94 Per Cent, 0.6 Per Cent: The Two Numbers Driving the IPRED Reform Debate

Permanent injunctions granted in 94 per cent of the patent decisions examined. Proportionality expressly assessed in 0.6 per cent. Since January these two figures, taken from the executive summary of the Commission’s follow-up study on the Enforcement Directive, have become the shorthand of the European debate on patent remedies, and within weeks an industry coalition was citing them as proof that European courts grant injunctions automatically. A percentage of injunctions granted is not yet a finding about proportionality. Much of the case for amending the Directive depends on treating it as one, and the reply that all is well rests on no better evidence.

Both come from the Follow-up Study on the Application of the Directive on the Enforcement of Intellectual Property Rights, published by the European Commission at the end of January 2026.1 DG GROW tendered it in 2023, and a consortium of EY and Technopolis carried it out.2 Its object is narrower than its title suggests, since it does not evaluate Directive 2004/48/EC (IPRED) as a whole. Building on the Commission’s 2017 guidance on the Directive, it examines how selected provisions are applied in the Member States, across five priority areas: proportionality in patent injunction cases, patent assertion entities, dynamic blocking injunctions, the right of information and its interaction with data protection, and the cost of storing and destroying infringing goods.3 It remains a consultants’ report rather than a Commission proposal, and its legal notice attributes the views expressed to the authors.

The consultation was meant to gather evidence on implementation and to assess whether clarification or change was needed. Its purpose was to improve the effectiveness and consistency of enforcement, and reducing the number of injunctions was never a stated objective. Of the responses received, 114 were usable and 93 addressed proportionality, and the report itself acknowledges how limited their representativeness is.4

The figures travelled fast. In March 2026 a group of industry associations wrote to Executive Vice-President Séjourné that European courts consistently issue automatic permanent injunctions, and asked for IPRED to be amended. Their footnote draws on the same Darts-ip data but gives a different count: two assessments in 628 decisions from EU countries, described as cases in which infringement was found and an injunction requested, and read as an absence of any proportionality assessment in more than 99 per cent of them.5 Neither document supplies the list that would reconcile the two counts. Before that conclusion is drawn, one needs to know which decisions were counted and what the injunctions actually required.

For standard-essential patents the stakes are higher. Compliance with a FRAND undertaking and the consequences of the proposed remedy are separate inquiries, each with its own problem to solve. A reform that blurs them will correct neither, and no percentage of injunctions granted can tell the two apart.

The FRAND question comes first

Of the 635 decisions behind the headline figures, 37 refused an injunction, and the study’s footnote on those refusals mentions technical standards among the reasons.6 Constraints on relief can therefore escape a count that looks only for an express proportionality label. The footnote does not show that every such refusal applied Huawei v. ZTE, or that all FRAND decisions were excluded from the four assessments. That would require the underlying cases and coding rules.

Huawei v. ZTE addresses the circumstances in which a dominant SEP holder’s pursuit of an injunction becomes abusive, after a commitment to license on FRAND terms.7 I have argued elsewhere that the undertaking should also be understood through its contractual content.8 On that approach the inquiry begins with the access promised to implementers and with each party’s performance of its obligations. General balancing cannot decide, in place of that inquiry, whether a licensing offer or a response to it complied with FRAND. The contractual reading cuts both ways. It binds the holder to the access it promised as firmly as it binds the implementer to negotiate in good faith, and an injunction pursued against a willing licensee on terms the holder cannot justify is a breach of the undertaking before it is a question of proportionality.

The judgment of the German Federal Court of Justice of 27 January 2026 in VoiceAge EVS v. HMD keeps the two inquiries apart. Willingness is treated as continuing conduct. Security is addressed following the rejection of a counter-offer, and whether its amount must follow the holder’s last offer is left open. The hardship clause of section 139(1), sentence 3 of the Patent Act is then applied separately, and on the facts the implementer’s delaying conduct and an available workaround defeated the argument.9 The defence was rejected after it had been examined, which is precisely what a count of express assessments cannot see.

A recent post in these columns on the Munich guidelines concerned the sequence in which financial commitments and the examination of the licensing offer are required.10 If that sequence produces an error, the remedy lies in the FRAND analysis. A general invitation to weigh the parties’ economic interests cannot say which licensing obligation was breached.

Proportionality keeps a role where the circumstances establish additional hardship. Matthias Leistner’s July 2026 briefing for the JURI Committee argues for subsidiarity: the general principle should address circumstances beyond those characteristic of SEP enforcement, and should not compensate for shortcomings in the application of Huawei v. ZTE. He also argues that the protection of third-party and public interests should not depend on the infringer’s prior willingness.11

That distinction is sound. Whether an independent patient or customer needs protection does not depend on the weakness of an implementer’s negotiation record. A party seeking relief on that basis must identify the third party’s harm and explain the adjustment required, and ordinary loss of sales should not be dressed up as third-party necessity.

Numerators and denominators

The headline figures come from an analysis of 635 patent decisions handed down between 2015 and 2020, described as covering the EU27, Norway and the United Kingdom. No country breakdown shows how those decisions are distributed. Injunctions were granted in 598 of them, and proportionality was assessed in four. The study discloses that IP2Innovate commissioned the Darts-ip analysis and shared it with the consultants.12 On the day of publication, IP2Innovate welcomed the study and called for legislative amendments to IPRED.13

Commissioned evidence is common on every side of this debate, and it is not unreliable for that reason. Whoever pays for it, its accuracy still has to be shown, and that requires the case list, the selection rules and the coding decisions. The report supplies neither the case list nor a complete coding protocol. The distribution by country and sector remains unknown, and so does the classification of FRAND objections and of partially granted or deferred injunctions.

The 94 per cent rate is therefore neither an overall litigation success rate nor, by its size alone, evidence of excessive relief. If the population consists of cases in which infringement has been established and an injunction requested, a high grant rate is what one would expect. The association letter describes such a population, but the report does not disclose enough to verify that description for all 635 decisions. The relevant question is whether relief was justified in those cases.

A low grant rate would prove no more. Where an action fails because the patent is invalid or not infringed, the court never has to decide whether an injunction should be limited, so adding those cases to the denominator would lower the rate without saying anything about remedial control. A withdrawn action is just as ambiguous, since it may reflect a weak claim or an agreed licence.

Which population is relevant depends on the question asked. Access to injunctions across litigation calls for a cohort of filed actions and their outcomes. Proportionality calls for decisions in which infringement was established, injunctive relief was sought, and circumstances potentially justifying an adjustment were before the court. Swap one population for the other and the subject changes.

The dates limit the inference as well. A corpus ending in 2020 can say nothing about the express hardship provision added to section 139(1) of the German Patent Act, in force since 18 August 2021, nor about the UPC, which opened in June 2023.14 The study was researched in 2024 and its manuscript completed in March 2025. Neither development proves that practice has improved, but both call for current evidence before an older sample becomes the basis for a diagnosis of the present system.

Express reasoning and effective protection

The 0.6 per cent figure concerns recorded assessment and nothing more. It does not show how often defendants pleaded disproportionality, which facts they established, or whether an order was adjusted without being coded under that heading.

The study’s own Table 4 points the other way, recording the application of proportionality in almost all the selected patent, copyright and trademark cases, and ten of eleven responding judges reported applying it always or very often.15 The selection explains much of that result, since Annex I takes in only decisions in which proportionality was raised or applied, or an injunction adjusted. That sample is selected and mixed, partly concerned with provisional relief, and cannot replace the Darts-ip figures with a more reassuring percentage.

Infrequent pleading would prove nothing either. Litigants may see no relevant hardship, or lack the evidence to establish it, and each explanation calls for a different response. Silence in published reasons cannot tell them apart. It cannot be read as a defect, but it cannot be read as control either. A party that has pleaded a defined adjustment and supported it with evidence is entitled to reasons, and a judgment that disposes of the point in a sentence leaves the appellate court, and any evaluator, with nothing to review.

Article 3(2) of IPRED requires enforcement measures to be effective, proportionate and dissuasive. An evaluation must therefore look for failures of protection in both directions, and fewer injunctions cannot serve as a measure of compliance.16 The Directive also contains a more specific tool, which the headline figures ignore. Under Article 12, Member States may allow courts to order pecuniary compensation instead of an injunction, at the request of an infringer who acted unintentionally and without negligence, where execution would cause disproportionate harm and compensation appears reasonably satisfactory.17 Any reform proposal has to say whether it means to widen that exception or to build a different one. A count of express assessments does not answer that question.

When measurement becomes prescription

Arithmetic can be exact while the legal categories being counted remain contestable. Deciding what counts as an assessment, treating silence as the absence of control, recording a tailored injunction as a simple grant: each of these is an interpretive choice. The danger of turning law into a technical exercise is that such choices vanish behind the apparent neutrality of a percentage, and a lower injunction rate quietly becomes a policy target, whether or not the individual orders were justified. A high grant rate defended as proof of effective enforcement would commit the same error in reverse.

Nor does the authors’ standing settle anything. The report was prepared by EY Consulting and Technopolis, with EY intellectual property lawyers and judicial and academic contributors.18 The question is whether the analysis demonstrates the legal relevance of its categories and the steps that connect its evidence to its recommendations.

Some passages fall short of that standard. The report presents the Darts-ip results as empirical support for criticism of quasi-automatic injunctions, while its own footnote explains that the vocabulary comes from interviewees and that no legal system has truly automatic injunctions.19

The conclusions on patent assertion entities go further, connecting certain financing structures with possible money laundering or the funding of criminal organisations, and the passage carries no citation. An allegation of that gravity needs evidence before it can shape a recommendation on patent remedies. Opaque litigation funding is a legitimate subject of inquiry in its own right, which is all the more reason not to raise it without support.20

The terms of the order

The UPC Court of Appeal’s decision in Meril v. Edwards shows what a binary count misses. On 25 November 2025 the Court explained that a permanent injunction ordinarily follows an established infringement, subject to special reasons, disproportionality among them. The word “may” in Article 63 does not create an unrestricted choice whether to enforce a valid and infringed patent.21 Nor is the balancing required for provisional measures to be treated as the complete test for final relief.22

The Court nonetheless preserved access to particular XL prosthetic heart valves supplied by Meril, excluding those devices from the injunction and from the corrective measures where a physician confirmed that they were the only available treatment for an individual patient. The exception in the operative part reached beyond patients whose implantation had already been scheduled at the date of the first-instance judgment.23

In a database, that is simply an injunction granted. For the patient concerned, what the order still permits matters far more than the binary outcome. The same blind spot recurs whenever relief is confined to particular products or uses, or its operation postponed. A high grant rate may coexist with substantial adjustment, and a formally narrow order may impose severe disruption. Evaluating a remedy means recording what the order actually does and whom it protects.

The objection that survives the statistical critique

The study does raise a serious concern, even if its headline evidence cannot measure its breadth: a defence formally available may have little practical effect. It reports eleven published German SEP cases in which the defence failed after the reform, without a reproducible list of those decisions.24 Why the defence failed, and on what evidence, deserves investigation. If the hardship clause almost never succeeds, the legislature is entitled to ask whether the threshold sits where it intended in 2021, and that question does not depend on the Darts-ip figures.

Those failures do not show that an injunction was legally inappropriate in each case. Nor does the study’s methodological weakness make further guidance unnecessary. Replacing an inference of systemic failure with one of systemic health would simply repeat the error.

Beyond its headline figures the study offers useful material. Its account of the factors national courts consider, among them third-party interests and the effects of relief, can inform a concrete request for adjustment, and its support for exchanges between judges points to a practical route towards consistency.25

Guidance could usefully clarify the evidence needed to establish third-party harm and the forms of adjustment available. A problem with access to comparable licences or with the assessment of willingness, however, needs a response aimed at that problem. Policy should follow an identified defect in legal reasoning or remedial design.

Measure the decision the court had to make

A better evaluation would distinguish final from provisional proceedings, disclose its selection criteria, and record whether infringement was established and an injunction requested.26 It would then identify the legal basis of each objection, the court’s response and the terms of the operative order, and for SEP decisions give a separate account of the FRAND issue that determined access to relief.

Particular attention belongs to cases in which a party requested a defined adjustment and supplied evidence for it. Those cases reveal whether a refusal reflects a demanding but coherent rule or an approach that leaves a relevant interest unprotected. The study’s own recommendation to observe the UPC’s developing practice before any amendment offers a starting point.27

For practitioners the lesson is immediate, and it applies on both sides. Defendants should bring a proposed order alongside the defence, identify the products, uses or period for which an exception is needed, and prove the availability and timing of alternatives. Claimants have every reason to engage with that proposal rather than resist it wholesale, since a tailored order is easier to obtain and harder to overturn. In SEP litigation, keep the record of negotiations, offers and security separate. Hardship evidence cannot repair that record, and the record cannot establish every consequence of an injunction.

The real question for reform is which legally justified adjustments courts fail to make, and whose interests are left unprotected as a result. That question deserves an answer, and implementers who raise it are not asking for weaker patents. Answering it requires reading the decisions, their evidence and their effects. The percentage of injunctions granted can only come afterwards.

• 1European Commission, DG GROW, Follow-up Study on the Application of the Directive on the Enforcement of Intellectual Property Rights, Final Report, EY-Technopolis Consortium, Publications Office of the European Union, 2026, DOI 10.2873/9680583, p. 12; Commission announcement, 27 January 2026, Publications Office.
• 2European Commission, DG GROW, contract notice, Follow-up Study on the Application of the Directive on the Enforcement of Intellectual Property Rights (IPRED) and Contribution to the EU Toolbox against Counterfeiting, OJ S 99/2023, 24 May 2023 (TED 309612-2023), TED; Follow-up Study, pp. 4 and 10 (legal notice attributing the views to the authors; research conducted in 2024, manuscript completed in March 2025).
• 3Communication from the Commission, Guidance on certain aspects of Directive 2004/48/EC of the European Parliament and of the Council on the enforcement of intellectual property rights, COM(2017) 708 final, 29 November 2017, EUR-Lex; European Commission, European IP Helpdesk, “New Commission Study Highlights Diverging Application of IPRED Across the EU”, 9 February 2026, IP Helpdesk.
• 4Follow-up Study, pp. 17-18, 23-26; Annex II, PDF p. 140, on the purpose of the evidence gathering. The survey ran from 3 April to 3 May 2024.
• 5Multi-association letter to Executive Vice-President S. Séjourné on the Intellectual Property Rights Enforcement Directive, 3 March 2026, p. 1 and nn. 1-2, letter. Footnote 2 of the letter refers to Darts-ip data for 2015 to 2020 and reports proportionality assessed in 2 of 628 cases (0.3 per cent) in EU countries; compare the study’s four assessments in 635 decisions covering the EU27, Norway and the United Kingdom (discussed below).
• 6Follow-up Study, p. 28, n. 22. The footnote identifies the involvement of technical standards or the expiry of the patent as reasons for refusing relief; it does not identify or classify the individual cases.
• 7CJEU, 16 July 2015, C-170/13, Huawei Technologies v. ZTE, ECLI:EU:C:2015:477, paras 53-55 and 60-71, judgment; ETSI IPR Policy, clause 6.1. Decisions applying this framework, before the UPC and in national courts, are collected in the FRAND Litigation Watch.
• 8M. Dhenne, “FRAND Is Not a Number: The Prior Question About Standard-Essential Patents”, Kluwer Patent Blog, 28 May 2026; M. Dhenne, A Contractual Theory of FRAND: From Private Normativity to Legal Obligation in Technical Standardization, Wolters Kluwer, 2026.
• 9BGH, 27 January 2026, KZR 10/25, FRAND-Einwand III (VoiceAge EVS v. HMD), ECLI:DE:BGH:2026:270126UKZR10.25.0, paras 44-46 (continuing willingness), 85-92 and 96 (security and the question left open), 97-108 (proportionality and application to the facts), judgment.
• 10M. Dhenne, “Munich’s FRAND Guidelines: A Compass Drawn Too Finely”, Kluwer Patent Blog, 1 October 2026.
• 11M. Leistner, Legal Assessment of the Intellectual Property Rights Enforcement Directive, briefing requested by the JURI Committee, PE 787.219, July 2026, pp. 15 and 24-25; see also pp. 10-12 on the UPC and unresolved third-party-interest questions, briefing. The views are the author’s, not the official position of the European Parliament.
• 12Follow-up Study, p. 28, especially nn. 21 and 22. The published report describes the 635-decision corpus but does not supply its case list or its complete selection and coding protocol.
• 13IP2Innovate, “IP2I’s statement on the European Commission publication of the follow-up study on the application of the Directive on the Enforcement of Intellectual Property Rights (IPRED)”, Brussels, 27 January 2026, statement.
• 14Zweites Gesetz zur Vereinfachung und Modernisierung des Patentrechts, 10 August 2021, BGBl. I 2021, p. 3490, art. 1 no. 36 and art. 13(1), promulgated on 17 August 2021 and in force for this amendment on 18 August 2021; see the final parliamentary text, BT-Drs. 19/30498, pp. 12 and 40, Bundestag, and DPMA press release, 18 August 2021, DPMA. The UPCA entered into force on 1 June 2023.
• 15Follow-up Study, pp. 32-34, Table 4; Annex I, PDF pp. 133-135, and Annex V, PDF pp. 162-163, for the inclusion and exclusion criteria and information to be extracted. The statement concerning the eleven responding judges is on p. 32.
• 16Directive 2004/48/EC of the European Parliament and of the Council of 29 April 2004 on the enforcement of intellectual property rights, art. 3(2), EUR-Lex.
• 17Directive 2004/48/EC, arts 11 and 12, EUR-Lex.
• 18Follow-up Study, pp. 20-22; Annex II, PDF p. 140, identifying EY Consulting, Belgium, and Technopolis Group. The commissioning service is DG GROW, Directorate D, Unit D.3, Intellectual Property (p. 2).
• 19Follow-up Study, p. 28 and n. 23.
• 20Follow-up Study, p. 68, under “Consideration of PAEs’ nature”.
• 21UPC Court of Appeal, 25 November 2025, Meril v. Edwards Lifesciences, EP 3 646 825, UPC_CoA_464/2024 and connected appeals, including UPC_CoA_21/2025 and UPC_CoA_27/2025, paras 188-190, decision; Agreement on a Unified Patent Court, art. 63(1).
• 22UPCA, arts 42, 62(2) and 63(1); Rules of Procedure, R. 211.3 (provisional measures), EUR-Lex.
• 23Meril v. Edwards, cited above, paras 191-202 and operative part, points IX-X. The exception concerns Meril’s XL heart valves and requires a physician’s confirmation that the device is the only available treatment option for the individual patient.
• 24Follow-up Study, p. 63 and n. 163. The report dates the German amendment to 2022, whereas the relevant change entered into force on 18 August 2021.
• 25Follow-up Study, pp. 35-37 on national factors and p. 49 on judicial knowledge sharing.
• 26For a published example of this disclosure, limited to validity and infringement outcomes before the UPC, see the method notes of the UPC statistics in the UPC Litigation Watch.
• 27Follow-up Study, p. 69, under “Impact of the UPC”, recommending observation of the developing case law before a potential amendment.