EPO FRAND Study: Comparables, Top-Down and a Questionable Convergence Narrative

The EPO’s recent study, Methodologies for FRAND determination: evidence from global case law, is an extensive account of judicial approaches to evaluating fair, reasonable and non-discriminatory terms (FRAND) for standard essential patents (SEPs). It covers 65 decisions from seven jurisdictions and identifies comparable licences and top-down analysis as the principal methods used in FRAND-rate litigation.

Its breadth makes it a useful reference point of what courts are currently doing. But the study is less convincing as an account of what courts should do and sometimes presents contested legal and economic choices as proof of settled international consensus. Its treatment of comparables, US law and “willingness” requires some clarification.

Comparables are not necessarily primary

The study presents comparable licences as the dominant valuation method, with top-down analysis often secondary or a cross-check. This reflects, for example, InterDigital v Lenovo (2024), where the UK Court of Appeal preferred a rate derived from comparables when it diverged from a top-down calculation.

Yet judicial recurrence does not establish methodological superiority. Courts may turn to comparables because parties put them in evidence or because there is evidence that the circumstances surrounding the licenses indicate they represent what the market has accepted to be fair market value. Neither point shows that valuation based on comparable licenses is more reliable than a structured top-down exercise.

The study accepts the methodology that courts must determine whether licences are genuinely FRAND comparable, derive an effective royalty rate and adjust for differences in portfolios and transaction terms. These steps are not merely technical adjustments. They are all critical and non-trivial matters for courts to adjudicate.

Instead, before a license should be even accepted as a FRAND comparable license for establishing a FRAND rate, it is critical that the court first inquire whether the circumstances under which the license was entered into were conducive to establishing FRAND terms.  For example, the rates used in a license with a total royalty value below the cost of negotiation and/or litigation are likely to be based on avoidance of transactional costs. Likewise, even a licence concluded in the shadow of litigation may be driven by the parties’ assessment of procedural risk, costs, delay and the commercial disruption of an injunction. In these cases, the terms are based on factors extrinsic to the economic value of the underlying patents which forms the basis of the FRAND value. Treating such agreements as clean price signals may therefore risk converting leverage into a benchmark for subsequent licensees.

Even if the circumstances of a license proffered as a comparable indicate the license should not be rejected, determining the actual terms is far from trivial. Lump sum license agreements require assessing the number of units covered, which means assessing the number of units the parties anticipated the license would cover at the time of the agreement. This can frequently require looking beyond what was explicitly licensed, given that license agreements frequently include creative and non-obvious ways of establishing broader ‘patent peace’ that operate as a broader de facto license. Likewise, the existence of a cross-license agreement requires establishing the value of the counterparty’s portfolio, and differences in the geographic scope of the licensor’s patent coverage and licensee’s market distribution add further complicating factors.

Top-down analysis is also difficult. It requires a credible aggregate royalty burden and a rational allocation to the relevant portfolio. Essentiality, portfolio strength and relative technical contribution need close scrutiny. The study is right to warn against crude patent counting. Still, top-down has an independent function. It begins with the aggregate burden imposed by the standard and creates a logical ceiling against royalty stacking. That systemic perspective is obscured when it is treated merely as a tool for confirming a result generated by comparables.

The point is not that a top-down approach invariably supplies the right answer. Its aggregate-rate premise may itself be contested. But it forces the court to confront a question that bilateral comparisons can leave unresolved: i.e. whether the cumulative demands of all relevant SEP holders remain consistent with a commercially sustainable royalty burden for the standard’s implementers.

If top-down and comparables converge, two distinct methods support the outcome. Where they diverge, courts should scrutinise both, instead of presuming that opaque prior licences prevail. The earlier WIPO study on SEP valuation is more helpful here: it analyses methodologies through their economic rationale and explains how structured valuation may improve objectivity and dispute resolution.

US law: apportionment, not symmetry

The study suggests that US law moved from concern with hold-up towards a more balanced approach, exemplified by the Federal Circuit’s decisions in Ericsson v D-Link (2014) and CSIRO v Cisco (2015), requiring evidence before a jury is instructed on hold-up or royalty stacking. That is a thin basis for a broader convergence narrative. It elevates an evidential rule about jury instructions over the Federal Circuit’s principal substantive message: apportionment.

In Ericsson v D-Link, the court held that an implementer seeking a hold-up or stacking instruction must offer evidence connecting the concern to the particular dispute. This was a procedural evidentiary ruling about jury instructions, not a broad doctrinal shift in how FRAND is interpreted. Specifically, D-Link was centrally concerned with damages. The court found that the instructions did not adequately reflect the RAND commitment and the need to apportion the royalty. A SEP royalty must capture the value of the patented invention, not the value created by incorporation into an industry standard. The court required two forms of apportionment:

• The patented feature must be separated from unpatented features in the standard.
• The calculation must exclude value created by standardisation rather than the patented technical contribution.

That rule does not depend on proof of hold-up by the SEP holder. It is a structural limit on reasonable royalty damages. A patent holder cannot recover the market-wide value generated by adoption of the standard merely because its technology became essential to it.

CSIRO v Cisco confirms this position. The Federal Circuit held that the royalty must track the approximate incremental value of the patented invention and exclude additional value resulting from incorporation into the standard. This rule applied although the SEP lacked a RAND undertaking, confirming that the principle rests on the risk of standardisation inflating apparent patent value beyond technological contribution.

The study is correct that abstract assertions cannot justify a hold-up instruction. But it should not recast this procedural proposition as evidence that US law now treats hold-up and hold-out symmetrically. D-Link and CSIRO are primarily authorities on valuation and the exclusion of standardisation rents.

Willingness serves different functions

The study treats the willing-licensor/willing-licensee construct used to calculate reasonable royalty damages in U.S. cases as analytically equivalent to the UK and EU constructs, used to set prospective FRAND terms or evaluate injunctive relief. However, although the frameworks discuss “willing” parties, they serve different purposes.

In the United States, the Georgia-Pacific factors are used to calculate reasonable royalty damages for past patent infringement through a hypothetical negotiation between a willing licensor and willing licensee. SEP cases adapt these factors to reflect RAND constraints and prevent capture of standardisation leverage. After the court in D-Link criticised the insufficiently adjusted use of conventional factors in SEP cases, courts began adapting the factors to reflect RAND constraints and prevent capture of standardisation leverage in SEP cases. The hypothetical negotiation is a compensatory fiction used as a tool to quantify damages; it does not determine whether an implementer is, in fact, a willing licensee.

The UK and EU willingness framework plays a different role. English courts use the willing-party construct to set prospective FRAND terms, as Birss J explained in Unwired Planet. Under the EU Huawei v ZTE (2015) precedent, willingness may also affect whether pursuit or enforcement of an injunction abuses dominance. Willingness in the UK and EU determines price, negotiation conduct and access to injunctions, not damages.

The remedies contrast is stark. In the US, injunctions are governed by the equitable four-factor test in eBay v MercExchange (2006). None of the eBay factors involve a Huawei-style willingness enquiry to assess negotiating conduct.   

The frameworks should therefore not be treated as interchangeable. Conflating these enquiries also risks overlooking their different temporal orientations. The US hypothetical negotiation conventionally reconstructs a bargain at the date infringement began; the European inquiry considers the parties’ actual negotiation conduct and prospective licensing relationship. A shared contractual metaphor does not erase the distinct remedial and competition law consequences of the two frameworks.

A cautious conclusion

The study’s descriptive material may be useful, but should not be used to conclude that comparables are superior, that top-down is inherently secondary, or that US and European frameworks have meaningfully converged.

Comparable licences can be relevant evidence, but only after transparent scrutiny of their economic content and negotiating context. Top-down is imperfect but provides a distinct check on aggregate royalty demands and standardisation value. In the US, the more immediate lesson from D-Link and CSIRO is that abstract claims of hold-up do not justify jury instructions, while SEP damages must still exclude value attributable to standardisation.