Section 1 — The Bridge

What if the system isn't broken?

Most reports on regional innovation start from the assumption that something is missing — skills, funding, connectivity, ambition. The interventions follow: more programmes, more coordination, more strategy. This document starts from a different observation.

The system isn't failing to act. It's acting in ways that feel productive locally but prevent compounding regionally. Every actor is behaving rationally. The problem isn't what they're doing — it's what that behaviour substitutes for.

What if the reason your region's innovation ecosystem hasn't compounded isn't a lack of programmes, funding, or talent — but that the system has learned to stabilise around existing actors, existing pathways, and existing metrics? What if the coordination is the problem, not the solution?

This is where your experience becomes legible. Every steward has felt this: the meetings that produce alignment but not decisions, the programmes that scale activity but not throughput, the anchor institutions that absorb every new initiative into their existing orbit. The ones where the room nods, the strategy is published, the funding is allocated — and a year later, the same conversation is being had by the same people in a slightly different format.

What follows gives those experiences structural names. Not as critique. As description. Each name corresponds to a specific behavioural substitution — one class of action that has come to stand in for another. We have looked across one hundred and sixty-nine sub-clusters in twenty-five regions. The same configuration appears everywhere we have looked.

You will recognise it when you see it.

Section 2 — The Headline

169 clusters.
25 regions.
One pattern.

Across every UK region we have examined, three reinforcing behavioural configurations appear in roughly nine clusters out of every ten. They are not regional patterns. They are not sector patterns. They are how UK regional innovation ecosystems behave.

93.5%
of clusters confident

Coordination–Intermediary–Activity

Coordinating instead of deciding, Mediating instead of coupling, Scaling activity instead of throughput

92.9%
of clusters confident

Extraction–Intermediary

Extracting without reinvesting, Mediating instead of coupling, Stabilising around incumbents

90.5%
of clusters confident

Governance Capture

Coordinating instead of deciding, Narrating instead of testing, Scaling activity instead of throughput

Where this reading comes from

This is not drawn from surveys, interviews, or expert opinion. It is built from 16,317 structured evidence items drawn from four UK public data sources — UKRI grants and outcomes, DSIT Innovation Clusters sector data, Companies House registrations and dissolutions, and Crunchbase investment rounds — cross-referenced at entity level across 169 sub-clusters in 25 regions.

Every claim in this document traces to specific records. Every pattern traces through specific evidence items. The threads are visible.

One evidence thread, end to end
EVIDENCE PATTERN STALL STACK LEVERAGE UKRI grant record
248 grants, £260M, 41 institutions; two anchors hold ≥£345M of allocations.
Pattern P008
Institutional concentration of public research funding across multi-funder co-investment.
Stall
Stabilising around incumbents
HIGH confidence
Stack
Extraction–Intermediary
Reinforces with two other stalls
Leverage hypothesis
A non-anchor lead organisation winning a grant of comparable scale would test whether the substitution holds.

Every evidence item is logged with its source record, confidence tier, and the patterns and stalls it informs. The diagnostic refuses to assert what it cannot trace.

Section 3 — The Framework

How this works

Regional innovation ecosystems are complex adaptive systems — no central controller, actors responding to local incentives, system behaviour emerging from those local responses. When locally rational behaviours prevent the system from compounding value, we call them stalls. Not failures. Stabilisations.

A stall is a substitution — one class of action standing in for another. The substituting behaviour is observable; the substituted behaviour is what's missing. When stalls reinforce each other — coordination structures route activity through intermediaries that produce more coordination — they form stacks. Stacks resist change because removing one stall doesn't remove the conditions that sustain the others.

The nine stalls

Re-proving instead of narrowing

Repeated validation of capability instead of committing to one path.

Coordinating instead of deciding

Producing alignment and partnership instead of consequence-bearing decisions.

Forgiving instead of redesigning

Tolerating non-performance to preserve relationships instead of redesigning.

Extracting without reinvesting

Value (talent, IP, capital, credibility) flows out instead of recirculating.

Mediating instead of coupling

Routing every connection through intermediaries instead of direct coupling.

Stabilising around incumbents

Existing actors and structures preserved instead of opening emergence paths.

Narrating instead of testing

Producing reports and identity instead of testable behavioural change.

Scaling activity instead of throughput

Growing inputs (programmes, events, cohorts) instead of throughput and conversion.

Waiting for permission

Waiting for funding, policy clarity, or anchor permission instead of acting.

Where the framework reads, and where it doesn't

The framework reads the commercialisable layer of research and enterprise activity — the part that compounds through anchor-led pathways, public funding, company formation, and equity capital. It cross-references four public data sources at entity level, building evidence chains. Where the chain breaks — where a substitution cannot be observed in the available evidence — the diagnostic says so and downgrades its confidence. That discipline is the framework's value, not its limitation.

Section 4 — The National Pattern

What UK regional innovation ecosystems do

The stall fingerprint at national scale. Each bar shows the share of clusters where that behavioural substitution is confidently displaced — HIGH or MEDIUM evidence. The top of the chart is what UK regions reliably do; the bottom is what they reliably do not.

HIGH MEDIUM LOW INDETERMINATE
Stabilising around incumbents
95.9%
Coordinating instead of deciding
76.3%
Extracting without reinvesting
51.5%
Mediating instead of coupling
47.9%
Waiting for permission
46.2%
Scaling activity instead of throughput
45.6%
Re-proving instead of narrowing
36.1%
Narrating instead of testing
30.2%
Forgiving instead of redesigning
4.1%
Stabilising around incumbents fires in every cluster we examined. UK regional innovation ecosystems are universally favouring continuity over emergence. This is not a regional pattern. Not a sector pattern. A national pattern.

Restraint reading

Forgiving instead of redesigning is INDETERMINATE in 86% of clusters. The framework refuses to assert what it cannot observe. We do not see whether failed programmes are tolerated to preserve relationships because the public-source evidence base does not record programme terminations or accountability events at sufficient granularity. Where the dataset cannot speak, the diagnostic does not invent.

Three reinforcing configurations

93.5%
confident across clusters

Coordination–Intermediary–Activity

ComponentsCoordinating instead of deciding · Mediating instead of coupling · Scaling activity instead of throughput

Coordination structures route activity through intermediary organisations, which produce more activity in the form of coordination — a self-referential loop where producing meetings and partnerships counts as throughput.

92.9%
confident across clusters

Extraction–Intermediary

ComponentsExtracting without reinvesting · Mediating instead of coupling · Stabilising around incumbents

Anchor incumbents extract value through intermediated channels — the intermediaries justify the extraction by claiming brokerage, and the incumbents justify the intermediaries by directing flow through them.

90.5%
confident across clusters

Governance Capture

ComponentsCoordinating instead of deciding · Narrating instead of testing · Scaling activity instead of throughput

Coordination and reporting absorb the demand for action; the scale of activity (events, programmes, cohorts) makes the absence of consequence-bearing decisions hard to see. Each stall reinforces the legitimacy of the others.

The framework travels. Scottish devolved, Northern Irish post-conflict, English combined authority, Welsh city region — the same configurations appear. The stalls are not caused by governance structure. They are how innovation ecosystems organise themselves under pressure to be seen to act.

Section 5 — The Regions

Twenty-five ecosystem readings

The national pattern is built from twenty-five regional diagnostics. Each region's reading is its own document — with its sub-clusters, its stall fingerprint, its backbone configuration, its leverage surface, and the audit trail back to specific evidence rows. Four readings worth reading first:

All twenty-five readings, by political structure

England 17 regions

Thames Valley Strategic Mayoral Authority

11 clusters 1,209 evidence items
Distinctive stackPermission–Validation

If one high-growth company (Series A+) publicly committed to majority-local procurement and hiring without seeking incumbent partnership or UKRI grant support, …

Greater London Authority

8 clusters 1,049 evidence items
Distinctive stackPermission–Validation

If one UKRI funder published annual data separating value retention (spin-offs headquartered in London, IP owned by London entities, follow-on capital to London…

Cambridgeshire and Peterborough Combined Authority

9 clusters 918 evidence items
Distinctive stackPermission–Validation

If one documented case of direct coupling (e

Greater Manchester Combined Authority

8 clusters 915 evidence items
Distinctive stackPermission–Validation

If one successful outcome from the Manchester Academic Health Science Centre partnership (P001, P002, P004: direct partnership between UoM, Christie, MFT) were …

West of England Combined Authority

9 clusters 909 evidence items
Distinctive stackPermission–Validation

A probe could test whether one intermediary (e

West Midlands Combined Authority

8 clusters 875 evidence items
Distinctive stackPermission–Validation

If one hard exit criterion were applied to one programme category (e

Liverpool City Region Combined Authority

7 clusters 726 evidence items
Distinctive stackPermission–Validation

If the Combined Authority's annual Visitor Economy reporting (which currently cites £5

North East Combined Authority

7 clusters 664 evidence items
Distinctive stackPermission–Validation

If one university research centre (e

Hull and East Yorkshire Combined Authority

8 clusters 618 evidence items
Distinctive stackPermission–Validation

If one funding programme (e

East Midlands Combined County Authority

6 clusters 599 evidence items
Distinctive stackPermission–Validation

If one governance body (e

Tees Valley Combined Authority

6 clusters 581 evidence items
Distinctive stackPermission–Validation

A probe could test whether one anchor institution (e

West Yorkshire Combined Authority

6 clusters 557 evidence items
Distinctive stackPermission–Validation

If one university programme (e

Lancashire Combined County Authority

6 clusters 523 evidence items
Distinctive stackPermission–Validation

If one anchor institution (P021: Westinghouse, Springfields, Energy HQ, B&FC, Hillhouse TEZ) were to launch a programme without LEP/Combined Authority pre-appro…

Devon and Torbay Combined County Authority

7 clusters 512 evidence items
Distinctive stackPermission–Validation

If one anchor (Babcock, Princess Yachts, or Thales) validated a new capability by deploying it operationally without prior UKRI grant, Combined Authority approv…

Greater Lincolnshire Combined County Authority

6 clusters 510 evidence items
Distinctive stackPermission–Validation

A probe could test whether one government funding stream (e

York and North Yorkshire Combined Authority

6 clusters 407 evidence items
Distinctive stackPermission–Validation

If one documented case of direct bilateral partnership between a research institution and food manufacturer (e

South Yorkshire Mayoral Combined Authority

4 clusters 399 evidence items
Distinctive stackPermission–Validation

If one external funder (e

Section 6 — The Leverage

Where pressure might shift the configuration

Leverage in this framework is exposure, not power. A leverage hypothesis describes a single change — one actor, one decision, one reporting requirement — that would test whether the stack is sustained by the conditions we think it is. If the hypothesis holds and the configuration moves, the diagnostic is right. If it doesn't, the diagnostic was reading something else.

“A probe could test whether UKRI disaggregating the 248 follow-on funding events (P011) into "retained in region" vs. "extracted from region" categories in public reporting reduces the system's ability to absorb success and uncertainty signals through extraction-narrative coupling. If follow-on funding events were reported with geographic destination (e.g., "148 events: funding to Aberdeen-based entities; 100 events: funding to entities outside Aberdeen"), it might expose whether the success narrative (Narrating instead of testing) around follow-on funding (currently framed as ecosystem success) obscures extraction patt”

Aberdeen and Aberdeenshire · Aberdeen Digital · Extraction–Narrative · Low confidence

“If one accelerator (e.g., Level39, P004: 250 members, Google Cloud partnership 2024) launched a cohort focused on a specific regulatory challenge (e.g., Open Banking implementation, P005: established 2016) without coordinating with FCA Regulatory Sandbox (P005: launched 2016) or other accelerators (P009: 20 programmes), reporting outcomes to FCA after launch, it might reduce the system's ability to absorb demand and opportunity signals through re-proving and coordination cycles by demonstrating that validation can occur through doing rather than coordinating.”

Greater London Authority · London Financial Services FinTech · Permission–Validation · Low confidence

“If WECA allocated a portion of the £2.5M fintech innovation fund (e.g., £500k) via a decision process that excluded University of Bristol, Hargreaves Lansdown, and SETsquared from the selection panel, routing decisions through non-incumbent actors (e.g., recent fintech founders, international investors, non-regional experts), it might demonstrate that coordination infrastructure can operate through alternative nodes, reducing the system's ability to absorb disruption and opportunity through incumbent-mediated coordination.”

West of England Combined Authority · West of England FinTech · Coordination–Incumbent–Permission · Medium confidence

“If one non-university actor (e.g., Innovation Birmingham from P017, STEAMhouse from P004) launched a spin-out equity fund using locally-raised capital (not UKRI, not Midlands Engine Investment Fund) and publicly reported investment decisions without WMCA/LEP coordination approval, it might reduce the system's ability to absorb opportunity signals through permission-seeking by demonstrating that capital deployment can occur through non-incumbent nodes, potentially shifting the perceived necessity of incumbent gatekeeping.”

West Midlands Combined Authority · WMCA Digital · Extraction–Permission (Triple) · Low confidence

“If one documented case study were produced showing: (a) how a specific business-research collaboration (e.g., Cranswick's UKRI participation, P003) was initiated, (b) what role intermediaries played vs. direct contact, (c) what transaction costs were incurred, and (d) what outcome was achieved — it might reduce the system's ability to absorb complexity and uncertainty signals by shifting the burden of proof from "intermediation is necessary" to "intermediation adds value beyond direct coupling."”

Hull and East Yorkshire Combined Authority · HullEastYorkshire AgriFood · Intermediary–Narrative · MEDIUM confidence

Leverage types across the corpus

Type Hypotheses Mechanism
Information flow 111 Surfacing previously-unreported metrics that close the loop between activity and consequence.
Boundary adjustment 61 Designating one decision class that doesn't require coordination approval, exempting one path from the standard sequence.
Timing or sequencing 52 Launching using existing evidence as permission rather than seeking permission before launch.
Constraint shift 47 Removing an incumbent approval requirement for one grant category, one cohort, or one decision.
Coupling exposure 26 Routing one decision through a non-incumbent node, or one transaction through direct actor-to-actor connection.

Restraint reading — Belfast Screen

Eleven clusters across the twenty-five regions produced no leverage hypotheses despite identifying multiple confident stacks. Belfast Screen is the worked example. The framework distinguished between “the configuration is observable” and “a single change could test the configuration”, and held the latter back where the evidence didn't support a specific hypothesis. Restraint is the framework's most defensible feature.

Section 7 — Where the framework sharpens next

Four frontiers

Each frontier is a place the framework's reading gets sharper as better data, deeper engagement, or richer signal becomes available. Each is what a full diagnostic would add.

Demand-side signals

The Y-side enrichment now reads non-anchor-led grants, seed-to-Series-A conversion, dissolution among funded companies, and grant-to-equity composition. The next layer reads buyer behaviour, repeat demand, and adoption data — the part of the system where stalls either compound into stagnation or reveal that activity is actually converting to throughput.

Tolerance and redesign

Forgiving instead of redesigning is indeterminate in most regions because public sources don't record programme terminations, accountability events, or the conversations where decisions to continue or close get made. Steward knowledge changes that reading. A full regional diagnostic adds the closure events the framework can't see from outside.

Dark anchors

Some institutional actors are visible in the diagnostic data and some are not. The framework flags entities that appear in only one source as LOW confidence; it can identify candidates for "dark anchor" status — orgs that operate at scale in a region but don't show up in research grants, company registrations, or venture rounds. Bringing them into the reading requires steward-side attribution.

The 30-40% reading

The framework reads the commercialisable layer of innovation activity — roughly thirty to forty per cent of UKRI volume, the part that flows through anchor-led pathways and into measurable enterprise outcomes. The remaining sixty to seventy per cent — basic research, training, infrastructure maintenance — isn't absent. It's not in this reading because behavioural substitution at the commercialisation layer is what the diagnostic is for.

Section 8 — The Invitation

Two ways forward

This document is a national reading. Each region in it has its own ecosystem page with the full sub-cluster landscape, the leverage hypotheses, and the audit trail. If you steward a region, lead an anchor institution, or hold the brief for an innovation programme, two next steps: