The €157 Billion Problem: Why Research Waste Is Structural, Not Financial

Conversations about the state of European research almost always begin with the same question: is there enough money. It is a reasonable question, and it produces reasonable answers — larger grants, broader consortia, better instrumentation, longer funding horizons. It is also the wrong question to ask first, because it assumes the binding constraint is input. It is not. Roughly €157 billion of research value is lost every year, and very little of that loss is caused by work that was never funded. It is caused by work that was funded, performed, completed and then never reached anyone who could have used it.

That figure covers three things: experiments duplicated because nobody knew they had already been run, results never published because they were negative, and the absence of infrastructure that would let researchers collaborate on work while it is still in progress rather than after it is finished. None of those are failures of scientific rigour. They are failures of plumbing. The distinction matters, because underfunding and misrouting call for entirely different responses — and the sector has spent two decades applying the remedy for the first problem to a system suffering mainly from the second.

A community of 8.8 million, organised as 21,000 islands

There are approximately 8.8 million active researchers worldwide, distributed across more than 21,000 institutions. On paper this is the largest and most interconnected knowledge-producing community in human history. In practice, the unit of coordination is still the individual laboratory, and the mechanism by which one laboratory learns what another is doing is still the peer-reviewed journal article — a document that appears, on average, twelve to thirty-six months after the work described in it was actually carried out.

Consider what that embargo window means operationally. For between one and three years, the existence of a completed experiment is known only to the group that ran it, their funder, and possibly a handful of conference attendees who saw a partial poster. Every other group in the world is, during that period, working with a picture of the field that is by definition out of date. They are not being careless. They are being structurally denied the information that would let them allocate their own resources sensibly.

The result is not a rounding error. It is the dominant inefficiency in the modern research economy, and it decomposes into three fairly distinct categories of loss.

Category one: parallel duplication

The most common form of waste is also the least visible, because nobody involved ever finds out it happened. Two groups, in different countries, working from the same published literature and the same funder priorities, independently identify the same promising avenue. Both design a baseline trial. Both spend in the region of €150,000 running it. Neither has any way of knowing the other exists, because there is no register of work in progress — only a register of work completed and cleared for publication.

Eventually one of them publishes. The other group discovers, some months into their own analysis, that their contribution has been pre-empted. Their data is not wrong. It is simply no longer novel, which in the current incentive system is functionally the same as being worthless. The €150,000 is written off, the postdoctoral years spent on it produce nothing publishable, and the field gains one paper where it paid for two.

Notice what would have prevented this. Not more funding. Not better peer review. A single verifiable signal, visible early, saying: this line of enquiry is occupied, at this level of maturity, by a group willing to be contacted. That signal does not exist as shared infrastructure, and so it does not exist at all.

Lifecycle state	Traditional destiny	What infrastructure changes
Successful, unpublished	Held in secrecy for 12–36 months	Timestamped proof allows disclosure without loss of priority
Active but blocked	Reagents and labour burned on one problem	The problem is posted; an outside specialist can solve it
Parallel duplication	The slower team is scooped; budget written off	Matching flags the collision before either team loses the work
Null or inconclusive	Filed on local servers, repeated elsewhere	Listed as licensable data, raw datasets encrypted
Abandoned programme	Written off; notebooks archived	Packaged and transferred as an asset with its full history

Category two: the negative result that never leaves the building

The second category is more widely acknowledged and no better solved. A well-designed clinical trial costing around €250,000 returns a negative primary endpoint. The compound does not work, or the intervention shows no effect against control. This is a genuine scientific finding — arguably a more decisive one than a marginal positive — and it has a clear economic value to every other group considering the same hypothesis.

It is also, in practice, almost unpublishable. Journals select for novelty and positive effect. Institutional reputation management discourages advertising failures. Principal investigators facing a finite writing budget rationally prioritise the manuscripts most likely to be accepted. So the trial is filed, the dataset sits on institutional storage, and within a few years another group runs a materially identical study and reaches the same negative conclusion at the same cost. The cycle can repeat several times before the null result becomes tacit knowledge in the field, transmitted by conversation at conferences rather than by any formal record.

The waste is not the product of scientists behaving badly. It is the product of a system in which the rational individual choice — wait, protect, publish later, publish only what will be accepted — aggregates into a collective loss on the order of €157 billion a year.

Category three: the stranded asset

The third category sits mostly inside industry, and it is the one research offices and technology-transfer staff will recognise immediately. A pre-clinical programme accumulates perhaps €500,000 of characterisation work, assay development and animal data. Then the sponsoring organisation changes strategic direction — a portfolio review, a merger, a shift in therapeutic focus — and the programme is discontinued. The science was not invalidated. It was simply no longer aligned with the owner’s priorities.

What happens next is almost always nothing. The asset is written down. The data stays on an internal server under an indefinite embargo, retained for defensive reasons and used for none. Somewhere else, an academic group or a smaller company would find that same package genuinely valuable, and would pay for it. But there is no market, because a market requires three things this one lacks: a way to describe an asset without disclosing it, a way to prove who owns it, and a way to transact without a bespoke six-month legal negotiation for every deal.

Why the obvious fixes have not closed the gap

None of this is newly observed. The sector has responded with trial registries, preprint servers, open-access mandates, data-management plans and FAIR principles. Each has done real good, and none has touched the underlying mechanics, for a consistent reason: they all operate on work that the researcher has already decided to disclose.

The disclosure decision is where the loss originates. A researcher who cannot prove priority will not share early, because sharing early without proof means risking the only currency the career system recognises. A pharmaceutical company will not describe a shelved asset in a public forum without a defensible ownership record and a controlled disclosure path. Asking people to be more open, without first giving them the means to be safely open, is asking them to absorb an individual risk in exchange for a collective benefit. Most will decline, and they are not wrong to.

What infrastructure would actually have to do

If the problem is structural, the fix has to be structural too. Working backwards from the three categories, a piece of infrastructure capable of closing the gap would need to satisfy a fairly demanding and quite specific set of requirements.

• Establish priority at the moment work is done, not at the moment it is published — so that early disclosure carries no career cost.
• Produce proof that is verifiable by a third party without trusting the institution that generated it, since the whole point is to coordinate between organisations that have no reason to trust each other.
• Keep the underlying data confidential, because no laboratory director will place patient data, unpublished protocols or commercially sensitive results into a public system.
• Make the terms of a collaboration enforceable and legible in advance — contribution shares, attribution, licensing conditions — rather than renegotiated by counsel for every partnership.
• Cost little enough per record that timestamping a routine milestone is an unremarkable operational decision rather than a budget line requiring justification.

That last requirement is stricter than it sounds, and it is where most proposed solutions quietly fail. Infrastructure only changes behaviour when using it is cheaper than thinking about whether to use it.

The uncomfortable part of the answer

The second requirement is the one that rules out the conventional approach. A registry run by a university, a publisher or a national funder is only as credible as that organisation, and its records are only as durable as its budget. For coordination between competing institutions across jurisdictions, an authority that any participant could plausibly influence is not sufficient.

This is where distributed ledgers become relevant, and it is worth being precise about how little is actually being claimed. The mechanism is narrow. A cryptographic hash is a fixed-length fingerprint of a file: the same input always yields the same fingerprint, any alteration yields a completely different one, and the fingerprint reveals nothing about the contents. Publishing that fingerprint to a public ledger creates a record that a specific dataset existed, in a specific state, at a specific moment — a record no single party can revise afterwards, including whoever operates the platform. Verification is arithmetic. It does not require anyone’s assurance.

That is the entire contribution. It does not make results true, replace peer review, or say anything about scientific quality. It removes one specific obstacle — the inability to prove priority without publishing — and that obstacle happens to sit directly upstream of a very large amount of the waste described above.

Why the Netherlands is a sensible place to start

Infrastructure of this kind succeeds or fails on density rather than reach. The Netherlands offers an unusual concentration: 31 globally ranked universities, the cluster at Leiden Bio Science Park, the European Medicines Agency, and a research culture that has already internalised FAIR data principles at policy level. It is a jurisdiction where the argument for machine-readable, verifiable research provenance does not need to be made from first principles, and where academic groups, contract research organisations and regulators sit close enough together for a shared standard to become useful quickly.

Where this stands today

DimenChain operates this layer as live infrastructure for research institutions, laboratories and R&D groups. Datasets, protocols and project milestones are hashed and timestamped on Arbitrum, an Ethereum Layer-2 network whose transaction costs keep per-record notarisation compatible with ordinary academic budgets. Raw data never leaves the researcher’s control: only encrypted metadata, hashed proofs of ownership and smart-contract events are written on-chain. Collaboration terms and licensing of unpublished or discontinued work are governed by contract logic rather than by correspondence. The underlying protocol is patent-protected under WIPO PCT/IB2024/058791.

None of this makes anyone’s science better. It makes the surrounding system less lossy — which, given where the €157 billion actually goes, is the more consequential of the two.