Proven Technology, Broken Pathway: The Structural Gap Holding Rail Back

Rail is more open than it was a decade ago. I want to say that plainly at the start, because everything that follows depends on it being true.

The sector now attracts proven companies from defence, aviation, highways, and technology. Firms arrive with real cross-sector track records, deployed systems, and references that would satisfy any buyer in their home market.

And rail still treats them as new entrants.

Same procurement hoops. Same assurance queues. Same distance between a successful trial and a contract. Our tenth anniversary review at the Rail Innovation Group names this directly, and it comes from member research rather than impression.

The maturity gap is real and it is documented

The report, Small Voices Big Impact, draws on member research from SMEs, scale-ups, and innovators working between 2016 and 2026. It finds clear cultural progress. It also finds that the progress has not been matched by structural change.

Members describe a system that remains difficult to enter and a persistent gap between successful trials and full adoption.

The industry has a name for that gap. Companies reach what is widely called the valley of death, where an idea completes its trial or demonstration and then fails to progress into operational deployment.

Here is the part I keep returning to. When a proven supplier stalls in rail, the reason is rarely a decision to reject them.

The deal is never rejected. It disperses.

Momentum leaks away across separate reviews, different stakeholders and parallel assurance conversations, and no single person is responsible for the point where it stops moving.

What other sectors settled long ago

Look at nuclear. Its Generic Design Assessment settles in advance which parts of a safety case are generic and which are site-specific. The split is a property of the framework. Nobody has to win that argument again on every scheme.

Look at aviation. Type certification attaches approval to the design itself, and a supplemental type certificate does the same for a modification. The approval belongs to the design and carries across installations rather than being earned again at each one.

Rail has analogous machinery. RISAS assesses a supplier once through an accredited body, so that duty holders and rolling stock owners can rely on a common assessment instead of each running their own. Network Rail's product acceptance route has a reduced path for products already proven elsewhere. Both grew up around physical products and rolling stock components.

Neither was built for a software, data, or an AI company.

For a software supplier the burden shows up as information assurance, security review, and integration approval. Each operator asks for it in a slightly different form. None of the answers count anywhere else. No accredited body issues anything another organisation will accept.

So the supplier proves the same thing again, and then again.

The first one doesn't travel as evidence. The burden of proof never reduces. It just resets.

Who actually pays for that repetition

The cost is not distributed evenly, and that is the point most often missed.

A large supplier treats duplicated assurance as overhead. It has the balance sheet to absorb the repetition and the staff to run it in parallel.

An SME pays in runway, which is the one resource it cannot replace. A cost that a large firm books as overhead can end a small one. The sector then filters for size, and the result gets described as a fair process.

If you are a founder reading this: budget for the repetition before you enter. The assurance you earned in one operator will not carry into the next. Plan your runway around proving the same claim several times over.

The problem moved. It did not go.

Cultural progress has not been matched by structural change. That single line from the review carries the whole diagnosis.

The sector is open, engaged, and actively pursuing innovation. Nobody now needs convincing that new ideas belong in rail. Concept to trial is achievable, and often achieved well.

Trial to adoption is where it breaks.

We won the argument about welcoming outsiders, and that success relocated the failure to a later stage. The handover from a proven trial to a real contract sits with no clear owner.

What a borrowed model would look like

The fix has a shape. I can describe it in three parts, and each part names something concrete.

Portable assurance for software and data. Evidence generated once should count more than once. An accredited finding that one operator accepts should carry weight with the next. This is the piece nuclear and aviation solved in their own domains.

A named owner for the trial-to-adoption transition. Someone with budget and authority whose job is the handover. Not a committee. A person with a line to spend and the standing to say yes. Rail already accepts this principle in a narrower setting. A product acceptance application at Network Rail has to be sponsored by a suitable person inside the organisation who is willing to act in that capacity. The principle is established. It does not apply to the decision that matters most.

A decision date registered before the trial starts. Not after it succeeds. The moment a trial begins, the date on which someone decides to buy or not should already exist on a record.

Some of this is already being addressed. In July 2026 GBRX published a preliminary market engagement notice setting out its intention to establish a dynamic market for rail innovation and technology. Suppliers would meet proportionate conditions of membership, join at any time and compete against defined problem statements. That is a serious answer to the access problem, and the design is still open to input.

The notice describes contract awards happening when funding and business need are confirmed. I understand why it is worded that way. It also describes the gap precisely. The route to market is being built well. The route to adoption sits behind a condition that nobody is yet accountable for satisfying.

I want to be careful here, because portability alone does not finish the job. An approval that travels removes a cost. Somebody still has to decide to buy the thing.

The convener and the matchmaker

There is a workaround that already keeps rail moving, and I know it well because I have stood inside it.

Rail is a closed industry. Getting an idea adopted usually depends on knowing the right people. Introductions are where those conversations start.

A trade body funded by members to represent them has to treat them equally. It can advocate. It cannot select. Membership buys visibility, access, and the chance to be considered. It does not buy an introduction.

Selection is what buyers actually want. Not a list of everyone. Three companies who could plausibly solve the problem, brought by someone who has thought about it before the meeting.

When that introduction works, it works because credibility stands in temporarily for the supplier's missing rail track record. That is a person accepting risk that no process is currently absorbing.

Neutrality is the load-bearing condition. If membership ever bought preference, the whole thing collapses.

Why I distrust the thing that works

When we asked members what they valued most about the Rail Innovation Group, credibility and introductions came before anything else. That is because the formal routes are hard to find.

Read that carefully. It describes a system with a gap in it. Nobody should mistake it for a compliment to the convener.

What I am describing is people using a relationship to do a job that a mechanism should be doing. A relationship reaches as far as one person's standing and lasts as long as they stay in the room.

A mechanism reaches further and outlasts the individuals involved.

So the honest ambition is uncomfortable. The aim is to make the role less necessary over time.

What this asks of you

If you sit in procurement or lead innovation for an operator, register the decision date before the next trial begins, and put one name against the handover.

If you shape policy, treat portable assurance for software and data as infrastructure worth building, because non-technical barriers around funding, procurement, and adoption are where proven firms currently stall.

If you are entering rail, plan for the wait. The welcome is genuine. Nobody yet owns the handover.

Rail no longer needs persuading to open its doors. What it has not settled is who says yes once a trial has done its work, and on what date. That is the next decade's problem.