How to Identify Spending Intent After a Grant: Reading the Signals Correctly

A signed grant is not just money received — it is money committed to be spent. The beneficiary has agreed, in a legal document with a budget annex and a work plan, to incur specific categories of cost within a specific period, subject to audit. That makes public grant data one of the few places where future B2B purchasing is visible before it happens.

But the signal is only as good as your reading of it. Grant records tell you some things with certainty, other things with useful probability, and some things not at all. This guide walks through the funding-to-spending logic for the EU's flagship programme, Horizon Europe, and shows — with explicitly labeled examples — where fact ends and inference begins.

What a grant budget actually commits to

Every Horizon Europe grant reimburses eligible costs, defined in the Model Grant Agreement and explained in the Commission's Annotated Grant Agreement (AGA, v2.0, 01.04.2025). The budget is structured into standard categories:

• A. Personnel costs — employees, seconded staff, SME-owner day rates. Typically the largest line.
• B. Subcontracting costs — parts of the project work carried out by third parties against payment.
• C. Purchase costs, split into:
  • C.1 Travel and subsistence — trips, accommodation, conference fees;
  • C.2 Equipment — instruments, machinery, IT hardware. Note the rule: where equipment outlives the project, only depreciation for the project period is eligible, not the full purchase price;
  • C.3 Other goods, works and services — consumables, cloud and software, testing and certification, dissemination, IP costs.
• D. Other cost categories — e.g. financial support to third parties, internally invoiced goods and services.
• E. Indirect costs — a flat 25% of eligible direct costs (excluding subcontracting and a few other items), covering overheads with no receipts required (see also the Austrian funding agency FFG's summary of Horizon Europe eligible costs).

For a supplier, the mapping is direct. Personnel budget implies hiring (a signal for recruiters and HR tooling). Subcontracting budget implies outsourced work packages — engineering services, clinical work, specialised development. Equipment budget implies capital purchases. "Other goods and services" is where cloud infrastructure, lab consumables, software licences and certification services live.

Two structural facts sharpen the equipment signal. First, the depreciation rule creates a timing incentive: since only depreciation during the project is eligible, equipment bought late in a project recovers less of its cost — so beneficiaries who need equipment tend to buy it early. Second, purchases and subcontracts must be awarded to offers providing best value for money, which in practice means beneficiaries collect competing quotes — a supplier who shows up while quotes are being gathered is in the deal; one who arrives after the audit file is closed is not. (Both rules: AGA, Articles 6.2.C and 6.2.B.)

The timeline: when funding turns into purchasing

Horizon Europe operates under a service standard the Commission calls time-to-grant: applicants are informed of the outcome within five months of the call deadline, and grant agreements are signed within eight months (Commission GAP presentation; see also the University of Cambridge's grant-preparation guide).

A realistic sequence for a project with an equipment or infrastructure component:

1. Call deadline → month 5: evaluation results. (Winners often know before any database does.)
2. Month 5–8: grant agreement preparation and signature. The budget is now fixed.
3. Project start (the start date is published in CORDIS): personnel costs begin immediately; recruitment often starts here.
4. First months of the project: equipment and infrastructure purchases concentrate here, driven by the depreciation rule and by the fact that later work packages depend on the equipment.
5. Mid-project: subcontracted work packages, trials, testing and certification services.
6. Reporting periods: each reporting cycle re-opens budget discussions; underspent lines get reallocated.

Meanwhile, the publication of the grant lags the signature: CORDIS bulk data refreshes roughly monthly, and the Financial Transparency System publishes a financial year's commitments only in the following year. Practical consequence: by the time a grant is visible in public data, the beneficiary is typically at step 3–4 — exactly the window where purchasing decisions are being made but not yet all closed. The lag hurts less than it seems, but it is real, and any vendor of funding intelligence should tell you so.

Reading a CORDIS record: fact versus inference

CORDIS publishes, for every Horizon project: total cost, EU contribution, all participants with per-participant contributions, start and end dates, the topic, and an objective — the project's own summary of what it will do. The objective text is the closest public proxy to the work plan (full work packages are not published).

Here is a real record, and what can and cannot be concluded from it.

Example 1 — a real record that should not trigger a sales alert. CORDIS project 101069750, the Clean Energy Transition Partnership (CETP):

• Fact: total cost EUR 561,150,162.21; EU contribution EUR 136,317,392.84; coordinated by an Austrian federal ministry; runs 2022–2030. (Source: the CORDIS fact sheet.)
• Fact: the participants are predominantly national ministries and funding agencies.
• Correct inference: this is a co-funded partnership — a vehicle through which public funders pool money and issue their own calls. The half-billion euro figure is not one organisation's spending budget.
• Incorrect inference (the false positive to avoid): "an organisation just received EUR 136 million and will now buy equipment." The purchasing signal here is downstream, in the projects this partnership later funds — not in the record itself.

The lesson: entity type matters. A ministry, a university consortium, and a 40-person deep-tech company with the same grant size imply completely different purchasing behaviour.

Example 2 — the typical single-beneficiary pattern (illustrative, not a real project). Suppose a CORDIS record shows a German battery-materials SME as sole beneficiary of a EUR 2.4 million EIC Accelerator grant; the objective states the company will "scale production of its cathode coating process from laboratory to pilot line and validate output with automotive customers."

• Fact: the company received EUR 2.4 million; the project runs 24 months from the published start date; the objective names a pilot line and validation.
• High-confidence inference: a pilot line implies process equipment, instrumentation and facility work in the first project year — the depreciation rule and the work-plan logic both push purchases early.
• Medium-confidence inference: "validate with automotive customers" implies external testing/certification services and travel — category C.3 and C.1 spending.
• Low-confidence inference: which specific vendor category wins (new machinery vs. retrofitting existing lines; buying vs. leasing) — the record does not say, and only a conversation can resolve it.
• Not inferable: budget split across categories (the budget annex is not public), supplier shortlists, or whether the purchase already happened before you read the record.

This labeling discipline — fact, then inference with a stated confidence, then explicit unknowns — is the difference between a signal and a guess. Any funding-intelligence workflow, manual or tooled, should preserve it.

Failure modes to respect

Honest use of this method means knowing when it breaks:

• Consortium dynamics. In multi-partner projects, equipment may be bought by one partner and used by all; the per-participant contribution in CORDIS tells you who holds the budget, not who signs the purchase order.
• Existing capacity. Organisations with the equipment already in place budget depreciation of existing assets, generating no new purchase.
• Timing misses. Some purchasing closes before public data surfaces the grant; a fraction of signals will always arrive late.
• Grant amendments. Budgets shift between categories mid-project; the original signal decays.

None of these invalidate the approach — they set its ceiling. Grant data yields prioritisation, not certainty: a list of accounts where the probability of near-term, category-specific purchasing is measurably elevated, each traceable to a public record you can cite in the first email.

The method in five lines

1. Detect the funding event (CORDIS monthly data, national registers, FTS annually).
2. Classify the beneficiary (company vs. institute vs. agency — Example 1 above is why).
3. Read the objective for purchasable nouns: pilot line, platform, trial, deployment, infrastructure.
4. Map them to AGA cost categories and to the project timeline from the published start date.
5. Label every conclusion as fact or inference, with confidence — and let the sales conversation do the rest.

SubsidySignal automates steps 1–4 across funding programmes and keeps step 5's labels attached to every signal we surface. But the method itself requires nothing more than the public records linked above — and if a signal can't be traced back to one, it isn't evidence.

---

Sources: EU Grants Annotated Grant Agreement v2.0, 01.04.2025 · CORDIS and project 101069750 · Commission grant agreement preparation (GAP) guidance · FFG — eligible costs in Horizon Europe · Financial Transparency System.