Eureka's Eurostars programme funds roughly 29% of the applications it receives. Innovate UK's Smart Grant, in its September 2024 round, funded 46 applications out of 1,645 — about 2.8%.
Both are competitive R&D funding for smaller companies. Both require a serious written application. One gives you roughly ten times the odds of the other.
Grant success rates vary this wildly across instruments, and the variation is almost entirely knowable in advance. That the choice matters is not news — forecasting which funder to target and hunting for the less-crowded instruments inside Horizon Europe are both well-trodden ground. What is missing is a table you can actually read across, and a second axis that almost nobody prices before committing: how long until the money arrives.
Grant success rates by instrument, with dates and sources
| Instrument | Success rate | Period | Notes |
|---|---|---|---|
| Eureka Eurostars | ~29% | 2024–25 | SME-led international consortium required |
| Innovate UK Smart Grant | 2.8% (46 of 1,645) | Sept 2024 round | Scheme later paused for evaluation |
| Horizon Europe (overall) | ~16% | first years, to May 2023 | Up from ~12% under Horizon 2020; more recent calls reported nearer 14% |
| ERC Starting Grant | ~12% | 2025 | Single PI, no consortium |
| NIH R01 | ~20% | recent years | Includes A1 resubmissions, which inflates the headline |
| EIC Accelerator | 3–7% | recent rounds | Multi-stage, heavy pitch component |
| Sapere Aude (DFF, Denmark) | 11% (38 of 336) | 2024 | Identical 11% for men and women |
| LF Experiment (Lundbeckfonden) | 12% (30 of 250) | 2024 | Short application; judged on the idea, not track record |
| Innobooster (Innovationsfonden) | Eligibility-gated | 2025 onwards | 35% max subsidy, DKK 200k–5m, not scored against a field |
Before drawing conclusions from that table, the necessary warning: these numbers are not apples to apples, and anyone presenting them as a straight ranking is misleading you.
The denominators differ. Some schemes report success against expressions of interest, others against full proposals; a two-stage programme always looks kinder at stage two. The eligibility gates differ enormously. And a 29% rate on an instrument that demands a three-country consortium with matched industrial partners is not "easier" than a 12% rate on a grant one person can write alone. The quantity you actually want is effort-adjusted odds, and no funder publishes it.
Which is why the table is a starting point for a decision, not the decision.
What actually drives the spread in grant success rates
Four mechanisms explain most of the variation, and telling them apart is what makes the numbers usable.
Gate effects. Innobooster is the clearest example. To apply you must have attracted at least DKK 100,000 in risk capital in the past three years, or shown DKK 250,000 in gross profit, and be an SME under 250 staff. That gate removes most of the population before anything is scored. A high apparent rate behind a narrow gate is not generosity — it is selection. The useful corollary is that you want instruments whose gate excludes people unlike you, because that is a filter working in your favour at zero cost.
Denominator and stage effects. A scheme that triages 80% of applicants at a light-touch first stage will report a flattering second-stage rate. Read which stage the published number refers to.
Budget rather than quality. In Horizon Europe's first two years, 71% of proposals that scored above the quality threshold went unfunded for budget reasons alone, with oversubscription running around 4.7×. In heavily oversubscribed programmes, the marginal decision is not about merit, and no amount of redrafting reaches it.
Fashion. Whatever is thematically hot draws the crowd. The instrument nobody has heard of, in a theme nobody is excited about, is where the odds live.
The axis nobody prices: time to money
Success rate tells you whether you win. It says nothing about when you get paid, and for anyone whose salary or a postdoc's contract depends on the answer, the second question is often the binding one.
Horizon Europe sets an official time-to-grant target of eight months from call deadline to grant signature. According to Science|Business reporting, nearly six in ten grants miss that target, and processing under Horizon Europe now runs roughly 23 days longer than it did under Horizon 2020. The direction of travel is the wrong way.
NIH R01 runs about nine to eleven months from submission to first-year funding, and ten to twelve months to the project start date.
Then there is the number that actually governs planning, and it is much larger than either. A realistic path for a first-time R01 applicant is 18 to 24 months from idea to first funding, because the sensible assumption is that an A1 resubmission will be needed. The headline NIH success rate of around 20% includes those resubmissions; first submissions do worse. Given how little a borderline rejection tells you about quality, planning for one round is planning to be surprised.
Put the two axes together and the arithmetic gets uncomfortable. An instrument with a 15% success rate and a twelve-month lag has an expected time to first euro measured in years, not months. Treating the submission date as the funding date is not optimism — it is a forecasting error with a number attached, and it is the reason projects die in the gap between winning money and being paid it.
Effort-adjusted odds: the number you have to compute yourself
Since no funder publishes it, here is the arithmetic. It is crude, and crude is enough to change decisions:
Expected value per hour = (success rate × grant size) ÷ hours you will personally invest
Two illustrations, with assumptions stated openly because the assumptions are doing most of the work:
| Success rate | Grant size | Your hours | Expected value | Per hour | |
|---|---|---|---|---|---|
| ERC Starting Grant | 12% | €1.5M | 200 | €180,000 | €900 |
| A national scheme | 30% | €150,000 | 60 | €45,000 | €750 |
| A heavily oversubscribed SME call | 3% | €500,000 | 120 | €15,000 | €125 |
Treat those rows as illustrative, not authoritative — plug in your own figures, because grant sizes and realistic hour counts vary enormously by field and by how much of the writing you personally do rather than delegate.
Two things fall out of the exercise that are not obvious from the success-rate column alone.
The prestigious low-odds grant often wins on expected value, because the prize is large enough to survive the division. A 12% shot at €1.5M beats a 30% shot at €150,000 on this measure, which is the opposite of what "pick better odds" advice implies. Odds arbitrage is a real strategy, but it is a strategy about effort, not about maximising money.
And the genuinely bad bets are the ones that are simultaneously low-odds and modest-prize. Those rows are where portfolios quietly bleed — a 3% call for half a million, entered because the deadline was there and the theme fit loosely.
One caveat that matters more than the arithmetic: expected value assumes you can run the bet many times. A researcher with one shot before a contract ends is not playing an expected-value game at all, and should weight probability and speed far above prize size. The formula is a tool for portfolio thinking, not a substitute for knowing which situation you are in.
Where to get the real numbers
Most of the figures circulating in conference corridors are stale, third-hand, or quietly refer to a different stage of a two-stage process. The primary sources are public and worth bookmarking:
- EU programmes — the European Commission's Horizon Dashboard publishes success rates by call and programme part, which is the only way to see how much a specific call diverges from the programme average.
- NIH — the NIH Data Book and RePORT give success rates by activity code and by institute, and the institute-level variation is much larger than most applicants assume.
- NSF — the annual Merit Review reports break rates down by directorate and division, which is where the enormous internal spread becomes visible.
- Denmark and the Nordics — Research Portal Denmark at
grants.forskningsportal.dklists individual awarded grants rather than rates, which is arguably more useful: you can see who won, for what, and how much. - National contact points — for EU calls, the NCP network and bodies like EuroCenter hold call-level intelligence that never reaches a published table.
The habit worth building is checking the rate for the specific call rather than the programme. A programme averaging 16% can contain calls at 40% and calls at 5%, and the average tells you nothing about which one you are about to spend 150 hours on.
What the ticket costs before you win anything
A typical Horizon Europe collaborative proposal consumes something like 100 to 150 hours of coordinator time before submission, and each partner contributes on a similar scale. That cost is incurred whether or not you win, and it is multiplied across the majority of high-quality proposals that fail on budget grounds.
I will not restate the institutional economics here, because the true cost of chasing grant funding deserves its own treatment and has one. The point relevant to instrument choice is narrower: an application is a priced bet, not a free lottery ticket, and the price varies by an order of magnitude across the rows in that table.
The discipline that follows is unglamorous and effective. Write the go/no-go decision down before partner agreements are signed — expected cost in hours, honest probability, and what else those hours would have bought. Do it in writing, because a verbal go/no-go always says go. The related question of whether your scope and budget fit the instrument at all is worth settling in the same sitting.
How to use this in practice
Four rules, in the order they should be applied.
Choose the instrument in week one. Not the narrative, not the title, not the figures. The single highest-leverage decision in non-dilutive funding happens before any prose exists, and most rejected applications fail at the choice of scheme rather than the quality of the writing.
Optimise against your actual constraint. If you have eighteen months of runway, time-to-grant dominates and a fast national scheme beats a prestigious slow one. If you need the line on your CV, the reverse. These give different answers, and picking without naming your constraint means letting the deadline calendar decide for you.
Prefer gates you clear and others don't. Every eligibility criterion you satisfy that most applicants do not is free competitive advantage. Career-stage-restricted schemes are the obvious case — which is why the timing question of whether to apply now or wait is worth real thought rather than defaulting to the earliest eligible round.
Date every number. Success rates move, sometimes sharply — Innovate UK's Smart Grant fell from 3–5% to 2.8% and was then paused entirely. A table without years is worse than no table, because it invites confident decisions on stale data. Everything above carries a period for exactly that reason, and it will need revisiting.
None of this makes a weak project fundable. What it does is stop a good project being spent on the wrong scheme, which is a more common failure and an entirely avoidable one. Grant success rates are public, the gates are published, and the timing is knowable. Spending the first week reading those three things — rather than drafting — is the cheapest improvement available anywhere in this process.