Frascati Manual: What Counts as R&D for a BSFZ Certificate

The Frascati Manual is the OECD's reference guide to measuring research and development, and Germany's research allowance borrows its idea of R&D from it. The Bescheinigungsstelle Forschungszulage (BSFZ), which certifies projects for the Forschungszulage, treats the manual as a supporting source. The binding test comes from German and EU law, and it uses three criteria instead of five. This guide shows how the two fit together and where applications fail.

What is the Frascati Manual?

The Frascati Manual is the OECD's methodology for collecting and reporting statistics on research and experimental development. It defines R&D as creative and systematic work that adds to the stock of knowledge, or that finds new applications for knowledge that already exists. Germany's certification body uses it as a supporting source when it decides what counts.

The Frascati R&D definition sits in paragraph 2.5 of the manual. It includes knowledge of humankind, culture and society alongside science and engineering, so the definition itself is broad. The criteria that follow it do the narrowing.

The Organisation for Economic Co-operation and Development (OECD) first published the manual in 1963, and it takes its name from the Italian town of Frascati. The 2015 edition is still the current one. The OECD has also published it in German as the Frascati-Handbuch.

The Frascati Manual is a statistical standard, and it does not decide tax claims by itself. Tax claims in Germany run through the research allowance, one of the instruments in our overview of government funding in Germany.

The five Frascati criteria, explained for engineers

An activity counts as R&D under the Frascati Manual only when it meets five criteria together. Paragraphs 2.7 and 2.8 list them and ask for all five to be met, at least in principle, every time R&D is carried out.

Novelty sets the target, and uncertainty shows the target was hard to reach. Germany's certification body reflects that balance. It scores novelty and risk directly and does not test creativity or transferability on their own, as the next section shows.

How the BSFZ turns five criteria into three tests

The Bescheinigungsstelle Forschungszulage (BSFZ) does not score the five Frascati criteria one by one. Its assessment guideline, the Prüfleitfaden (PDF, October 2025), sets three tests: novelty, technical risk and a methodical plan. The guideline names the Research Allowance Act (Forschungszulagengesetz, FZulG) and the EU General Block Exemption Regulation (AGVO) as decisive, and uses the Frascati Manual as a supporting source.

The BSFZ publishes its application volumes quarterly. According to the BSFZ, 26,042 companies applied for certification of 63,718 projects between 16 September 2020 and 30 June 2026.

The BSFZ's Prüfleitfaden lists only three criteria, so creativity and transferability are not scored separately. That is our reading of the text. A paragraph praising your team's creativity therefore adds little. Spend the space on novelty, risk and the plan.

Novelty: new to your company and to your sector

Following the Frascati Manual, the BSFZ tests novelty on two levels. The result must be new to your company, and it must not yet be state of the art (Stand der Technik) in your sector. A feature that is new to your product but standard across your industry fails the second test.

To separate R&D from routine work, the BSFZ looks at three parts of a project: the starting point, the path to a solution and the goal. At least one of them must differ enough from earlier work that further development under uncertainty is needed.

Technical risk: what the BSFZ accepts

The BSFZ counts only scientific and technical risk. Its test is whether the original solution approach might prove unworkable and force an alternative, extended or modified approach.

Write each risk the way you would brief another engineer. Name the approach, the point where it could break, and what you would try next.

Methodical plan: one defined task with clear objectives

Section 2(3) of the Research Allowance Act requires a precisely defined, indivisible task with clear objectives. The BSFZ links this to the Frascati requirement that R&D is planned and budgeted. For the full certification sequence, from the portal to the tax office, see our guide to the BSFZ research allowance for SMEs.

What is R&D, and where does routine work start?

Under German law, R&D is work that falls into at least one of three categories: basic research, industrial research or experimental development. Section 2(1) of the FZulG, the Research Allowance Act, takes these categories from the AGVO. Routine work begins once a product or process is essentially fixed and the aim shifts to market development or smooth production.

The Frascati Manual uses a close parallel in paragraph 2.9: basic research, applied research and experimental development. The BSFZ certificate does not say which category a project belongs to. Our FZulG guide to the R&D tax incentive explains the three categories in more detail.

The cut-off comes from section 2(2) of the FZulG. Once a product or process is essentially fixed, and the aim becomes market development or getting production to run smoothly, the work is no longer R&D. The BSFZ places the end of the R&D phase at the test of a prototype.

The BSFZ's Prüfleitfaden also names activities it will not certify, even inside an eligible project:

• feasibility studies, including "proof of concept" work that only prepares a go or no-go decision
• market research
• project management
• patent searches and patent filing work
• tests run only to generate data for approval, certification or standards

Software: the hardest boundary

The Frascati Manual sets a clear test for software projects: "its completion must be dependent on a scientific and/or technological advance". The OECD ran a public consultation on software and R&D measurement that closed on 31 March 2025, with a dedicated annex to the manual in view. We could not confirm that the final annex has been published.

The BSFZ draws the same line for software. Routine software tasks, porting existing software and configuring established tools do not qualify, and data-protection concerns never count as technical risk. For software teams weighing this against other money, our overview of startup financing in Germany shows where the allowance fits alongside equity and grants.

Why good projects still fail at the BSFZ

In our experience, BSFZ refusals usually trace back to the description rather than the project. The BSFZ can only certify what the text shows, and its guideline adds procedural rules that catch capable teams. Our BSFZ certificate guide explains who the BSFZ is and why the tax office is bound by its decision. These traps decide most Frascati-type questions:

• One criterion clearly missing. If novelty, risk or the plan is plainly absent, the BSFZ sends no follow-up request on the other criteria and refuses the project.
• English text. German is the official language of the procedure. The Prüfleitfaden excludes non-German content from assessment, so an English attachment is not read.
• Commercial risk presented as uncertainty. A market that might not buy is a business risk, and the BSFZ ignores it.
• Two projects in one description. The BSFZ must assess every criterion for each project, and a text that bundles two separable projects cannot be assessed.
• Market novelty. Being first on the market says nothing about the state of the art in your sector.

A real case: two projects certified, one rejected

An app startup building a place- and time-bound product filed three development strands for the research allowance. Two carried genuine technical risk: a data model for short-lived spatiotemporal states and an asynchronous orchestration layer across distributed systems. The third, an internal visualisation of existing data, had no open technical question.

We filed each strand as a separate project and flagged the third as weak before submission. The BSFZ certified the two risky strands and rejected the visualisation, as expected. The rejected strand did not reduce the allowance on the other two. The company received €725,000 on an eligible base just under €3 million, at the 25% rate that applied to those years. The full story is in our research allowance requirements case.

Does meeting the Frascati criteria guarantee R&D tax credit eligibility?

No. A BSFZ certificate confirming your project as R&D is the first condition, and the money depends on a separate second stage. The certificate binds the tax office on the R&D question. The tax office then decides costs, rate and payout.

According to the BSFZ, the allowance is 25% of eligible costs, or 35% for SMEs on activities after 27 March 2024. The BSFZ also states that for costs incurred after 31 December 2025, the assessment base is capped at €12m a year, and projects starting after that date add a 20% overhead flat rate. At 35% of €12m, an SME can receive up to €4.2m a year. Company size changes only the rate, and our page on funding for large companies explains how the allowance works above the SME threshold.

Earlier years stay open for a limited time. Our article on claiming the research allowance retroactively sets out which years and rates still apply.

For rates, caps and eligible costs across every period since 2020, see the Forschungszulage guide.

How to test your project against the criteria before you apply

Run each project through these questions before anyone writes the application. If one answer is vague, fix the project scope first.

• What was the state of the art when you started, and what could it not do?
• Which technical question was open at the start?
• Which approach could fail, and what would you try instead?
• Is the result new to your company and to your sector?
• Can you state one task with clear objectives and a work plan?
• Do the staff months and costs fit the work described?
• Have you removed feasibility, market, patent, certification and routine software work?
• Is the description written in German?

A project can meet all five Frascati criteria and still be refused if the BSFZ description does not show them. Our research allowance consulting team checks every project against the BSFZ's three tests before anything is filed. We separate eligible strands from routine work and draft the technical description with your engineers in a few short interviews. The fee is success-based, so nothing is due until the tax office has paid out the allowance. If a project will not pass, we tell you before you spend time on it.

This article explains general rules and is not legal, tax or financial advice. Check your own case with a qualified adviser before you apply.