- An assortment model forecasts demand for every item at every store, so each location gets a mix that fits its own shoppers.
- H2O Driverless AI builds the demand and fit models from your sales, space, inventory and local data, with reason codes on every forecast.
- The same forecasts show where to move stock between stores before a local trend turns into a stockout or a markdown.
- H2O is the scoring layer under the planning and allocation systems you already run, not a replacement for them.
One assortment for stores that are not the same
You plan by season and send most stores a similar mix. But each store has different shoppers, weather, shelf space and stockroom capacity, so the same plan runs out of some items and leaves others on the shelf.
A stockout loses the sale and sends the shopper somewhere else. A markdown gives away margin you already paid for. Both show up in the numbers after the season, when it is too late to fix the plan. You already hold the signals to plan by store: they sit in point-of-sale, inventory, space planning and e-commerce systems that were never joined into one view of what each location will sell.
How assortment optimization works
Your planning team builds explainable demand models in H2O Driverless AI and uses them to shape each store's mix.
Step 1: join the signals
Bring together sales history by store and item, display space, inventory, online behavior, local events and weather forecasts.
Step 2: forecast demand by store and item
H2O Driverless AI builds time series models and regression models with automatic feature engineering on seasonality, price, space and local trends.
Step 3: score the fit of each item for each store
The forecasts rank which items belong in each location's mix, and how deep to stock them.
Step 4: explain every recommendation
Reason codes from explainable AI show your planners why an item was added, cut or deepened for a store.
Step 5: rebalance during the season
Updated forecasting shows where to move stock between stores to follow local demand.
Working alongside the systems you already run
H2O is the scoring layer under the planning and allocation systems you already run, not a replacement for them. It sends forecasts and ranked assortments to those systems, and your planners keep the final decision. A first project usually compares the model's plan with last season's plan for one category, so the change is measured before it is trusted at scale.