Most companies treat funding as a single application: one programme, one deadline, one decision. Anyone running innovation systematically sees a different picture: a sequence of programmes along the stages of a project, bound together by cumulation rules, state aid limits and an evidence obligation that runs ten years beyond the decision. This guide describes the method behind a funding strategy and points to the deeper articles for the detailed cases. It is the only hub in this series that describes an approach rather than a programme.
Why a funding strategy?
Funding isn't an event, it's a process. The empirical evidence from academia shows this with remarkable consistency: since 2005 the Federal Demonstration Partnership has measured across several survey waves that researchers spend around 42 to 44 per cent of their research time on administrative tasks around funding[10]. The larger part of that effort sits not in the application but in the delivery afterwards: reporting duties, amendment requests, audit preparation. That asymmetry between roughly 20 per cent application and 80 per cent aftermath is the blind spot of conventional funding consultancy, which ties its fee to approval and ends with it. The article The 20/80 asymmetry in funding consultancy dissects the mechanism.
A funding strategy therefore means: thinking of programmes not individually but as a portfolio along the innovation stages; factoring cumulation and state aid limits in from the outset; and setting up the documentation from the start so that it survives a tax audit or an EU audit years later. The European Commission now puts the administrative burden in Horizon Europe at a median of 6 to 10 per cent of the project budget[11].
Mapping the stages
Innovation projects go through typical stages, and each stage has a different funding instrument attached. In the pre-incorporation and early phase, grants and venture capital subsidies apply: EXIST and INVEST. As soon as a company carries out R&D, the Forschungszulage comes in, which unlike grants has no deadline and remains available retroactively for every financial year. For demanding, market-close development, ZIM is the central grant instrument; for projects with a European dimension, Horizon Europe.
The mapping isn't a rigid scheme but a heuristic: which stage is the project at, how close to market, how international, how capital-intensive? We keep a separate guide for each of those axes: on the Forschungszulage, on ZIM, on Horizon Europe and on startup funding. The art of strategy lies not in choosing a programme but in sequencing and combining them over time.
Combinations and de minimis
Programmes can be combined, but not freely. The central limit is the GBER's ban on cumulation: the same eligible costs may not be funded twice from EU, federal or state money[2]. The Forschungszulage and ZIM can be combined as long as the same person-hour is only claimed once, and that separation has to happen at the level of cost type, not project name.
The second, often underestimated limit is the de minimis rule. Since Regulation (EU) 2023/2831 the ceiling is €300,000 per company over a rolling three-year period, raised from the previous €200,000 and now designed as a moving window that shifts with every new award[1]. What counts is the moment of award, meaning the funding decision, not the payment[3]. The group view is the tricky part: linked companies count as a single undertaking, so several small de minimis grants across corporate boundaries are added together. How quickly you can walk into that trap unnoticed is shown in the article The de minimis trap.
Ten years of compliance
The longest phase begins with the decision. There is no single retention period but a cumulative application of several rules, of which the longest governs. The basic grant law period under ANBest-P is five years after the statement of use is submitted[7]; the EU Financial Regulation requires five years after final payment under Article 132[6]. But tax and commercial law reach further: § 147 (3) AO and § 257 HGB provide for books, records and annual accounts to be kept for up to ten years[4][5]. In practice the longest applicable period governs, and that is usually ten years. The article Ten years of audit obligations after the funding decision sets out exactly how that cascade of deadlines is derived.
One modern compliance aspect concerns using AI in the application and reporting process. Anyone processing client or personal data covered by § 203 StGB — as a tax adviser or advisory service provider, say — may not use the public consumer version of an AI service unchecked; the German data protection conference considers that fundamentally impermissible under data protection law[8][9]. Its use only becomes permissible through informed consent, expressly bringing the provider within § 203 (3) StGB, an EU-hosted model with a data processing agreement, or anonymisation before input. The article ChatGPT in funding applications and § 203 StGB sets out the legal detail.
Infrastructure, not consultants
From all this follows the central proposition of this method: funding isn't a consultancy case, it's an infrastructure task. Conventional funding consultancy optimises the roughly 20 per cent of effort before approval and ends with the decision, exactly where the 80 per cent begins[10]. A funding strategy that addresses only the application leaves the larger part of the risk uncovered: cumulation errors that only surface in an audit; documentation gaps that lead to clawback years later; deadlines that pass unnoticed.
What an infrastructure does that a one-off consultancy cannot: it keeps the status of every application item traceable across the entire process, from delimiting the project through approval and drawdown to being audit-ready ten years later. It makes cumulation and de minimis usage visible at any time, instead of reconstructing them afterwards. That is exactly what upsmart is building the AI infrastructure for innovation management for: beyond consultants, not one application but the process across the whole term.
FAQ
How high is the de minimis limit? Since 2024 it has been €300,000 per company over a rolling three-year period; what counts is the moment of award, not the payment[1].
May I use ChatGPT for funding applications? For data covered by § 203 StGB the consumer version is fundamentally impermissible; its use becomes permissible through consent, bringing the provider within § 203 (3), EU hosting with a data processing agreement, or anonymisation[9].
How long do I have to keep funding records? The longest applicable period governs: five years under grant law, and up to ten years under tax and commercial law per § 147 AO and § 257 HGB[4].
Can I combine several funding programmes? Yes, as long as the same costs aren't funded twice and the de minimis limit is observed[2].
- [1]Commission Regulation (EU) 2023/2831 of 13 December 2023 on de minimis aidEUR-Lex · Official Journal of the EU · 2023Open source
- [2]Regulation (EU) No 651/2014 (GBER) — de minimis versus block exemption; the ban on cumulation for the same eligible costsEUR-Lex · Official Journal of the EU · 2014Open source
- [3]Förderdatenbank des Bundes: de minimis aid — national guidanceBMWK · Förderdatenbank · 2024Open source
- [4]Fiscal Code (AO), § 147 — rules on the retention of recordsFederal Ministry of Justice, gesetze-im-internet.de · 2024Open source
- [5]Commercial Code (HGB), § 257 — retention of records; retention periodsFederal Ministry of Justice, gesetze-im-internet.de · 2024Open source
- [6]Regulation (EU, Euratom) 2018/1046 (the Financial Regulation), Art. 132 — retention of recordsPublications Office of the EU, EUR-Lex · 2018Open source
- [7]General ancillary provisions for project funding grants (ANBest-P), no. 6Federal Ministry of Finance, Annex 2 to administrative provision no. 5.1 to § 44 BHO · 2019Open source
- [8]§ 203 StGB — violation of private secretsFederal Ministry of Justice, gesetze-im-internet.de · 2024Open source
- [9]Guidance from the German data protection conference: artificial intelligence and data protection (version 1.0, 6 May 2024)Conference of the independent data protection supervisory authorities (DSK) · 2024Open source
- [10]2018 Faculty Workload Survey — Primary Report (Federal Demonstration Partnership)Federal Demonstration Partnership (thefdp.org) · 2020Open source
- [11]Commission Staff Working Document — Interim Evaluation of the Horizon Europe Framework Programme (2021–2024), SWD(2025) 110 finalEuropean Commission, EUR-Lex · 2025Open source
- The de minimis trap: when one funding programme blocks the next
- ChatGPT in a funding application: why it breaches § 203 StGB
- The ten-year rule: audit obligations after the funding decision
- What the Federal Audit Office's KTF report really says
- The 20/80 asymmetry: why writing the application is the smallest job
- Human in the loop in the application process: what the Dell'Acqua study at BCG measured
- The reuse effect: how an application platform pays for itself from cycle 2–3
- Staff turnover in the innovation team: five questions about retained knowledge
The de minimis trap: when one funding programme blocks the next
The €300,000 threshold counts across three tax years, and across group boundaries. How to avoid application B devaluing application A.
ChatGPT in a funding application: why it breaches § 203 StGB
Tax advisers, lawyers and auditors risk criminal consequences when they use ChatGPT. What the DSK held in 2024.
The ten-year rule: audit obligations after the funding decision
In the European Court of Auditors' sample of audited research transactions, one in four shows measurable errors. Staff costs are the number one source of error, and the largest budget block.
The 20/80 asymmetry: why writing the application is the smallest job
The FDP faculty burden survey, the Commission's Horizon Europe interim evaluation and the auditors' error statistics show where the effort actually sits, and where conventional consultants structurally can't reach.
From analysis into the application.
We'll show you the platform on your actual case.