How we work
Start with the decision or outcome. Define the inputs and constraints, test what matters, and deliver something that can be inspected and used.
Input → evaluation → delivery
A clear scope before sensitive details.
- Define
- The question, available material, intended environment and disclosure limits.
- Evaluate
- The baseline, acceptance criteria and limits of what the evidence can show.
- Deliver
- The agreed output, supporting records, responsibilities and handover.
Quantitative project
- Output
- A model, simulator, report, analysis tool, bot or integration.
- Acceptance
- Defined baseline, evaluation sample, costs and operating conditions.
- Delivery terms
- Software ownership, integration, handover and any ongoing operation.
Data work or licence
- Output
- Processed records, a scoped dataset delivery or a licence to specified material.
- Acceptance
- Coverage, field definitions, validation, gaps and correction handling.
- Delivery terms
- Source rights, permitted use, delivery cadence, distribution and any exclusivity.
Infrastructure rental
- Output
- Access to specified compute and storage, plus the included infrastructure tasks.
- Acceptance
- Workload fit, allocation, access and configuration checks.
- Delivery terms
- Rental period, administrators, retention, support and end-of-rental handling.
Confidentiality and scope
Client identities and project details are kept out of public case studies. Before confidential data is shared, the scope identifies permitted sources, recipients, access, retention and any publication rights.
The acceptance test also needs a next step if the available inputs cannot support the intended result. For a model, that may mean comparing against a baseline on unseen data; it does not mean guaranteeing a profitable outcome.