Understanding the data and models behind our recommendations
Tru-Stock AI analyzes four core data streams to generate recommendations:
Sales History
Historical transaction records showing what sold, when, and in what quantity. Used to calculate demand velocity and identify trends.
Inventory Balances
Current on-hand quantities for each SKU across locations. The foundation for days-of-stock calculations.
Vendor Lead Times
Average time from placing an order to receiving stock, per vendor. Critical for timing reorder points correctly.
Purchase Orders
Open POs showing what’s already on order. Prevents duplicate ordering and factors incoming stock into recommendations.
We calculate average daily demand using a weighted rolling window of your sales history. More recent sales are weighted more heavily than older data, allowing the model to adapt to changing demand patterns. The system also identifies seasonal trends by comparing month-over-month and year-over-year patterns where sufficient historical data exists.
The reorder point for each SKU is calculated as:
Safety stock is a buffer calculated based on demand variability—SKUs with inconsistent sales patterns get higher safety stock to account for unpredictability. When your on-hand quantity drops below the reorder point, the system flags it for reorder.
Suggested order quantities aim to bring your stock level up to a target that covers expected demand through the next lead time cycle, accounting for what’s already on order. The formula considers:
The target stock level typically covers 2–3 lead time cycles worth of demand plus safety stock, ensuring you have enough runway between orders.
Every SKU is assigned a risk category based on its days of stock remaining relative to vendor lead time:
Auto glass parts often have multiple variants (left/right, heated/non-heated, with/without sensors). Tru-Stock AI automatically groups related SKUs by their parent part number and analyzes them both individually and as a group, giving you a complete picture of your coverage for each inventory or glass type.
As you import more sales data over time, the system’s predictions become more accurate. We recommend at least 90 days of sales history for reliable demand forecasting, with 12+ months being ideal for capturing seasonal patterns.
A Tool, Not a Replacement
These models provide a strong analytical foundation, but they work best when combined with your industry knowledge. Factors like upcoming promotions, weather events, new vehicle model releases, or supplier issues should always be considered alongside our data-driven suggestions.