Know what to reorder, when, and where
Custom AI that forecasts demand per SKU and location, sets reorder points and safety stock that update themselves, and tells you which stock is never going to sell.
How it works
Inventory decisions are made weekly on numbers that were true a month ago. Reorder points get set once and never revisited, safety stock is a round number someone picked in 2019, and the real cost sits in two places at once: the SKU that ran out in the branch that could have sold it, and the pallet that has not moved since spring. We build AI planning systems on top of your ERP or WMS that forecast demand at the level you actually buy and ship, recalculate the thresholds as conditions change, and surface the exceptions worth a planner's attention.
Demand forecasts at the level you actually plan
Per SKU, per location, per week, with seasonality, promotions, and new-product cold starts handled rather than smoothed away.
- Models seasonality, trend, and promotional lift separately
- Handles intermittent and lumpy demand where simple averages fail
- Cold-starts new SKUs from attribute similarity to existing lines
- Publishes forecast accuracy by segment so planners know what to trust
Reorder points and safety stock that recalculate themselves
Thresholds are derived from live demand variability and real supplier performance, then pushed back into your ERP instead of a side spreadsheet.
- Safety stock sized to your target service level, not a flat multiplier
- Tracks actual versus promised supplier lead times and their variance
- Adjusts buffers automatically when a supplier gets slower or less reliable
- Writes updated min/max and reorder points straight into your system of record
Dead stock and multi-location allocation
Finds the inventory that is not going to sell where it currently sits, and proposes the transfer, markdown, or write-off before it ages further.
- Flags slow, obsolete, and excess stock with the capital tied up in each
- Recommends inter-branch transfers to where demand actually is
- Balances allocation across DCs, stores, and third-party warehouses
- Prioritizes exceptions by cost impact so planners work the top of the list
Where teams use it
Built for real revenue work.
Distributors running dozens of branches
Every branch buys for itself, so the same SKU is short in one location and overstocked in three others. Network-level forecasting plus transfer recommendations recovers stock you have already paid for instead of reordering it.
Manufacturers exposed to supplier lead-time swings
Buffers are set against quoted lead times that suppliers stopped hitting two years ago. Tracking real delivery variance per supplier and part sizes the buffer to the risk that exists rather than the one on the contract.
Retail and ecommerce with heavy seasonality
Peak forecasts built from last year's totals miss the shape of the curve and the promotions inside it. Weekly SKU-location forecasts let buying commit earlier with less hedging, and flag the leftovers early enough to clear at a reasonable margin.
“The transfer recommendations paid for the project before the forecasting did. We were sitting on stock in the wrong branches and reordering the same parts nationally, and nobody could see it across sixteen locations until the system laid it out.”
FAQ
Questions, answered.
Do we have to replace our ERP or WMS?
No. We build on top of what you run, whether that is NetSuite, SAP, Dynamics, Odoo, or something custom, and read stock, sales, and purchase history through APIs or a nightly extract. Recommendations are written back as reorder points, min/max levels, or draft purchase orders inside your system of record. Planners keep working where they already work.
How much history do we need for the forecasts to be useful?
Two years of transactional history is ideal because it gives us two passes at seasonality, but we regularly start with twelve to eighteen months. For SKUs with little or no history, the model cold-starts from similar products by attribute and tightens as real demand arrives. We will tell you upfront which parts of your catalogue we can forecast well and which need a simpler rule for now.
Will it place orders automatically?
Only if you want it to, and usually not at the start. Most clients run in recommend-and-approve mode, where the system proposes quantities and a planner confirms, then move low-risk categories to automatic once they trust the numbers. High-value or long-lead items typically stay under human approval permanently, which is a reasonable choice rather than a limitation.
What does a project cost and how long does it take?
Inventory builds usually run four to twelve weeks depending on how many locations, systems, and SKU classes are in scope, quoted as a fixed project cost rather than an hourly burn. You get a dedicated team, code and models you own outright, and infrastructure operated under SOC 2 aligned controls with 99.9% uptime. There is no per-SKU or per-seat licence waiting after go-live.
Keep exploring
More from the Leadify platform.
Custom AI Software
The engineering practice behind every build: fixed project cost, dedicated team, and code you own.
AI Automation
Wire forecasts into the purchase orders, transfers, and approvals they should trigger automatically.
AI for Retail
Store-level demand, assortment, and replenishment patterns specific to multi-store retail operations.
Stop planning inventory in a spreadsheet
Give us a demand and stock extract and we will show you the forecast accuracy and the dead stock hiding in it.