Most of the industry is still working around it.
There’s a paper Colin Mason and his collaborators published in 2017 that does something quietly devastating. They asked 238 UK angel investors why they’d declined deals. Seventy-one per cent of the rejections came down to one or two reasons. Across the two surveys, between fifty-nine and ninety per cent of the reasons were about people. Market, product and unit economics ran a long way behind. What carried the decision was character. Whether the founder seemed honest. Whether they seemed to know what they were doing. Whether their sense of their own valuation tracked reality. Whether the angel actually liked them.
Sit with that for a second. Most of the decision is happening in a register the pitch process barely exposes. The angel is making a character assessment, in something close to real time, from a deck and a coffee. The data they would actually need (twelve months of how this person behaves when the numbers go sideways) doesn’t exist in the room.
Mason has been documenting this, in one form or another, for thirty years. The corpus is substantial. The findings are consistent. The implementation has barely moved.
Five findings, five failure modes
Drawn from a body of work that runs from the late 1990s to a Henley Business Angels study published this year. Each has been visible long enough to have been responded to. Each is still load-bearing in 2026.
One. Investment readiness programmes deliver the cheap dimension and skip the expensive ones.
Mason and Kwok’s 2010 paper made the taxonomy explicit. There are three dimensions to investment readiness: equity aversion (the founder’s willingness to give up shares at all), investability (the substantive structural and analytical work that makes a business actually fundable), and presentational failings (the pitch and the deck). The first is cultural. The third is cheap. The middle one is where most of the value lives, and it requires expert delivery teams over months. The paper named eight critical success factors for getting it right. The binding constraint, in every case, was the quality of the delivery team and the duration of the engagement.
Their conclusion: programmes that deliver awareness and presentation but not diagnostic and business support are unlikely to be effective. That’s a direct quote from a UK government-funded study sixteen years ago.
What’s happened since. UK accelerators have multiplied. Most compress the engagement into ten or twelve weeks, focus heavily on Demo Day preparation, and rely on cohort peer learning to substitute for practitioner mentoring at scale. The investability work Mason and Kwok named as the load-bearing dimension is still mostly absent.
The Startup Race runs for fifty-two weeks. Every Startup Racer is operating Dave McClure’s Pirate Metrics and Ash Maurya’s Customer Factory on a working-backwards model from a public ARR target. Every week the model meets reality and the variance gets diagnosed: channel hypothesis wrong, early-adopter assumption wrong, value proposition wrong, pricing wrong. That’s the investability dimension, running as the operating substrate for the entire year.
The honest acknowledgment: the £100,000 Startup Race entry threshold is £5,000 ARR and a UK Limited Company with banked revenue. Pre-revenue founders are excluded by design. The Mason-Harrison five-element model, properly delivered, is the right intervention for them. The Startup Race solves the problem that begins at first verified pound.
Two. People factors dominate the decision, and the standard process barely exposes them.
Mason, Botelho and Zygmunt’s 2017 finding (the one I opened with) is the cleanest version of an observation that runs through the literature. The angel decision is dominated by a small number of high-weight people factors. Under conventional process, the substrate for evaluating those factors is one window: the founder pitching for forty minutes across a table. The statistical test in the paper found that only eight per cent of investor characteristics significantly predicted which rejection reasons mattered most. The angels are looking at roughly the same things. They’re trying to read them through a slit.
What an angel actually wants to know about a founder is: when the channel that was supposed to work doesn’t, do they name it or hide it? When the line-in-the-sand revenue goal isn’t being hit, do they update the model or rationalise the gap? When a major customer leaves, what do the next four weeks look like? These are behaviours visible only across time, under conditions of variance the founder can’t choose. None of them surfaces in a single sitting.
The Startup Race gives an angel fifty-two of those windows. The leaderboard updates weekly. Revenue is pulled from the founder’s bank account through open banking, so there’s no narrating around it. The week one founder loses a customer is visible alongside the next four weeks of how she handles it. The McClure-Maurya substrate produces a continuous record of hypothesis revision an angel can read after the race ends, or follow live across the year. The recorded angel Q&A in week 48 captures the founder under interrogation, in a form any angel who wasn’t watching can consume. Self-presentation in one window gets replaced with documented behaviour across fifty-two of them.
The honest acknowledgment: personal rapport (one of the four people-factor sub-themes Mason names) is harder to evaluate at distance than the other three. The IPP syndicate’s interaction with racers across the year handles some of it. It doesn’t handle all of it.
Three. Angels don't actually use the valuation tools the industry sells them.
Harrison and Mason’s 2017 paper should be on the wall of every early-stage advisory firm in London. They asked 127 UK angels what information sources they used to value an investment opportunity, on a five-point importance scale. The interview with the entrepreneur came out at 4.64 out of 5, marked as vital by seventy-five per cent of the sample. Third-party due diligence ranked fifteenth of twenty. Management projections beyond a single year ranked fourteenth. Backward-looking financial data was weighted more heavily than forward-looking projections.
The authors’ own conclusion is worth reading in their words: angels’ approach to valuation is not a function of the conventional protocols of financial analysis, but of personal relations and assessment. The Berkus method, the Payne Scorecard, Risk Factor Summation, Monte Carlo simulations, the new generation of AI-augmented platforms. None of them operates on a substrate the angels themselves substantially use. Twenty years of valuation methodology has been refining the wrong instrument. Every method on that arc was a workaround for missing data. Every method on that arc is still a workaround for missing data. The faster the telescope, the more obvious the fog. (I wrote a longer piece on that arc last month, from Berkus in 1996 through the AI tools of 2025.)
The Startup Race produces the thing angels were already valuing most heavily and rarely had access to: backward-looking verified revenue, twelve months of it, at the moment of first cheque. The valuation narrative milestone forces the racer to articulate the theory behind the number. The verified revenue tape is the data that theory has to fit. The Harrison-Mason finding that angels weight the founder interview at 4.64 out of 5 doesn’t get displaced. It gets supplemented with data that was previously absent.
The honest acknowledgment: absolute numbers at £5K to £100K ARR are small. The methodology can’t predict outliers and won’t suit companies whose value really does exist primarily as scientific conviction. Deep tech is a different problem. The methods Berkus invented are still the right tools for that problem.
Four. Due diligence rigour has not closed in eighteen years, despite syndicate professionalisation.
Wiltbank and Boeker’s 2007 Kauffman Foundation work on three thousand angel investments is one of the most-cited findings in the field. Angels who spent more than twenty hours on due diligence achieved deal-level returns 5.9 times higher than angels who spent less. The failure rate on the under-diligenced cohort was sixty-five per cent. The variable was hours of structured engagement before the cheque cleared.
That paper is eighteen years old. The UK syndicate market has, in the intervening period, professionalised. Cambridge Angels, Henley Business Angels, the Angel Investment Network, half a dozen others now run more structured screening processes than existed in 2007. The natural assumption is that the Wiltbank finding has been internalised by now.
Arundale and Mason’s 2025 paper on Henley Business Angels (one of the better-organised UK groups, by reputation) tests that assumption directly. Median due diligence time was five to ten hours per member. No external advisers were used. Information was taken on trust and not independently verified (their phrasing). Sixty-nine per cent of members relied on herd signals from previous investors who had committed earlier. The contemporary group runs at a quarter of Wiltbank’s threshold, eighteen years after the threshold was identified.
There’s a parallel solve worth flagging. Revenue-based financing has grown into a $9.8 billion global market on the same epistemological foundation this post is describing. Capchase, Wayflyer, Re-cap, Uncapped, Outfund. None of them runs pitch processes. All of them underwrite on verified revenue data pulled through open banking, with continuous monitoring across the lending period. The product is non-dilutive growth capital rather than equity, but the substrate is identical. The world has validated the verified-revenue substrate at substantial scale in adjacent territory. The Startup Race is what the equity version of that solve looks like.
The race takes a different route through the same gap. Twelve consecutive months of verified weekly revenue, McClure-Maurya hypothesis-revision artefacts, IPP syndicate interaction records, peer-observable behaviour across the cohort. That’s a year of evidence, generated before the cheque clears rather than alongside it. Wiltbank’s twenty-hour figure was about how long one angel spent producing private evidence for one deal. The race produces twelve months of public evidence for fifty deals. The evidence gets produced once and consumed by every syndicate member. The herd signal Arundale and Mason document gets replaced with direct observation that doesn’t require trust because it’s inspectable.
The honest acknowledgment: collective action across a syndicate is hard. Henley Business Angels’ co-ordination problems will exist in The Startup Race syndicate too. The Commitment Letter structure formalises the financial commitment in advance, which helps. Post-investment co-ordination (board representation, mentoring, follow-on decisions) still requires active management. The race doesn’t claim to have solved collective action. It addresses the substrate problem, which is the part angels can actually use data on.
Five. Gatekeeper screening is dominated by tacit personal experience, not articulable criteria.
Botelho and Mason’s 2024 paper used verbal protocol analysis on 21 gatekeepers across 19 UK angel groups, capturing 1,475 thought units during live deal screening. Only nine per cent of the thoughts related to the shared interests of the group members the gatekeeper was screening on behalf of. Sixty-two per cent of gatekeepers, when asked directly, didn’t acknowledge the collective nature of their role. Within the small subset of shared-interest thoughts, more than half were about fit with the group’s investor attributes. The gatekeeper was deciding what the members needed, not what they wanted.
This matters because the criteria gatekeepers use aren’t articulable. They emerge from the gatekeeper’s personal investment history and tacit pattern-matching. Two different gatekeepers screening the same deal would produce two different verdicts, and neither would be able to explain the difference in terms that the syndicate’s members could engage with. The investment opportunity gets filtered through a layer the members can’t see into.
This is the failure mode the race’s architecture sidesteps most completely. There is no gatekeeper screening. The £5K ARR entry threshold is objective; the race itself is objective (the leaderboard is the verified revenue tape); the prize allocation is objective (the winner is the founder who delivered the trajectory). The IPP syndicate’s role isn’t to screen founders into the race. Its role is to be the named prospective investor on the SEIS Advance Assurance application, to provide the post-race investment to the winner, and to provide infrastructure across the year. The gatekeeper-cognition failure mode Botelho and Mason describe doesn’t have a place to operate.
The honest acknowledgment: removing gatekeeper screening creates a different exposure. The race depends entirely on its objective entry criteria being well-calibrated. If the £5K ARR threshold is too low, the cohort is weak; if it’s too high, the cohort is small. Trading subjective gatekeeper screening for objective entry criteria solves one problem and creates another.
What the eight milestones do, given all of that
The Investment Readiness Constellation is a set of eight artefacts each racer produces over the year. The race is doing the people-factor evaluation across fifty-two weeks; the milestones do the work the observational substrate doesn’t.
Three are legal-readiness artefacts. A cap table reconciled with Companies House. Intellectual property formally assigned to the company. A founders’ agreement with vesting on a four-year schedule. These are the structural items Mason and Kwok named as the expensive dimension, the ones programmes routinely skip because they’re harder to deliver than pitch coaching. Three separate milestones (rather than one bundled requirement) signals that legal hygiene is the dimension racers most under-invest in unless held to it.
One is structural and race-facilitated: SEIS Advance Assurance, which gates more than ninety per cent of UK angel investment at this stage and requires a named prospective investor on the application. The IPP syndicate provides that.
One is the traction artefact: documented customer evidence following the traction hierarchy, recurring revenue at the top and letters of intent treated as worth almost nothing. This is the Harrison-Mason finding made operational. The artefact compresses fifty-two weeks of backward-looking data into a form an angel who wasn’t watching can read.
One is the analytical artefact: a defensible valuation narrative, with comparables and explicit assumptions. Not a number. A theory of the number, derived from the working-backwards model the racer has been running all year. The 2024 UKBAA finding that overvaluation is the most-cited founder mistake by thirty-one per cent of angels usually means founders unable to defend their valuation under pressure. The milestone forces the articulation in advance.
One is the costly-signal artefact: a weakness-disclosure document. The racer names their weaknesses in writing before due diligence finds them. Naming weaknesses disadvantages the racer’s negotiating position, which makes the disclosure credible in the way Michael Spence’s signalling theory describes. An independent reviewer certifies the document names at least one weakness that would actually cost the racer something to disclose, preventing decay into performative humility.
One is the stress-test artefact: a recorded Q&A with working UK angels who weren’t on the IPP syndicate. Any angel who watches the recording can see whether the founder folds, deflects, or holds. The race-watching angels don’t need this. The mid-race or post-race angels do.
The architectural framing matters more than the milestone list. The race is the people-factor evaluation. The eight milestones are the legible distillation of fifty-two weeks of underlying work, in forms that survive the year.
What's still genuinely uncertain
Kerr, Lerner and Schoar’s 2014 finding that angel funding’s main effect on outcomes is non-financial (mentoring, networks, governance) means a different selection mechanism, on its own, doesn’t capture what angel groups actually contribute. The race’s positioning (the IPP syndicate provides post-investment support alongside the capital) is consistent with that finding. The IPP syndicate has to actually deliver the support. The architecture allows it. It doesn’t guarantee it.
The Mason corpus documents the gaps and the empirical evidence that the gaps reduce returns. It doesn’t test a performance-gated first-cheque mechanism of this kind. The race’s proposition is consistent with the documented gaps without being validated by them. That validation is what the £100,000 race itself is designed to produce.
The Wiltbank IRR of twenty-seven per cent is an upper-bound number. Angel-group members tend to outperform solo angels, and the headline figure is probably at the top end of what’s achievable even with good due diligence. The race’s claim is narrower than the headline IRR: a tighter distribution around it. Fewer zeros, not necessarily higher peaks.
Why this hasn't been done already
The question comes up in every conversation about the race. If thirty years of research has been saying the same thing, why hasn’t someone built this before?
Three answers.
- The regulatory architecture favoured pitch-based models. Article 48 and 50A exemptions in the Financial Promotion Order were written around the historical syndicate category. Building a performance-gated structure required finding the legal pathway for an objective-criteria selection mechanism inside the existing framework. The architecture didn’t forbid it. It didn’t anticipate it.
- The data infrastructure didn’t exist until recently. Open banking went mainstream in 2018. The accounting integrations (Xero, QuickBooks) that make verified revenue available at first-cheque scale are roughly the same age. Pre-2018, what the race does would have required forensic accounting that was prohibitively expensive at first-cheque scale.
- The intellectual climate to make the case is recent. Mason’s most pointed papers post-date 2017. Behavioural economics (Kahneman, Sutherland) has only recently filtered into investment writing in a form that lets the substrate argument be made compellingly. McClure published Pirate Metrics in 2007; Maurya published Running Lean in 2010. The full conceptual stack the race draws on has only been available for about a decade.
Calling the race a brilliant insight overstates it. The race is the first coherent response to a problem the literature has been documenting more clearly each decade, built on infrastructure and intellectual scaffolding that have only recently converged. The people who saw the problem clearly long before there was a way to address it deserve the credit for the diagnosis. The data that lets the diagnosis be acted on is finally here.
James Shoemark is CEO of The Startup Race, a performance-gated angel investment platform headquartered in Edinburgh. The £100,000 Startup Race begins Summer 2026.
This article is general commentary on UK early-stage investment dynamics. It is not investment advice, a recommendation, or a financial promotion. Founders considering fundraising, and investors considering an angel commitment, should seek independent professional advice before acting on anything described here.