Our approach

A question, a test, a reasoned next step.

Agree what the investigation must resolve, examine the evidence and leave with findings your team can use.

A concept diagram routing source information through checks and optional human review into business records
From source to business record · Concept diagramSource information passes through extraction, validation and routing. Cases needing human judgment take a review path before the business record is updated.

Agree the scope before the investigation.

Start with the task and the decision it affects. We establish who will use the output, what material is available and which constraints matter. An early conversation may identify that data preparation or clearer requirements should come first.

The proposal sets out the question, method, deliverables, exclusions, fee, dependencies and review dates. Access, data handling and ownership are agreed before the work begins.

  1. Frame the question

    Define the intended task, acceptable output, baseline and what evidence would support proceeding or stopping.

  2. Prepare the material

    Establish permissions, source versions, sample coverage and assessment examples. Record where the material differs from intended use.

  3. Investigate or build the experiment

    Use the agreed approach. Keep configuration, intermediate observations and material changes visible at review points.

  4. Interpret the findings

    Separate observation from inference. Review failures, trade-offs and limits alongside successful examples.

  5. Hand over the decision record

    Deliver the agreed report and shareable artefacts, explain the recommendation and identify remaining questions. The next step may be further research, implementation or stopping.

Your role and the handover.

A decision owner sets priorities. A domain reviewer helps judge substantive correctness. A technical contact establishes access where systems are involved. Agreeing those roles early prevents an experiment being assessed against an unstated expectation.

The handover records assumptions, methods, relevant versions, observations and limitations. Code and datasets are transferred where included in the scope and where permissions allow.

A prototype is not a production release, and an evaluation is not certification. Production engineering, security review, hosting and ongoing support are separately scoped. Findings can be used by your own engineers or an implementation partner.

See our worked example, explore the engagements or describe your question.

A considered next step

What do you need to find out?
Start with the question.

Tell us the task, the material available and the decision you need to make. We will discuss whether a focused investigation is the right next step.

Discuss your question hi@manaslabs.co.uk
01   Describe the task02   Agree the investigation03   Review the evidence