Skip to content
Wellbridge Investment Technologies, home

Approach

Modelling is designed to be reviewed, not defended after the fact.

Our approach rests on five principles and a simple engagement structure, written from the reviewer’s side of the table: what would they need to see to trust this number?

Principles

  1. 01

    Evidence before claims

    We describe capabilities, results and clients only when they are confirmed and we are permitted to share them.

    In practice

    • Illustrations are labelled as illustrations.
    • Figures in demonstrations are synthetic unless stated otherwise.
    • If we are unsure, we say so.

  2. 02

    Lineage you can follow

    Every figure should lead back to the data and the model that produced it.

    In practice

    • Sources, timestamps and transformations travel with the value.
    • Stale, missing or modelled inputs are marked, not hidden.
    • Calculations are explained in plain language next to the result.

  3. 03

    Review by design

    Model review is part of the workflow, not a disclaimer at the end.

    In practice

    • Consequential actions require an explicit approval step.
    • AI suggestions show their sources and can be rejected without friction.
    • Every decision keeps a record of who approved what, and why.

  4. 04

    Calm, accessible interfaces

    Dense information should still be easy to read and operate.

    In practice

    • We design to WCAG 2.2 AA as a baseline.
    • Motion explains relationships and can be reduced or switched off.
    • Keyboard and screen-reader use are planned into testing from the start.

  5. 05

    Minimal data

    We collect only what a task requires.

    In practice

    • No advertising trackers on our website.
    • Enquiry data is limited to what we need to reply.
    • Retention periods are defined before collection begins.

Engagement

How an engagement is structured

A typical outline. Scope and timing are agreed for each engagement.

  1. 01

    Discover

    Understand the decision, the people involved and the model or analysis it depends on.

  2. 02

    Map

    Trace data sources, model inputs, hand-offs and review points. Agree what a defensible figure means for this work.

  3. 03

    Prototype

    Build a working model or illustration with synthetic data, so feedback is concrete from the start.

  4. 04

    Review

    Validate with the people who will use and sign off on the output. Confirm assumptions, controls and accessibility.

  5. 05

    Deliver

    Implement, document and hand over — with the lineage, model documentation and review records intact.

Two colleagues reviewing a printed document across a pale oak table, a glass-walled office behind them.
Engagements start in the room where the work already happens, not in a slide deck.

Boundaries

What we will not do

  • Present illustrative data as live results.
  • Let an automated system take an irreversible step without a person’s approval.
  • Publish a client name, quote or result without permission.
  • Collect personal data we do not need.

Want to see how this would apply to your work?

Describe the decision and who is involved. We will suggest where to start.

Start a conversation