Four Questions That Decide a UPC Market Entry

The first question before a European launch is not whether the product infringes. It is where, against whom, on what record and on what timetable that question will be tested. Market-entry opinions nonetheless tend to collapse into a single sentence: launch, wait or redesign. Management likes the compression because it looks like a decision. Before the Unified Patent Court, it often postpones one.

The reason is structural. A single product now generates acts in several Contracting Member States, may involve several entities of the same group, and can be met by one application for provisional measures whose effect extends to every territory in which the patent has effect.1 The same patent may, however, be opted out, validated in only part of that territory, held by someone other than the registered proprietor, or exposed to a revocation action that the implementer can bring before the central division without first filing an opposition at the European Patent Office.2 The exposure is real, and so is the implementer’s room for maneuver. Neither appears in a single rating.

An in-house team that asks outside counsel in the spring whether it can launch, and schedules the stock release for the fall, has already made its most consequential choice. It has allowed the calendar to be written by the party that does not bear the launch.

Four questions, one chain

A UPC market-entry decision should answer four linked questions. Which claimed technical feature or combination could the commercial product actually practice? Which acts, entities and territories create exposure? What evidence can each side already rely on, and what further evidence will the launch itself create? Which procedure could alter the commercial timetable?

The questions form a chain. The claimed feature determines which acts matter, the acts determine which records a court will read, those records determine how early a credible application can be filed and that date decides whether the analysis of infringement and validity reaches the business while it can still change anything. Procedure determines the commercial and temporal consequences of that analysis, which is what a launch decision is about. The output is a launch architecture: conditions that can be audited, assigned, dated and reviewed when the facts move.

The claimed feature behind the patent list

The first question asks which claimed technical feature, or which claimed combination of features, the commercial product could actually practice. The unit of analysis is the claim as properly construed. The Court of Appeal has made the point that the claim is not merely the starting point but the decisive basis for determining the scope of protection under Article 69 EPC and its Protocol, with the description and drawings serving as explanatory aids.3 What matters is therefore neither the title of the patent nor the technical idea behind it, but the construed claim, read against the product as it will ship.

The exercise is iterative. A first reading of the product identifies the features that could matter, a search identifies the claims that recite them, a closer reading of those claims sends the engineers back to the product to check a feature nobody had thought relevant. For each claim, the team should map the product version, component, process step or system interaction said to satisfy each feature. Three categories of fact should be kept apart: facts visible from public material, facts held only internally, and facts a claimant could obtain by testing the product or through an order to preserve evidence, which the Court may grant before any action on the merits on reasonably available evidence of actual or imminent infringement.4

The same map should record the principal non-infringement construction, the realistic design alternatives and the prior-art or entitlement issue most capable of changing the balance. Validity is fought in provisional proceedings as well as on the merits. The Court of Appeal has held that the sufficient degree of certainty required for provisional measures is lacking where the Court considers it more likely than not that the patent is invalid.5 A documented and serious validity attack therefore has immediate procedural value, on one condition: it must exist when the application arrives, not be commissioned after service.

A long patent list with a single rating does not tell management which feature carries the exposure. A feature-level map tells it which engineering decision would remove it, and at what cost.

Acts, entities and territories

The second question asks which acts, entities and territories create exposure. Patent exposure follows acts, and each commercial act must be characterized before it can be counted. The Agreement gives the patentee the right to prevent the making, offering, placing on the market or use of the patented product, and its importation or storage for those purposes. It also allows the patentee to prevent the supply of means relating to an essential element of the invention, but only where the recipient is not entitled to work it, where the supplier knows or should know that the means are suitable and intended for that purpose within the relevant territory, and outside the exception for staple commercial products.6 Each condition is a question of fact. A component shipped to an installer is not an act of indirect infringement until those requirements have been examined, and a demonstration at a trade fair is not an offer until its content has been.

The group chart is only a starting point. Manufacture may take place outside Europe while importation, offer, tender participation, storage, installation, software activation or use occurs within it, each performed by a different entity, and a feature unlocked remotely after delivery may complete the claimed combination in a territory the commercial team never planned to enter. Some of these steps are preparatory acts that create no liability on their own; others may already constitute infringement. The distinction governs both exposure and the moment from which the patentee can act.

Four distinct issues are then easily confused and should be kept apart. International jurisdiction is governed by the Brussels I bis Regulation, as adapted for the Court by Articles 71a to 71d, and turns principally on the defendant’s domicile, the place of the harmful event and the joinder of co-defendants.7 The allocation of a case between divisions follows Article 33 UPCA: an infringement action may be brought where the actual or threatened infringement occurs, or where one of several defendants is established, provided that the defendants have a commercial relationship and the action relates to the same alleged infringement.8 The territorial scope of a decision on a European patent is that of the Contracting Member States for which the patent has effect,9and for a unitary patent that territory is fixed at the date of registration of unitary effect.10 The applicable law, finally, is the law of the patent concerned: Union law, the Agreement and the EPC for the designations within the Court’s territory, national law for the others.11

The first of these questions has moved. The Court of Justice held in BSH Hausgeräte v. Electrolux that the court of the defendant’s domicile retains jurisdiction over the infringement of a patent granted for another State even where invalidity is raised as a defense.12 The Court of Appeal has since confirmed, in Fujifilm v. Kodak, that the Court may hear an infringement claim concerning the United Kingdom designation of a European patent against defendants domiciled in a Contracting Member State, Article 34 UPCA governing the effect of decisions rather than international jurisdiction; it set out how objections to the validity of the foreign designation are to be handled and set aside the injunction on the facts.13 In Dyson v. Dreame, it referred to the Court of Justice the question whether provisional relief can reach a third-State defendant and its EU-based authorized representative for acts in a Member State outside the Court’s territory, while granting provisional measures for the Court’s own territory; the reference is pending as Case C-196/26.14 Neither an automatically Europe-wide reach nor a general confinement to the participating States describes the position. What can be said is that an EU-domiciled entity in the launch chain can carry the dispute further than the map of the patent suggests, and that the answer for provisional measures against a third-State manufacturer is not yet known.

For a classic European patent, three further questions arise: whether an opt-out has been registered, whether it was validly lodged by all proprietors, and whether a national action has since been brought that would prevent its withdrawal.15 I have argued in these columns that the register entry answers none of them conclusively.16 The ownership chain, the national validations and any pending opposition belong on the same map.

The resulting map frequently reveals that a supposedly single launch consists of several legally distinct launch events. An order addressed to one entity may require operational changes by others, while a claim against one entity may not reach every market or every related right.

The record a launch creates

The third question asks what evidence each side can already rely on, and what further evidence the launch will create. Market entry generates evidence before litigation begins. Regulatory filings, pricing submissions, tender responses, customer letters, product pages, technical manuals, released stock and internal authorizations may show that a launch is imminent. The Agreement allows injunctions intended to prevent an imminent infringement.17 The Court of Appeal describes the situation as one in which the potential infringer has already set the stage, so that the infringement is only a matter of starting the action, and asks whether it is more likely than not that an infringement is imminent. For generic medicines, the mere application for or grant of a marketing authorization does not suffice, whereas completion of the national pricing and reimbursement procedures may.18

That case law grew out of a regulated sector in which the last administrative step before sale is public and datable. Its transposition to other industries is an argument, not a given. A confirmed purchase order with a delivery date, an acceptance test on the customer’s premises or a firmware release announced in a public changelog may play the role that reimbursement listing plays for a generic, but the analogy must be made act by act, by asking whether the document shows that nothing remains but to start. A demonstration unit or an internal go decision that still depends on external events will usually not.

The same record bears on urgency, though not mechanically. The Court must have regard to any unreasonable delay in seeking provisional measures. The Court of Appeal counts that delay from the day on which the applicant became aware, or should have become aware, of the infringement or of its imminence, such that an application could be filed with a reasonable prospect of success, and whether the delay is unreasonable depends on the circumstances.19 In Merz v. Viatris, the Court left open whether awareness of an imminent infringement also starts the clock for the subsequent actual infringement, and it required proof that the relevant knowledge had reached the patentee rather than an intermediary.20 A warning letter or a public statement is therefore a fact to be weighed for what it proves about knowledge and about the stage of preparation. By itself it starts and stops nothing. The drafting of launch communications in light of this case law was the subject of an earlier column.21

The company should keep two records, and keep them apart from its legal analysis. The technical record identifies the commercial configuration, its release date, the feature map, the tests performed and the design decisions taken, with the version-control references that will later prove what the launched product did. The launch record identifies the regulatory or contractual steps completed, the acts still blocked, the objective event that will release them, the person authorized to decide, and every external communication with proof of transmission. The legal assessment of those facts, the claim charts prepared with counsel and the related correspondence form a third body of material, which the Rules of Procedure protect. Confidential communications with a lawyer or patent attorney, including one employed by the client, and the resulting work product are privileged from disclosure before the Court; confidential communications with third parties for the purpose of obtaining evidence for proceedings enjoy litigation privilege.22 The protection does not extend to the underlying facts. The Court may order a party to produce specified evidence in its control, may order the preservation of evidence and the inspection of premises, and manages confidential material through protective regimes rather than by excluding it.23 Neither the assumption that internal documents are unreachable nor the assumption that everything is discoverable is a safe planning basis.

One further point should be made plainly. Documenting the launch as it is protects the company because the Court reconstructs the sequence from the record. A record built to suggest that the stage has not been set, when it has, does not create an absence of imminence. It creates an exposed witness.

The procedural clock

The last question asks which procedure could alter the commercial timetable. A position that is defensible on the merits may be worthless commercially if provisional measures can stop the launch before the validity debate has run its course.

Several instruments move the calendar, and each has a cost. A revocation action lies before the central division without prior opposition, but where an infringement action between the same parties on the same patent is already pending before a local or regional division, the revocation action must be brought there.24 An action for a declaration of non-infringement lies before the central division under the same reservation, and it is stayed if the patentee brings an infringement action between the same parties on the same patent before a local or regional division within three months.25 A protective letter places the implementer’s arguments and evidence before the judge who receives an application; the Court must take it into account and must in particular consider summoning the parties to a hearing, but the letter guarantees no hearing, and it is removed from the register after six months unless extended.26 Licensing discussions and a redesign program move the calendar too, only more slowly.

The patentee’s own constraints are precise. An application for provisional measures may be lodged before or after proceedings on the merits have started. Where it precedes them, the Court must revoke the measures at the defendant’s request if the applicant does not start proceedings on the merits within a period, fixed in the order, that may not exceed 31 calendar days or 20 working days, whichever is longer, counted from the date the order specifies; and where the measures are revoked, lapse through the applicant’s act or omission, or are followed by a finding that there was no infringement or threat of infringement, the Court may order the applicant to compensate the defendant.27 The Court may also require security, and must do so where measures are ordered without hearing the defendant unless special circumstances justify otherwise.28 They make a poorly prepared application expensive. They do little to deter a prepared one.

The matrix that follows gathers what these dates require. Procedure belongs inside the freedom-to-operate analysis from its first page, because it decides when that analysis can still affect the business.

The matrix

A worked example

The facts are invented. A manufacturer established outside the Union makes an industrial heat pump. Its controller runs a defrost routine that starts a cycle when the difference between coil and ambient temperature exceeds a threshold. Firmware version 4.2 adjusts that threshold with a humidity reading; version 4.3, planned for the spring, adjusts it with compressor run time as well. A European patent with unitary effect, registered in 2024, claims a heat pump system with a controller configured to initiate defrost on such a differential where the threshold is adjusted as a function of compressor run time. The claim does not mention humidity. Units go from the factory to a bonded warehouse in the Netherlands run by a group subsidiary, are offered by a German subsidiary, sold to installers in France and Italy who activate them with a service tool, and distributed in Spain by an independent distributor.

The claimed feature that matters is the construed run-time adjustment, not the defrost routine as a whole. On a first reading, version 4.2 falls outside the claim and version 4.3 inside it; the construction of “adjusted as a function of” against a firmware that combines several inputs is the point the patentee will litigate, and it is where the engineers’ description of the algorithm has to be exact. The acts are distributed: importation and storage by the Dutch subsidiary, offers by the German one, activation and use by installers, all within the unitary territory; Spain lies outside it, but if the same European patent was also validated in Spain, a claim against the two EU-domiciled subsidiaries may bring the Spanish part into the discussion on the basis of their domicile. The record will contain the changelog announcing 4.3 and the German subsidiary’s tender responses. A tender response describing the run-time feature as available from a stated date could support imminence; a roadmap slide shown at a trade fair is far weaker.

Two decisions are available, and each comes with its own review triggers. The first is to launch version 4.2 across the territory, keep version 4.3 out of the tender responses and the installer tool until the prior-art search on run-time compensation is complete, and fix a date for that search. The condition is reviewed if a warning letter arrives, if the search produces a document that makes the more-likely-than-not threshold realistic, or if the patentee files a limitation. The second is to launch version 4.3 with the run-time input disabled at the factory and activatable only outside the unitary territory, the German offers amended and the installer tool reconfigured before the first delivery. That condition is reviewed if the construction analysis concludes that a combined humidity and run-time adjustment is caught in any event, or if a customer in a UPC state requires the feature, at which point the decision returns to management with the validity file as it then stands. Neither decision is a rating. Both can be audited.

The patentee’s side, and the cost objection

The same four questions work in reverse. A well-advised patentee maps which feature of the forthcoming product falls within its best-construed claim, which entity it can reach through an EU domicile, which public record will show that the stage has been set, and which division offers the quickest credible hearing. Its best moment is the day the launch becomes provable while the implementer’s validity case is still a draft.

The serious objection to the method is its cost. A launch program that treated every family in a clearance search this way would never ship. The method is meant to be targeted: the feature-level analysis is reserved for the claims that survive a first screen for construction, validity and territorial reach, which is what a competent freedom-to-operate opinion already does, and the better ones already build the procedural timetable in. Nor is a synthetic indicator worthless. A rating is useful when it sits on top of a traceable analysis and can be unpacked into the facts that produced it. It fails when it replaces that analysis, because the Court will not weigh a rating. What the Court weighs, on any application, are the interests of the parties and the harm each would suffer from the grant or the refusal of the measure.29 That weighing rewards no one for a well-kept file. The record matters because it establishes the facts on which the weighing turns, including the time needed for a workaround and the reversibility of the harm.

Launch conditions, not risk labels

The most pressing use of the grid concerns products already on the calendar. A freedom-to-operate opinion written before the case law on imminent infringement and delay took its present shape, or before a competitor withdrew an opt-out, may describe a position that no longer exists: the transitional period runs for seven years from 1 June 2023, subject to extension, and a proprietor may withdraw its opt-out at any time unless a national action has already been brought.30 An opinion that cannot say which claimed feature matters, which entity performs which act, what evidence each side can rely on and when a court could first be seized is not wrong. It is incomplete in the one respect that decides the commercial outcome.

The four answers should be converted into conditions rather than into a label. Version A ships only in identified territories. Specified tender, shipment or activation acts remain blocked until an objective event occurs. The approved configuration and the launch communications are preserved. A targeted prior-art or experimental workstream is completed by a fixed date, before stock leaves the warehouse. Any change to the product or to the public message requires approval at decision level, and the first-response team and the redesign path are in place before the first delivery.

Each condition has an owner, a supporting record and a trigger for its review. The real launch decision concerns control: whether the company commands the acts, the evidence and the timing through which a patent risk becomes an injunction. Those three should be settled first, and the launch date fixed only once they are.

• 1Agreement on a Unified Patent Court, OJ C 175, 20 June 2013, p. 1 (“UPCA”), Articles 32(1)(c), 34 and 62(1).
• 2UPCA, Articles 32(1)(d), 33(8) and 83(3); Regulation (EU) No 1257/2012 of the European Parliament and of the Council of 17 December 2012 implementing enhanced cooperation in the area of the creation of unitary patent protection, OJ L 361, 31 December 2012, p. 1, Article 18(2), second subparagraph.
• 3UPC Court of Appeal, 26 February 2024, UPC_CoA_335/2023, APL_576355/2023, NanoString Technologies Inc. and others v. 10x Genomics Inc. and President and Fellows of Harvard College; European Patent Convention, Article 69 and Protocol on its interpretation.
• 4UPCA, Article 60(1) to (3).
• 5UPC Court of Appeal, 26 February 2024, UPC_CoA_335/2023, cited above; Rules of Procedure of the Unified Patent Court (“RoP”), Rule 211.2.
• 6UPCA, Articles 25 and 26.
• 7Regulation (EU) No 1215/2012 of the European Parliament and of the Council of 12 December 2012 on jurisdiction and the recognition and enforcement of judgments in civil and commercial matters, OJ L 351, 20 December 2012, p. 1, Articles 4(1), 7(2), 8(1) and 71a to 71d; UPCA, Article 31.
• 8UPCA, Article 33(1), first subparagraph, points (a) and (b).
• 9UPCA, Article 34.
• 10Regulation (EU) No 1257/2012, Article 18(2), second subparagraph.
• 11UPCA, Article 24(1).
• 12

  Case C-339/22, BSH Hausgeräte GmbH v. Electrolux AB, EU:C:2025:108, 25 February 2025, operative part, points 1 and 2. A long arm, with certain limits: SeeMatthieu Dhenne, 
  "The UPC’s Long Arm and the Limits of Its Reach", Kluwer Patent Blog, 10 March 2026.

• 13UPC Court of Appeal, 2 June 2026, UPC_CoA_312/2025, UPC_CoA_333/2025, UPC_CoA_880/2025 and UPC_CoA_882/2025, Fujifilm Corporation v. Kodak Holding GmbH and others.
• 14UPC Court of Appeal, 6 March 2026, UPC_CoA_789/2025 and UPC_CoA_813/2025, Dyson Technology Ltd v. Dreame International (Hongkong) Ltd and Eurep GmbH; Case C-196/26, Dreame International, reference lodged on 11 March 2026, pending.
• 15UPCA, Article 83(3) and (4); UPC Court of Appeal, 4 June 2024, UPC_CoA_79/2024, APL_9578/2024, Neo Wireless GmbH & Co. KG v. Toyota Motor Europe NV/SA; UPC Court of Appeal, 12 November 2024, UPC_CoA_489/2023 and UPC_CoA_500/2023, AIM Sport Development AG v. Supponor Oy and others.
• 16

  Matthieu Dhenne, “An Opt-Out Is Not a Register Entry: Ownership and the Limits of Procedural Choice Before the UPC”, Kluwer Patent Blog, 15 September 2026.

• 17UPCA, Article 62(1); RoP, Rule 206.2(c).
• 18

  UPC Court of Appeal, 13 August 2025, UPC_CoA_446/2025 and UPC_CoA_520/2025, Boehringer Ingelheim International GmbH v. Zentiva Portugal, Lda, headnotes and paras. 46 and 52. SeeMatthieu Dhenne, "Imminent Infringement: The UPC’s New Geography of Pharmaceutical Risk", Kluwer Patent Blog, 30 October 2025.

• 19RoP, Rule 211.4; UPC Court of Appeal, 25 September 2024, UPC_CoA_182/2024, APL_21143/2024, Mammut Sports Group AG v. Ortovox Sportartikel GmbH; UPC Court of Appeal, 27 April 2026, UPC_CoA_917/2025, Merz Pharmaceuticals LLC and others v. Viatris Santé, paras. 48 to 50.
• 20

  UPC Court of Appeal, 27 April 2026, UPC_CoA_917/2025, Merz v. Viatris, cited above, paras. 54, 57 and 58. However, caution is warranted, as the analytical framework proposed by the Court arguably creates more uncertainty for all market participants than it resolves: See Matthieu Dhenne, "Merz v. Viatris on Appeal: Urgency Is Still Everything But Not Every Signal Is Urgent", Kluwer Patent Blog, 19 May 2026.

• 21

  Matthieu Dhenne, “The Launch Letter: How Generics and Biosimilars Should Write to Avoid UPC Urgency”, Kluwer Patent Blog, 8 September 2026.

• 22RoP, Rules 287.1 to 287.4 and 288.
• 23UPCA, Articles 59 and 60; RoP, Rules 190, 192 to 199 and 262A.
• 24UPCA, Articles 32(1)(d), 33(4) and 33(8).
• 25UPCA, Articles 32(1)(b), 33(4) and 33(6).
• 26RoP, Rules 207.1, 207.9 and 209.2(d).
• 27RoP, Rules 206.1, 213.1 and 213.2.
• 28RoP, Rule 211.5; UPCA, Article 62(5).
• 29UPCA, Article 62(2); RoP, Rule 211.3.
• 30UPCA, Article 83(1), (4) and (5).