Skip to content

EPO guide

The problem-solution approach to inventive step at the EPO

By , Co-founder, PatentAssist.ai
Updated

Inventive step is the patentability requirement European applicants argue about most, and the EPO decides it with one method: the problem-solution approach. Examiners use it in search opinions, opposition divisions use it to revoke patents, and the Boards of Appeal have refined it for more than forty years. This guide explains each stage with the Guidelines and case law behind it, works through a full example, and lists the mistakes that lose arguments.

Quick answer

  • The law: an invention involves an inventive step if, having regard to the state of the art, it is not obvious to a person skilled in the art (Art. 56 EPC).
  • The method: (i) determine the closest prior art, (ii) establish the objective technical problem, (iii) decide whether the claimed solution would have been obvious to the skilled person (Guidelines G-VII 5).
  • The key test: would, not could. The prior art must have prompted the skilled person to make the change in the expectation of an advantage (T 2/83).
  • The trap: the problem must not contain a pointer to the solution (T 229/85), and only technical features count (T 641/00).

Where the approach comes from

Article 56 EPC is one sentence long. It says what inventive step is not (obvious) but not how to decide it. The early Boards of Appeal filled the gap. In T 1/80 (Bayer, carbonless copying paper, 1981) the Board held that the assessment must be preceded by determining the technical problem on objective criteria. In T 24/81 (BASF, metal refining, decided 13 October 1982, OJ EPO 1983, 133) it set out the core idea: start from the objectively relevant state of the art, determine the technical problem the invention solves from an objective point of view, and ask whether the solution was obvious to the skilled person. Looking for loose "indications" of inventiveness, the Board said, is no substitute for that analysis.

The Boards often anchor the method in Rule 42(1)(c) EPC, which requires the description to disclose the invention "in such terms that the technical problem, even if not expressly stated as such, and its solution can be understood". The approach is not strictly mandatory, but the Case Law book (I.D.2.2) notes that a department departing from it should say why, and in practice nobody at the EPO does. The Enlarged Board restated the stages in G 1/19 (point 26 of the Reasons).

The stages at a glance

The Guidelines (G-VII 5) describe three stages. The Case Law book splits the second into two (effects, then problem), which is how most practitioners actually work, so the table uses four.

The stages of the problem-solution approach
Stage The question Basis
1. Closest prior art Which single disclosure is the most promising starting point for getting to the invention? G-VII 5.1; T 606/89
2. Differences and their effect Which claimed features are missing from that document, and what technical effect do they produce? G-VII 5.2
3. Objective technical problem What task does that effect solve, stated without hinting at the solution? G-VII 5.2; T 229/85
4. Could-would Would the prior art as a whole have prompted the skilled person to make the change, in the expectation of that result? G-VII 5.3; T 2/83

Behind every stage stands the skilled person: a practitioner of average knowledge and ability in the field, aware of the common general knowledge, with access to the whole state of the art and the means for routine work and experimentation. In some fields this is a team rather than one person (G-VII 3). Prior European applications that count for novelty under Art. 54(3) are excluded from the inventive-step analysis altogether.

Stage 1: the closest prior art

The closest prior art is the single reference that discloses the most promising starting point for a development leading to the invention. The first filter is purpose: it should be directed to a similar purpose or effect, or at least belong to the same or a closely related field. Among the candidates that pass, the closest is normally the one with a similar use that needs the fewest structural and functional changes to reach the claim (G-VII 5.1, T 606/89). Shared purpose often weighs more than the raw count of shared features.

Three practical points follow from the Guidelines:

  • To refuse or revoke, one realistic starting point is enough. The examiner need not prove it is the closest; it only has to be a feasible springboard (T 967/97, T 21/08).
  • To grant, the claim may have to survive attacks from several equally valid starting points, but only if it has been convincingly shown that they really are equally valid. Opposition is not an open forum for unlimited attacks (T 320/15).
  • Prior art you acknowledge in your own description is treated as correct unless you say it was a mistake. Do not describe something as "known" casually.

Stage 2: the objective technical problem

Compare the claim with the closest prior art and list the distinguishing features, structural or functional. Then ask what technical effect those features produce in the claimed combination. The objective technical problem is the task of modifying or adapting the closest prior art to achieve that effect (G-VII 5.2).

This is rarely the problem stated in your application. You wrote the application against the art you knew; the examiner reformulates the problem against the art the search found. The reformulated problem can be less ambitious. If you claim an improvement but have no evidence of it over the closest prior art, the problem becomes "provide an alternative", and alternatives are often obvious (T 87/08).

Four rules decide most arguments at this stage:

  • No pointers. The problem must not contain part of the solution, because that guarantees hindsight (T 229/85).
  • Effects must be in the teaching. You can rely on an effect, even one shown with post-filed evidence, if the skilled person would see it as encompassed by the technical teaching and embodied by the same invention as originally filed. It need not be stated literally (G 2/21, Headnote II; T 116/18).
  • The whole claim must work. The problem counts as solved only if it is credible that substantially all claimed embodiments show the effect.
  • Partial problems are split. If different distinguishing features solve unrelated problems, each is assessed separately, and a different document can be combined for each (G-VII 6, T 389/86).

Stage 3: could or would

The final question is whether any teaching in the prior art as a whole would, not simply could, have prompted the skilled person facing the objective problem to modify the closest prior art so as to arrive at something within the claim (G-VII 5.3). The phrase comes from T 2/83 (Rider, simethicone tablet, 1984): the right question was not whether the skilled person could have added a barrier layer to a known tablet, but whether they would have done so in expectation of some improvement or advantage.

The prompt can be implicit (T 257/98, T 35/04). It can come from common general knowledge or a textbook, which are always obvious to combine with a document. And a chain of small steps is still obvious if each step follows from what has already been done (T 623/97). On the other side, needing two or more further documents to reach the claim can indicate an inventive step, and two documents that cannot work together because of inherently incompatible features are not normally combined (G-VII 6). Several Boards ask for "tangible evidence" of the prompt, not speculation (T 1014/07, Case Law book I.D.5).

Secondary indicators come in here as support: an unexpected technical effect, unless it is a mere bonus in a "one-way street"; a long-felt need; commercial success, but only together with a long-felt want and proof that the success comes from the technical features (G-VII 10.2 and 10.3).

A fully worked example

The example below is invented for illustration, but it follows the way an examiner's written opinion and a good reply are structured.

The claims

1. A drinking bottle comprising a container, a cap releasably attachable to the container, and a UV-C light source in the cap arranged to irradiate the contents of the container, characterised in that the cap comprises a sensor arranged to detect whether the cap is attached to the container, and a controller arranged to prevent the light source from operating when the sensor detects that the cap is not attached.

2. The bottle of claim 1, wherein the container comprises a level sensor and the controller is arranged to set the irradiation time according to the liquid level detected.

The prior art found in the search

  • D1: a bottle cap with a UV-C LED that sterilises the water for a fixed 60 seconds when the user presses a button.
  • D2: a countertop UV-C toothbrush steriliser whose lid carries a magnetic switch that cuts power to the lamp when the lid is opened, "so that the user's eyes and skin are not exposed".
  • D3: an industrial UV reactor for municipal water that adjusts the UV dose to the measured flow rate.

Claim 1, stage by stage

  1. Closest prior art. D1. It has the same purpose (sterilising drinking water in a portable bottle) and the same structure apart from the sensor and the interlock. D2 shares the interlock but has a different purpose; D3 is a different scale and field.
  2. Distinguishing features and effect. D1 lacks the attachment sensor and the controller that blocks the light source when the cap is off. The effect: the LED cannot shine when the cap is removed, for example when someone presses the button while looking into it.
  3. Objective problem. How to make the bottle of D1 safer for the user. A formulation such as "how to add an interlock to the cap of D1" would be wrong, because it names the solution.
  4. Could-would. Faced with a UV-C safety problem, the bottle designer would look at other consumer UV-C devices, a neighbouring field. D2 solves exactly this problem with an interlock and says why. The skilled person would therefore fit an attachment sensor and interlock to D1, arriving at claim 1. Claim 1 lacks an inventive step over D1 combined with D2.

Claim 2: where the argument becomes real

Claim 2 adds a level sensor and a level-dependent irradiation time. Its effect is that a half-full bottle gets a shorter cycle while a full bottle still receives an adequate dose, which saves battery charge without under-treating the water. Assuming the application as filed explains this, the effect is usable under G 2/21.

Note that the level sensor solves a problem unrelated to the interlock, so it is a partial problem: "how to reduce the energy used by D1 while ensuring sufficient disinfection". The examiner may cite D3, which also adapts dose to a measured quantity. The applicant's reply would argue the would-question on three grounds, each tied to the documents:

  • D3 is a remote field (continuous-flow municipal plant) and the bottle designer would only look there if prompted (G-VII 3, T 560/89). Nothing in D1 points to it.
  • D3 measures flow rate in a pipe, which has no counterpart in a closed bottle; adapting it needs more than routine work.
  • Neither D1 nor D3 is concerned with battery life, so neither gives the incentive that the objective problem requires.

Whether that succeeds depends on the evidence, but it is the right kind of argument: it accepts the examiner's framework and fights on the starting point, the problem and the prompt. The natural amendment is to combine claims 1 and 2, and to make sure the description supports the battery-saving effect. Drafting dependent claims with their own technical effects, so that such fallbacks exist, is covered in our guide on how to write patent claims.

Common mistakes, and what to do instead

Common mistakes in applying the problem-solution approach
Mistake Why it fails Instead
Writing the solution into the problem A problem such as "how to add a tilt sensor to the cap" makes any document with a tilt sensor look decisive. The Boards call this a pointer to the solution and reject it (T 229/85). State the problem by the effect: "how to stop the user being exposed to UV-C", not by the feature that achieves it.
Arguing that a different document is closer Once the examiner shows obviousness from one realistic starting point, it does not help to point to another document from which the claim would be inventive (G-VII 5.1; T 1742/12). Attack the chosen starting point itself: different purpose, different field, too many changes needed.
Relying on an effect the application never mentions A later effect counts only if the skilled person would see it as encompassed by the technical teaching and embodied by the same invention as filed (G 2/21, Headnote II). Describe the technical effect of each important feature in the application as filed.
Claiming an effect only part of the claim achieves A problem is solved only if it is credible that substantially all claimed embodiments show the effect (G-VII 5.2). Otherwise the problem shrinks to "provide an alternative". Narrow the claim to the range that works, or file data covering the whole range.
Treating an aggregation as a combination If features do not interact, each is a separate partial problem and can be met by a different document (G-VII 5.2, 6 and 7; T 389/86). Show the functional interaction and the combined effect, if there is one.
Counting non-technical features Features that do not contribute to technical character cannot support inventive step (COMVIK, T 641/00; G-VII 5.4). Tie each business or mathematical step to a technical effect in the claim.
Leaning on commercial success On its own it carries no weight. It helps only with a long-felt need and proof that the success comes from the technical features (G-VII 10.3). Use secondary indicators to support, never to replace, the could-would argument.

Mixed inventions and software

For claims that mix technical and non-technical features, such as most computer-implemented inventions, the approach is adapted under COMVIK (T 641/00, confirmed in G 1/19). Only features that contribute to the technical character of the invention can support an inventive step. A non-technical aim, such as a business requirement, can appear in the objective problem as a constraint given to the technical expert, which often makes the remaining technical implementation obvious (G-VII 5.4). Our guide to software and AI patents in Europe covers this in detail.

How it differs from US obviousness

US examiners apply 35 U.S.C. 103 through the factual inquiries of Graham v. John Deere Co., 383 U.S. 1 (1966), and the flexible approach of KSR International Co. v. Teleflex Inc., 550 U.S. 398 (2007), which rejected a rigid teaching-suggestion-motivation test (MPEP 2141). In practice US rejections combine references more freely, while the EPO's structure makes the reasoning more predictable and puts more weight on the effect you can show.

EPO inventive step compared with US obviousness
EPO USPTO
Legal test Art. 56 EPC: not obvious to a person skilled in the art, having regard to the state of the art 35 U.S.C. 103: not obvious to a person having ordinary skill in the art before the effective filing date
Framework Problem-solution approach, three stages (G-VII 5) Graham v. John Deere factual inquiries: scope and content of the prior art, differences, level of ordinary skill, plus objective indicia (MPEP 2141)
Starting point One closest prior art document, chosen for purpose and fewest modifications No single starting point; a primary reference is common, but any combination may be argued
Role of the problem Central: the objective technical problem is reformulated from the closest prior art One possible reason to combine among many; KSR rejected a rigid teaching-suggestion-motivation test
Standard for combining Would, not could: a prompt in the prior art, express or implicit Any articulated reasoning with rational underpinning, including "obvious to try" (MPEP 2141, KSR rationales A to G)
Non-technical features Ignored unless they contribute to technical character (COMVIK) Handled under 35 U.S.C. 101 eligibility, not in the 103 analysis

For how 103 rejections are answered in the US, see our guide to overcoming an obviousness rejection and the wider one on responding to a USPTO office action.

Drafting so the approach works for you

Inventive step at the EPO is won or lost largely in the application as filed, because effects and fallback positions cannot be added later (Art. 123(2) EPC). Before filing a European patent application:

  • Search first, so you can guess the likely closest prior art and describe your invention against it.
  • For each feature that might become the distinguishing feature, write down the technical effect it has and why.
  • State the problem broadly and in terms of effects, so it can be reformulated without losing support.
  • Include data or reasoning that makes the effect credible across the whole claimed range.
  • Build dependent claims that each solve a different partial problem, so there is always a fallback.

Find the closest prior art before the examiner does

PatentAssist.ai searches USPTO, EPO and Indian patent records, so you can see the likely closest prior art before you file. It also drafts claims and descriptions in EPO format, with a technical problem and advantageous effects section as Rule 42(1)(c) expects. It does not draft replies to EPO communications or argue inventive step for you; that remains the job of your European patent attorney.

Search EPO prior art free Try drafting free for 7 days

More on the search side: prior art search and European patent search. On drafting: AI patent drafting.

More EPO guides

All guides, by patent office

Problem-solution approach: frequently asked questions

What is the problem-solution approach?

It is the method the European Patent Office uses to decide whether a claim involves an inventive step under Article 56 EPC. The examiner picks the closest prior art, identifies the features of the claim that differ from it and their technical effect, formulates the objective technical problem that effect solves, and asks whether the prior art would have prompted the skilled person to solve that problem in the claimed way (Guidelines G-VII 5).

Is the problem-solution approach mandatory?

Not strictly. The Boards of Appeal have said it is not mandatory and that a department departing from it should explain why (Case Law book I.D.2.2). In practice examiners, opposition divisions and the Boards apply it in almost every case, so an argument that does not follow its structure is unlikely to persuade.

How is the closest prior art chosen?

It is the single disclosure that is the most promising starting point for a development leading to the invention. It should be directed to a similar purpose or effect, or at least belong to the same or a closely related field, and normally needs the fewest structural and functional changes to reach the claim (G-VII 5.1, T 606/89).

What does could-would mean?

The question is not whether the skilled person could have arrived at the invention by modifying the closest prior art, but whether they would have done so because the prior art prompted them, in the expectation of some improvement or advantage (G-VII 5.3, T 2/83). An implicit incentive is enough.

Can the technical problem differ from the one in my application?

Yes, and it usually does. The objective technical problem is based on the closest prior art found in the search, which the applicant may not have known about. It can be reformulated using any effect the skilled person would see as encompassed by the original teaching (G-VII 5.2, G 2/21). If no improvement is shown, the problem becomes providing an alternative.

How is the EPO approach different from US obviousness?

The EPO starts from one closest prior art document and asks whether the prior art would have prompted the skilled person to solve a defined technical problem in the claimed way. US examiners apply the Graham factual inquiries and, since KSR v. Teleflex (2007), may rely on any articulated reason to combine references, including obvious to try. The same claim can pass one test and fail the other.

Sources: Article 56 EPC and Rule 42 EPC; EPO Guidelines for Examination (April 2026), G-VII 1 to 12; Case Law of the Boards of Appeal, I.D.2 to I.D.5; Board of Appeal decisions T 1/80, T 24/81, T 2/83, T 229/85 and G 2/21; MPEP 2141. Other EPC provisions on patentability and procedure are annotated in our European Patent Convention reference. The worked example is a constructed illustration. This guide is general information, not legal advice. Ask a European patent attorney about your own application.