological approach (including interviews or surveys). Consequently, the findings should be interpreted as reflecting the economics of litigated cases, not the entire enforcement ecosystem. As for decided cases and published cost judgments, the data indicate that in some instances, litigation costs exceed the damages awarded. In many others, costs represent a substantial proportion of the awarded sum.

The median duration of proceedings is nearly identical for both claims having related cases (thus allegedly being more complex from a base management perspective) and independent/discrete claims. This suggests that the typical case length is not necessarily related to the number of related cases. Interestingly, the average (mean) duration is slightly lower for related cases, indicating that some shorter proceedings may be pulling the average down. However, related cases show a high degree of variability, as reflected in their standard deviation, which suggests that while many are resolved efficiently, others may still experience significant delays. These findings highlight that related cases are not inherently longer. Nonetheless, there remains potential to reduce costs and improve efficiency by better managing the coordination and handling of connected cases.

Additionally, this research provides evidence that the victorious party is often not fully compensated for expenses, as Rule 104 grants the CAT the power to order the reimbursement of proportionate and reasonable expenses, while also allowing for the reduction of the awarded amount if certain circumstances warrant it.

Finally, causation, while a fundamental element in competition law litigation, particularly in claims brought by indirect purchasers, umbrella buyers, competitors, counterfactual buyers, or other parties only indirectly linked to the defendant, is statistically a less prominent feature in practice.³⁰ This is because only a minority of cases progress to the stage where a full evaluation of causation is required.

9 Challenges and Future Research

As the project aimed to identify the barriers to effective competition law remedies for consumers and SMEs, we entered uncharted territory regarding the limitations of data that can be gleaned from court decisions. This study has crystallised and organised a valuable wealth of information concerning the CAT’s jurisprudence and has also provided evidence for some of the preliminary conclusions. While competition damages actions offer a theoretical route for redress, the empirical data from the CAT reveal significant procedural and cost-related hurdles that serve as de facto barriers for consumers and SMEs. Moreover, several other aspects of the litigation have been examined or subjected to more detailed scrutiny, such as case duration and the impact of follow-on litigation, to mention just two.

However, there are certain challenges related to data limitations that could not be addressed in this study alone, positioning this research more as a foundational work than a final study. The analysis indeed relies solely on publicly available CAT records. While this ensures transparency and replicability, it excludes confidential settlements and High Court judgments. Future research should consider complementary approaches, such as interviews with practitioners, judges, and third-party funders, freedom of information requests, or triangulation with data from the Competition and Markets Authority (CMA) to better capture the wider enforcement landscape.

The CAT decisions cover different stages of the proceedings, which are often difficult to compare in terms of complexity, duration, and costs. The drafting style and case management rules, as well as some procedural norms governing litigation before this Tribunal, have also evolved over the years. These are only some of the issues encountered in a longitudinal study that views CAT’s judgments as separate decisions. Additionally, there are undeniable influences among cases, either because they are formally connected or to some extent related. To some degree, this project has attempted to account for this complexity by highlighting the two main umbrella cases and considering a variable for ‘connected cases’. However, these variables contribute differently to the overall case complexity, which this project does not aim to define precisely. Further research should also aim to establish a benchmark for measuring case complexity.

Regarding the systematic analysis of the current case law available at CAT level, there was a significant limitation that needed to be addressed. Only a small proportion of cases reach a final judgment, limiting the ability to evaluate key variables such as damages awarded, litigation costs, and case duration. While this reflects an important procedural reality—most cases are withdrawn or settled—it also highlights the need for qualitative research. Surveys or interviews with claimants, funders, and legal representatives could offer valuable insights into settlement dynamics and decision-making processes.³¹

Moreover, a strategic decision had to be made in analysing certain specific variables, such as case duration, claim type, and party type. The exclusion of large umbrella proceedings, such as the Truck Cartel and MIF litigation, from some analyses was necessary to prevent skewing the averages. These cases have been grouped under a common umbrella because each comprises dozens of individual cases, all stemming from two key decisions of the European Commission: the Truck Cartel case³² and the MIF litigation.³³ Although they cannot be technically considered as a single case, for the purpose of analysing variables like case type, it is useful to consider scenarios both including and excluding these umbrella cases. This approach ensures that their impact on systemic trends is thoroughly understood.

From a more conceptual point of view, the present study adopts a narrow definition of “access to justice,” focusing on procedural and cost-related barriers. While this approach aligns with the study’s objectives, it overlooks qualitative dimensions such as behavioural aspects (e.g., fear factor and perceived fairness), informational effects, and deterrence effects. Future research should employ mixed-method designs, combining interviews, surveys, and behavioural analysis to explore these aspects.

The cost estimator developed in this study provides a practical benchmark but is based on simplifying assumptions, including static solicitor rates, a fixed staffing mix, and heuristic multipliers for complexity. It also excludes disbursements such as expert and counsel's fees and court charges. Future iterations should incorporate dynamic rate adjustments, sensitivity analysis, and all other costs and fees that are typically part of the litigation.

Finally, the analysis of causation and quantification issues is currently based on manual review of a limited number of judgments. A more systematic approach using natural language processing could identify patterns across the full corpus of CAT decisions, enabling a richer understanding of how these issues influence litigation outcomes.

In conclusion, this study marks an important step towards a systematic approach to analysing empirical evidence of the effectiveness of competition damages actions. However, several challenges remain, which also suggest promising avenues for further research.

Acknowledgments and Disclosures

The author would like to express sincere gratitude to Prof. Barry Rodger for his insightful comments on an earlier version of this paper. He also wishes to thank all participants at the ESELS 2025 Conference in Toulouse for the valuable discussions and feedback. Any errors that remain in the text are solely the author's own. The author declares no conflict of interest.

References

ARC Costs. n.d. ”Precedent H: How to Prepare a Costs Budget.” Accessed September 26, 2025. https://www.arccosts.co.uk/our-blog/precedent-h-guidance-how-to-prepare-a-costs-budget/.

BarnsChapel Costs Law. 2025. ”The King’s Bench Guide (April 2025) & Phase Guidance Summary.” https://barnschapel.co.uk/the-kings-bench-guide-april-2025/.

Carter Burnett. n.d. ”Costs Budget Guidance Note (KBD Phasing & Expectations).” Accessed September 26, . https://www.carterburnett.co.uk/costs-updates/costs-budget-guidance-note-issued-by-kbd/.

Centre for European Policy Studies, Erasmus University Rotterdam, and Luiss Guido Carli. 2007. Making Antitrust Damages Actions More Effective in the EU: Welfare Impact and Potential Scenarios. http://ec.europa.eu/competition/antitrust/actionsdamages/files_white_paper/impact_study.pdf.

Civil Justice Council. 2023. Costs Report. Civil Justice Council. https://www.judiciary.uk/wp-content/uploads/2023/05/Civil-Justice-Council-Costs-Report-FINAL-May-2023.pdf.

Companies House. 2022. ”A List of Small and Medium-Sized Enterprises (SME) in the UK.” FOI release, June 27, 2022. Accessed September 4, 2025. https://www.gov.uk/government/publications/companies-house-foi-responses-january-to-june-2022/a-list-of-small-and-medium-sized-enterprises-sme-in-the-uk.

Danov, M. 2024. ”Private Antitrust Enforcement–Improving Injured Parties’ Access to Legal Remedies.” Global Competition Litigation Review. https://oredev.exeter.ac.uk/repository/handle/10871/136238.

European Commission. 2013. Commission Staff Working Document, Impact Assessment Report: Damages Actions for Breach of the EU Antitrust Rules, SWD(2013) 203 Final, Accompanying COM(2013) 404 Final and SWD(2013) 204 Final.

European Commission, Directorate-General for Justice and Consumers. 2021. Mapping of Third Party Litigation Funding in the EU: Final Report. Publications Office of the European Union.

Jackson, R. M. 2010. Review of Civil Litigation Costs: Final Report. TSO.

Judiciary of England and Wales. 2022. Commercial Court Guide. 11th ed. https://www.judiciary.uk/wp-content/uploads/2022/06/Commercial-Court-Guide-11th-edition.pdf.

Justice UK. 2022. ”Practice Direction 3E – Costs Management.” Accessed March 6, 2026. https://www.justice.gov.uk/courts/procedure-rules/civil/rules/part03/practice-direction-3e-costs-management.

Lande, Robert H., and Joshua P. Davis. 2007. ”Benefits from Private Antitrust Enforcement: An Analysis of Forty Cases.” University of San Francisco Law Review 42: 879.

Lianos, Ioannis, and Claudio Lombardi. 2023. "Causation." In Research Handbook on Private Enforcement of Competition Law in the EU, edited by Barry Rodger, Miguel Sousa Ferro, and Francisco Marcos. Edward Elgar Publishing.

Lombardi, Claudio. 2020. Causation in Competition Damages Actions. Cambridge University Press.