UPC Appeal Rates and Case Distribution: Europe Is Not a League Table

A court of appeal is not a scoreboard, and it is worth saying so plainly at this stage of the debate on the Unified Patent Court. The discussion about how cases are distributed across the UPC’s divisions has lately drifted toward appeal statistics,1 which was probably inevitable: in a young court, figures are attractive. They seem to offer neutrality where impressions are contested, to replace anecdote with evidence, and to provide a way out of the national perceptions that have shadowed European patent litigation for decades.

Yet numbers do not speak for themselves. Appeal rates can be genuinely useful — they reveal pressure points, show where first-instance practice still needs clarifying, and trace the way the Court of Appeal is beginning to structure the new system. What they cannot do, on their own, is measure the quality of a division, still less justify recasting the UPC as a contest between national litigation cultures. The Court deserves a better use of its data than that.

From case distribution to appeal statistics

The debate began somewhere else. Its original subject was the concentration of infringement actions before a handful of local divisions2 — a sensitive enough topic in itself, touching on user choice, procedural confidence, inherited litigation habits and the institutional balance of a court shared by several Member States. It has since shifted, at least in part, toward a different question: what do appeal outcomes tell us about the quality of first-instance decisions?

The question is legitimate. Any serious assessment of the UPC should include some empirical monitoring of what happens on appeal, and there is nothing improper about looking at reversal rates. But there is a real difference between using data to understand a court and using it to rank the court’s divisions, and the two are easily confused. The first is useful; the second misleads.

What the figures can and cannot show

Taken properly, an appeal rate can tell us a number of things. It can show that a particular kind of decision is hard to stabilise, that certain procedural questions remain unsettled, or that the Court of Appeal is actively clarifying the standards governing urgency, evidence, confidentiality, claim construction, proportionality or remedies. It can show, in other words, where the law is being tested most intensely. Read this way, a higher number of appeals or reversals from a given division may reflect nothing worse than exposure: a division that hears the most complex and most urgent cases will naturally generate more appellate material; a division that users trust will decide more disputes; and a division that decides more disputes will, inevitably, appear more often before the Court of Appeal. That visibility is easily mistaken for weakness. It usually is not.

What an appeal rate cannot do is measure quality directly. A first-instance decision may be reversed because it was wrong; it may equally be reversed because the law was genuinely open, because the record was incomplete, because the matter was urgent, or because the Court of Appeal wanted to refine a test that the first-instance panel had been obliged to apply before any appellate guidance existed. In the opening years of a court, that last situation is common rather than exceptional. The Court of Appeal is not simply catching mistakes; it is building a body of law — confirming, reversing, narrowing and reformulating first-instance reasoning, and sometimes reordering it altogether. That is not a defect in the system but one of the ways the system acquires coherence. Early correction often signals formation rather than dysfunction.

What the numbers are actually counting

It also matters what one is counting. Statistics depend on their denominator, and the UPC’s figures mix categories that are not really comparable. A provisional measure cannot be set against a full merits decision: urgent proceedings do not pose the same questions as a dispute decided after complete written and oral argument, and an appeal against a preliminary injunction is not the methodological equivalent of an appeal following a final ruling on infringement and validity.3 Sample size compounds the problem. Where the relevant universe is a handful of decisions, percentages become fragile and a single case can move the result by a wide margin.4 None of this makes the figures worthless; it makes them a starting point rather than a conclusion — a way of raising questions, not of settling them.

Procedural posture deserves the same care. A reversal means different things depending on whether the decision under appeal was a provisional measure, a judgment on the merits, a procedural order, or a ruling on confidentiality, security, amendment, language or access to evidence. A reversal on urgency does not carry the message of a reversal on claim construction; modifying an injunction is not the same as holding that the applicable legal test was wrong; and reasoning that has been corrected has not necessarily produced an unsound outcome. If the UPC is going to use appeal statistics intelligently, it has to disaggregate them — otherwise figures that look precise will yield conclusions that are merely crude.

The national temptation

The deeper risk here is not statistical but conceptual. The UPC was created precisely to overcome the fragmentation of European patent enforcement, and it should not now be described as though its divisions were national teams competing for the best record on appeal. There is no need to defend one tradition against another. The German local divisions have done a great deal to establish the Court’s early credibility, and their experience and procedural discipline are plainly an asset to the whole system; the other divisions, including those still lightly used, are no less part of its promise, because a court meant to be common has to become common in practice. The right question is not which country is winning. It is whether the UPC is producing a body of law that can genuinely be recognised as common.

The role of the Court of Appeal

This is where the Court of Appeal’s function comes into focus, and it is more demanding than refereeing a national fixture. Its task is to turn first-instance disputes into shared doctrine: to decide which procedural instincts should travel across divisions and which should stay local experiments, to fix standards, and to discipline variation without flattening the diversity that a multi-divisional court necessarily contains. A first-instance decision is not just affirmed or overturned; it is absorbed into a larger work of legal construction, and even a reversal can contribute to that work if it clarifies the law for everyone. Read in that light, appeal statistics are less like marks awarded to pupils than like traces showing where the law is currently being made.

A better empirical agenda

The remedy is not to abandon statistics but to improve them. Serious empirical monitoring of UPC appeals would distinguish, at a minimum:

• the type of decision appealed;
• the division of origin;
• the procedural posture;
• the technical field;
• the legal issue at stake;
• whether the appeal turned on law, facts, evidence, procedure or remedy;
• whether the Court of Appeal reversed the outcome, modified the reasoning, narrowed the order or simply clarified the standard; and
• whether the appellate decision produced a rule capable of travelling across divisions.

Monitoring of that kind would not rank anyone. It would help us see how the UPC is actually forming its law.

Back to case distribution

This brings the argument back to case distribution, because the two are connected — though not in a simple way. If a large share of first-instance litigation is concentrated in a few divisions, those divisions will naturally supply much of the material from which the Court of Appeal builds its doctrine. That may be entirely normal, and in the Court’s early phase it may even be useful, since the most experienced divisions help generate the first layer of UPC case law. But it also confirms the institutional point: the way cases are distributed shapes the conditions under which appellate doctrine is made. The debate is therefore not only about workload, delay or user preference. It is also about the ecology of law-making in a common court.

Europe is not a league table

None of this is an argument for German-bashing, or for French-bashing, or for a contest between Paris, Munich, Düsseldorf, Milan, The Hague or anywhere else. That vocabulary misses what the UPC is. It is not a set of national courts under a shared logo but a common court whose European character still has to be built through practice — through first-instance case management, panel composition, the behaviour of users, appellate correction and the Court’s own self-assessment.

The UPC’s success is not in doubt. But success is a reason for closer analysis, not an exemption from it: a struggling court worries about survival, whereas a successful one has to worry about the quality, coherence and legitimacy of the law it produces. A young court needs data, transparency and empirical monitoring; it also needs concepts. Without data, analysis drifts into impression; without concepts, data hardens into ammunition.

So the question worth keeping in view is not which division has the best record on appeal. It is whether appeals, case distribution and judicial practice are together helping the UPC become what it set out to be — a genuinely European patent court. Europe is not a league table, and European patent law should not be built as one.

• 1

  M. Dhenne, “Du Hast, But Not So Fast: Lessons from Two Years of UPC Appeals,” Kluwer Patent Blog, 2 June 2025. See also M. Dhenne, “Where Should Cases Go? Or Where Should Patent Law Be Made?,” Kluwer Patent Blog, 7 May 2026.

• 2

  Unified Patent Court, Caseload of the Court (statistics update of 5 June 2025, covering 1 June 2023–31 May 2025): approximately 75% of all infringement actions were filed before the four German local divisions (Munich, Düsseldorf, Mannheim and Hamburg).

• 3Disaggregated by procedural posture, the published two-year figures show a markedly higher appellate reversal rate for provisional measures (approx. 46.2%) than for infringement actions (approx. 31.6%) or revocation actions (approx. 20%): Dhenne, “Du Hast, But Not So Fast,” op. cit.
• 4Ibid. (illustrating the fragility of small samples: a single Milan appeal at that stage translated into a 100% reversal rate — a single data point incapable of establishing a trend).