Meuze

Demand is local.
Intelligence should be, too.

Meuze brings the outside world into each location’s forecast. We aggregate variables across movement, markets, behaviour, weather, events and disruption, then pair them with the location’s historical data.

Across millions of candidate variables, our machine learning models test which relationships improve that location’s forecast and keep reassessing them as conditions change.

The world around a location,
made usable by its models.

We turn source data into variables aligned to a place and time: where something happens, when it happens, how long it lasts, its intensity and how it changes. These are examples of the wider variable universe. Each location’s model uses the signals selected for its forecast.

Footfall & traffic
Pedestrian volumes, traffic flow and changes in access around a location.
Agriculture & commodities
Agricultural conditions, harvest cycles and commodity movements that can affect supply, costs and consumption.
Consumer & transactional behaviour
Aggregated trends in GLP-1 usage, purchasing patterns, visit frequency, basket composition and channel mix.
Weather & seasonality
Local conditions and forecasts, including temperature, rainfall, seasonal patterns and extreme weather.
Large-scale events & campaigns
Major sporting events, festivals, public holidays and marketing campaigns, aligned to their timing, reach and proximity.
Political disruptions
Disruptions that change movement, trading conditions, supply or consumer activity—and the places and periods they affect.

Train. Test. Reallocate.
Keep learning.

Our machine learning process finds the combinations that improve predictions for a particular location, forecast target and time horizon.

Pair signals with local history.

We align relevant external variables with the location’s historical sales, orders, item mix and operating results. This gives the models a record of how local demand behaved under different combinations of conditions.

Train, compare and recombine.

We train and compare models, test combinations of variables, swap weaker inputs and reallocate their contribution. Learned patterns can be transferred into further tests, where they are evaluated against the location’s own results.

Select for predictive value.

We evaluate forecasts against historical periods kept out of training and select the drivers and combinations that improve performance. The useful signals can differ by location, item, season and forecast horizon.

Reassess as results arrive.

New actual results reveal where predictions held up and where they missed. We use that feedback to retrain, revisit the inputs and update their contribution as demand patterns and local conditions change.

A local forecast.
A clear operational plan.

Sales, order and item forecasts, down to 15-minute intervals, become specific preparation, replenishment and staffing actions.

Delivered through Meuze Core, Core+ with labour planning, or inside operators’ systems and software vendors’ products through Meuze API.

Interested?
Contact us.

Book a demo