Most of what a deep-tech programme does well or badly comes down to one thing: how it judges the ventures in front of it. Get that wrong and everything downstream inherits the error. The wrong teams advance, and the strong ones walk away discouraged.
Judging deep tech is hard in a way that judging software is not. A SaaS startup shows traction in months; you look at usage, retention, revenue, and form a view. Deep tech gives you none of that. The science can take fifteen years or more to resolve. On average a deep-tech company reaches Series A about eighteen months later than a software one, and needs roughly half as much capital again before it earns anything. So at the moment you assess it, there is little to measure and a great deal to weigh.
The usual tool is the expert panel. Put experienced people in a room, have them read the applications, average their scores. It feels rigorous. The evidence says it mostly isn't. Across large studies of grant review, independent expert scores agree only weakly to moderately, and the field's own bar for “good” reliability is rarely met. The interesting part is why. It isn't that the reviewers are weak. The task is multidimensional and built on judgment, so two careful, qualified people read the same venture and reach different conclusions. Training helps, which points at something uncomfortable: even experienced reviewers often misjudge their own scale.
Then there is TRL. The technology readiness level, one to nine, came out of NASA in the 1970s, and it is now how most programmes and funders talk about maturity. Useful shorthand. Weaker than it looks. A founder self-reporting TRL has every reason to round up, and the scale lets them; “we have a prototype” can mean almost anything. A system is only as mature as its least mature part, so a venture can claim a high level on the strength of one component while the integration that actually matters sits years behind. TRL tells you how far the technology has come. It tells you nothing about whether the venture is any good.
TRL tells you how far the technology has come. It tells you nothing about whether the venture is any good.
None of this is an argument against expert judgment. It is an argument for putting a consistent layer underneath it. That is what we built. Every application is read and assessed the same way, against the same framework, with no fatigue and no drift between the first one and the four-hundredth. It does not take a TRL claim at face value; it weighs the claim against the evidence in the application and forms its own view. This is our own assessment methodology, not a model answering questions. The work is in how it scores, and that part is ours.
For The Innovation Races, that is the whole point. We run two races, one for ventures at TRL 4 to 6 and one for TRL 7 to 9, and both are judged the same way for everyone in them. The founder pitching from Riyadh and the founder in a university lab meet the same bar, on the same day, with the same care. At the scale these calls run, hundreds of applications at once, that consistency is not a nicety. It is the only honest way to do it.
The last piece is the one I care about most. Almost every programme a founder applies to gives them nothing back; they submit, they wait, they get a yes or a no. We give every applicant an honest view of where they are and where to focus next, whether they advance or not. For most of them, that is the most useful thing they take from the process. The assessment technology behind it is TACTIC, at tactic.science, and we will say more about it as we go.
We will get plenty wrong and fix it in the open. But the judging, the part that decides who gets a shot, we have taken very seriously from the start.
See how The Innovation Races assesses every venture against the same bar, or get in touch.