Agricultural Knowledge Networks
Learn from every field, not just your own.
Farm evidence that lasts beyond the project that produced it, travels with its provenance and permissions, and stays under the control of the people who supplied it. VT3 helps research networks, funders and farmer organisations design and run the trusted data networks that make this possible.
Everyone agrees. Almost nobody is doing it.
Governments, funders and research bodies agree on the goal. Defra’s farming plans, the EU’s agricultural data spaces, the shift to FAIR research data, and new global work on portable, person-governed “memory” for AI all point the same way: farm knowledge should be shared, trusted and owned by those who produce it.
The reality hasn’t caught up. Trial results end up in a spreadsheet or a final report, and the next project starts from scratch. Evidence about a farmer stays locked inside whichever organisation or platform recorded it. Consent was given for one project, not for reuse. And the models now being trained on farm data often can’t tell where it came from, how current it is, or whether they’re allowed to use it.
The UK has already tried this the expensive way.
In 2015 the government launched Agrimetrics with £11.8m of public funding, as a flagship of its £140m Agri-Tech Strategy, to build a national farm-data platform. Despite tens of millions in public investment, it never became the national platform it was meant to be. The company pivoted to consultancy and was wound up in January 2026.
It wasn’t short of money, talent or technology. It started with the platform, not the people, and a centralised organisation is hard for farmers to trust with their data. A federated approach turns that around: every stakeholder’s rights are respected, and data keeps its provenance and stays with those who own it.
Sources: London Gazette notice 5045968 (MVL, 22 January 2026); AgTechNavigator, 9 January 2026.
| The centralised approach | The federated approach |
|---|---|
| Build the platform first | Agree governance and membership first |
| Pool data in one place | Data stays with whoever supplied it |
| One organisation owns the asset | Members govern it, and no one can capture it |
| Trust is assumed | Trust is proved: provenance and permissions travel with the data |
| Funded as a project | Designed from the start to outlive its funding |
| Technology-led: build it and they’ll come | Technology shaped by the community and its rules |
Farm memory that farmers own
Evidence stays with the farmer who produced it, and can be shared on the farmer’s terms, as often as it’s needed.
The farmer’s data
Verified records the farmer holds: trial results, practices, training, benchmarks, audits. They can share exactly what a buyer, lender, adviser or scheme needs, and nothing more, without re-proving it every time.
The network’s knowledge
Evidence that members choose to pool, under rules they set together, so researchers can compare across farms, seasons and regions, and every member can learn from results they could never have produced alone.
Context travels with the data
Both depend on it. Every item carries where it came from, when and how it was recorded, and whether it was verified, reported by the farmer, observed by a device, or inferred by a model. It also carries what it may be used for, for how long, and by whom.
Find it, trust it, ask to use it
Data nobody can find might as well not exist.
Find it
In an agricultural knowledge network, every dataset can publish a short description of itself: what it covers, where and when it was collected, which standard it follows, and who to ask for access. These descriptions can be indexed and searched across the network, so anyone can see what evidence exists while the data itself stays with its owner. Owners decide whether each description is public, visible to members only, or private. Discovery is provided by Murmurations, an open protocol for publishing and finding information without a central database, created by our team at The Open Co-op.
Trust it
Each description carries its provenance: who recorded the data, with what method or device, and whether it has been checked against the agreed standard. The authority behind it travels too. A signature shows who issued a piece of evidence, but not why they were allowed to. In a verifiable network, a farmer community can authorise an organisation to issue a certain type of evidence, under a published policy, for a set period. Anyone checking that evidence later, anywhere in the world, can follow the whole chain and see whether that authority is still valid.
Ask to use it
The data’s owners, or the rules their community has agreed, decide who gets access, for what purpose and for how long. Every grant of access can be recorded and withdrawn. The same pattern can link networks at community, national and global level, with no central database and no single gatekeeper.
Creating agricultural knowledge networks
We’ve seen why agricultural knowledge networks fail. These patterns help create networks that scale and deliver value for everyone involved.
- Governance first
- Before any technology, members agree who can join, what they contribute, what they get back, and how decisions are made. Membership classes for farmers, researchers, platforms and funders ensure no single party can capture what’s built together.
- Shared schemas and vocabularies
- The hardest and most important step is agreeing what gets recorded, so results from different farms, trials and platforms can be compared. We build on established standards such as AGROVOC, MIAPPE and the CGIAR ontologies, and on open farm data models like farmOS, rather than inventing new ones.
- Federated by design
- Data stays with whoever supplied it, and in the systems they already use. The network connects those systems; it never needs to hold everything.
- Provenance and permissions as credentials
- Membership, authority and consent are expressed as verifiable credentials that can be checked, time-limited and revoked. Members can prove what they need to without exposing everything.
- Checks against the standard
- Contributions can be checked against the agreed schemas, so data that enters the network is fit for reuse.
All of this builds on OpenVTC, open-source trust infrastructure built to open standards.
What it makes possible
- Better decisions on farm. Farmers benchmark against genuinely comparable farms and learn what works in conditions like theirs.
- Research at a new scale. Researchers pool trials across projects, seasons and regions for analyses no single project could run.
- Investment that keeps paying back. Funders’ trial evidence stays useful long after each project ends.
- Trustworthy inputs for models and AI. Predictive models and AI advisers learn from data whose source, quality and permissions are known.
- Proof that travels. Farmers reuse verified evidence with buyers, lenders and schemes, instead of re-proving the same thing for each.
Who does what
VT3
Governance and membership design, schema selection and standards, data architecture, the trust layer, with options to host and maintain networks.
Delivery partners
Recruiting and supporting farmers, field and offline delivery, agronomy, and domain-specific evidence.
The community
Sets the rules, and owns what it builds.
Experience that shaped our approach
We’ve spent years working on projects involving farm data, comparisons and trust.
The Yield Enhancement Network (YEN)
Created by Roger Sylvester-Bradley and Daniel Kindred at ADAS, YEN has demonstrated how competitions can make farmers want to share data. YEN defined what counts as a comparable result, built data models, and produced benchmark reports farmers genuinely valued. YEN now has more than 5,000 crop records and networks in several countries: living proof that agricultural data sharing creates value.
FarmPEP
A knowledge-exchange platform that showed how to structure knowledge across a fragmented sector, connecting topics, organisations, projects and people, and revealing what drives contribution and what doesn’t.
Oliver Sylvester-Bradley consulted on both projects to support the development of suitable data structures, user interface design and knowledge architecture, working alongside ADAS, researchers, farmers and developers, and bringing ADAS and Creative Co-op together to build platforms that worked.
What we learned
- When the funding ends, the data and the community fade, unless the network is designed to outlive the project.
- Consent has to be designed for reuse from day one, or valuable evidence gets stranded.
- Consistent data models are essential, to provide interoperable, queryable knowledge graphs.
- Data must stay with whoever supplied it. Centralising it erodes trust and creates a single point of failure.
- Open sign-up without verified membership invites noise, and activity follows funded programmes rather than growing on its own.
Since YEN and FarmPEP began, verifiable trust infrastructure has matured. Farming networks can now share and compare data without a central data store. VT3 brings that infrastructure together with our experience, to help design and run farm data networks that are more effective, more equitable and more secure.
Related initiatives and standards
We follow and build on work across the sector, including:
- Open farm data and tools
- farmOS, Proof.ag, AgStack (Linux Foundation)
- Data sharing and governance
- AHDB Farm Data Exchange, FIG/FDTP, the Farm Data Principles, FAIR principles
- Data spaces and knowledge systems
- Data4Food2030, the EU’s AKIS framework
- Vocabularies and ontologies
- AGROVOC, MIAPPE, the CGIAR ontologies
Talk to us
Planning a research programme, a consortium bid, or a farmer data network? We’d like to hear what you’re trying to achieve, and help you get the foundations right from the start.
Not everything is production-ready. OpenVTC is still developing, so each component is assessed for maturity, security and suitability.