A warehouse you can test the future on.

I was placed with Walmart US, and the brief sat deep inside DC operations. Every minute of dock time, every mis-slotted pallet, every under-staffed shift costs money downstream. The recurring question was simple to ask and brutal to answer: what's the best way to run this building today?

A DC receives freight (less-than-truckload and full-truckload), holds staple stock, assembles and ships to stores. To design for it, I first had to understand how goods actually flow through it.

I explained it the way I explained it to the team: with a rocket and a chessboard.

A digital twin is a virtual replica of a physical thing, same shape, temperature, wear, kept in sync by real-time data. On the virtual side you can model, simulate, monitor and predict without risking the real one.

A chess app quietly simulates thousands of futures for every move, then picks the best one. A twin does the same with a Monte Carlo method, running many possible futures to find the best call.

Each one is a hypothesis waiting to be tested. Treating them as "questions the user wants to ask the future" defined the whole product, and later became its central interaction.

And Curious Jane, the explorer

Jane wasn't content to react. She wanted to experiment, to poke at five scenarios in her head and see which one wins. Today that means spreadsheets and waiting on other teams. Her frustration became the question the product had to answer:

"What if there was a way to see the future, to take objective decisions backed by data that actually improve how the warehouse runs?"

An ad-hoc ask landed mid-research: show leadership what this could look like, now.

So I zoomed in. Sketching over a line-map of the warehouse, I built an outline, then added an isometric view of the building, a live viewport, and cards suggesting simulations. The first screens got a green light from higher up, and with it, room to redefine the brief and go deeper on research.

With the green light secured, I pulled the whole project onto a single synthesis map. The final shape needed three things: a live viewport, a hypothesis builder, and an outcomes page that optimises the answers.

Then Microsoft Ignite put the "industrial metaverse" on stage: digital twins for industry, exactly this area. Equinor, Mercedes and Coca-Cola were already building twins before building the real thing. Seeing that told me the bet wasn't mine alone.

The core interaction: write a hypothesis

The everyday questions from research become a conversational, fill-in-the-blanks sentence. Ask the future a question in plain language and the system does the rest, testing the decision before it's real. When the outcome looks good, one action pushes it to production.

The concept doesn't stop at a loading dock. The same three parts, a live twin, a plain-language hypothesis builder, an outcomes optimiser, could work anywhere: point them at a store, a transport network, a fulfilment line, and let anyone test decisions before making them.

Measuring the process honestly, I plotted the double diamond against the pages it took to document each phase. The back half of the second diamond, iteration and refinement, took by far the most time. The planned process and the real one rarely match, and that's the lesson.