From Data Sprawl to a Single Source of Truth

06Test before you build

Design, validate, deliver|Step 6

Test Before You Build

A prototype is the cheapest place to be wrong. Test it with the people who will use it.

The cheapest place to be wrong

CLEAR ends with a plan. This step is about what happens when that plan meets the people who will use it. A prototype is the cheapest place to be wrong.

Moving a metric, cutting a view or reworking a flow takes hours in a prototype. After development, the same change costs far more. A clickable prototype with realistic data feels close enough to the real thing that people react honestly, without anyone writing production code.

How to apply it

01 / 06
  1. Prototype

    Prototype one or two decision flows, not the whole platform.

    Use realistic data so gaps show up now rather than after launch.

  2. Test

    Test with the people who will use it.

    Give them real tasks rather than a demo, and watch where they hesitate.

  3. Hierarchy

    Ask the hierarchy questions.

    What deserves to be prominent? What’s secondary? What must be visible immediately, and what can live one level down? Does this match how people think about the operation?

  4. Refine

    Refine and test again.

    Do it before committing development resources.

  5. Blueprint

    Hand engineering a blueprint, not just screens.

    Include where each number comes from, how it’s defined, who owns it, and what data is still missing.

  6. Foundation

    Plan for what a trusted foundation makes possible.

    Predictive analytics are only as good as the data beneath them. Built on competing versions of the truth, a model predicts the wrong things faster. Built on agreed, owned data, it can see trouble coming.

In the field

At the oil and gas company, a validated clickable prototype became the blueprint for engineering. The platform was then developed and delivered.

With data it could trust, the operator went further and now runs predictive machine-learning models in-house that flag delivery and flow-management issues, the kind of problems that could otherwise shut down operations.

The blueprint for engineering

  • Interface
  • Interactions
  • Information hierarchy
  • User flows

Built on it

  • A delivered platform
  • Predictive machine-learning models, run in-house

From prototype to platform

Prototype, test and refine until the design holds. Then hand off a blueprint, deliver the platform, and build the predictive layer on data people trust.