Shelf layout
It shows in advance how moving an aisle affects customer flow and basket size.
Test it in simulation before changing your shop.
AI-Powered Retail and Warehouse Digital Twin
Builds a 3D copy of your store, warehouse or factory. Before you move a shelf, it calculates the sales impact, staff routes and picking time.
The problem
Shop layouts are changed on guesswork.
Changing a store layout is expensive: labour, downtime and sales risk. Yet the decision is often made from experience, which means guesswork. If it is wrong, you find out weeks later when revenue drops.
In the warehouse, the problem is easier to measure but the same: add a few metres to a picking route, multiply it by thousands of orders a day, and it turns into hours.
What it tests
We build a copy with the real store’s measurements. Changes are tried there first, then rolled out on site.
Customer flow
Movement patterns learnt from real camera and checkout data are replayed in the simulation.
High-traffic areas
A heat map reveals blocked aisles and dead zones nobody visits.
Staff routes
It works out the layout that shortens shelf-stocking and picking routes.
Robot integration
Routes for autonomous carriers are tested with human traffic before anything goes live on site.
Picking time
It measures walking distance and picking time per warehouse order.
Who is it for
Those who want to see it before changing the layout.
Retailers hesitant to change store layouts
Chains that only learn how moving an aisle affects sales after doing it, when reversing course is costly.
Warehouses where picking time is the bottleneck
Operations where walking distances grow with orders and layouts rely on gut feel.
Businesses considering robot investment
Anyone who wants to see whether automation will really pay off before buying.
Live demo
Sample run · no real data
Digital twin · heat map
Simulation output is illustrative
How it works
Four steps.
01
Measure
We take the store plan, shelf heights and checkout locations. Then we build a scaled copy.
02
Feed
Camera, checkout and stock data connect to the copy. The simulation imitates real behaviour.
03
Try
The new layout is tested in a virtual space. Results are compared by heat map and time.
04
Apply
Put the winning layout into practice. Feed real results back into the simulation.
Technology
What it runs on.
- NVIDIA Omniverse
- Isaac Sim
- Cosmos
- Python
- Go
- USD
The scene is built in USD format on NVIDIA Omniverse. Robot and human movement is simulated with Isaac Sim. Cosmos handles scenario generation and image synthesis. The data pipeline is Python, the service layer is Go.
Gallery
FAQ
What you may want to know before we talk.
What do you need to set up the model?
The space dimensions and shelf layout. We use architectural plans directly when available. If not, we measure and build it. The simulation becomes realistic once you connect product and sales data.
How reliable is the simulation?
Only as accurate as your data. With real sales and traffic data, it shows the trend accurately. We do not promise exact figures. We built it as a comparison tool.
Will it talk to our current warehouse software?
It works with systems that can export stock and order data. We build the integration during initial setup.
How many scenarios can we test at once?
There is no limit. Each scenario stays as a separate record. You can compare two side by side. The record also shows which one you chose and why.
Can I buy it now?
No, still in development. We are discussing a pilot.
Contact
Something like this?
do you want to build?
Tell us in one sentence. We will explain what you need and how long it will take.