Predictive Stock Replenishment Workflows
REPLENISHMENT FORECAST DECISION
Separate the demand forecast from the purchase decision
A forecast estimates future demand. Replenishment converts demand, current stock, inbound supply, lead time, safety policy and ordering constraints into an action. Review both stages so a plausible prediction cannot create an unjustified purchase.
01 – FORECAST
Estimate demand with context
Use clean history at the right item, location and period, then record promotions, stockouts and structural changes.
02 – PLAN
Calculate the net requirement
Combine available stock, reservations, trusted inbound, lead time, safety policy, minimums and pack multiples.
03 – APPROVE
Release a controlled order
Show recommendation evidence, route exceptions and keep supplier transmission under the approved authority.
Pilot forecast-to-order on stable items
Choose products with trustworthy history and supplier data, run recommendations in shadow mode and compare forecast, suggestion, decision and actual demand.
Predictive replenishment uses a demand estimate to support when and how much to buy, make or transfer. It can help an SME see future shortage earlier, but it also magnifies bad item identity, units, lead times and stock balances. The workflow should expose its inputs and create an approved supply document, not automatically convert a model number into a purchase.
Use the reorder-point automation guide for deterministic policy, the inventory rebalancing guide for inter-location transfers, and the purchasing and receiving service for execution controls.
1. Choose the forecast grain and horizon
| Decision | Example | Risk |
|---|---|---|
| Item grain | SKU and variant | Family totals hide size or colour demand |
| Location grain | Warehouse or branch | National demand hides local shortages |
| Period | Day, week or month | Too coarse misses lead-time timing |
| Horizon | Lead time plus review and safety window | Too short cannot change supply |
| Demand basis | Orders, consumption or shipment | Returns and stockouts distort history |
Microsoft’s Sales and Inventory Forecast documentation describes forecasting from item ledger history and notes that insufficient or highly variable data may prevent a prediction. It also describes aggregated limitations across locations and variants in that specific extension. Select a method and grain that match the business rather than assuming an AI label solves segmentation.
2. Prepare demand history
- Correct item, variant and unit mappings across the history.
- Mark stockout periods where sales understated demand.
- Separate returns, cancellations and one-off project orders.
- Record promotions, price changes and channel launches.
- Identify discontinued items and approved successors.
- Preserve the extraction date and source transactions.
3. Create and review the forecast
Generate a baseline, then show forecast interval or uncertainty where the method supports it. Compare with a simple baseline such as recent average or seasonal prior period. Let commercial and operations reviewers record known events without overwriting the original model output. Keep forecast version, method, training window, adjustments and approver.
4. Convert forecast to net requirement
| Component | Control |
|---|---|
| Forecast and actual demand | Avoid double-counting firm orders |
| Available stock | Exclude quarantine and protect reservations |
| Inbound supply | Trust only confirmed quantity and expected date |
| Lead time | Use supplier and route evidence, not a generic default |
| Safety policy | Define service rationale and review frequency |
| Order constraints | Apply minimum, maximum and pack multiple |
Business Central planning documentation describes balancing supply and demand and generating action suggestions such as purchase, transfer or production. Odoo reordering rules likewise use configured minimum, maximum, routes and quantities. Predictive demand can feed that policy, but the netting logic must remain testable.
5. Score the recommendation
Show current cover, projected shortage date, forecast demand, firm demand, available stock, inbound supply, proposed quantity, supplier, expected arrival, resulting cover and value. Flag low-confidence forecasts, new items, intermittent demand, unusual spikes, expired prices and supplier constraints. A recommendation with weak evidence should become a review item, not a precise-looking order.
6. Approve and release
Use purchasing limits by user, value, category and supplier. The approver may accept, reduce, defer, change supplier or reject with a reason. Changes must recalculate resulting stock and show whether constraints still pass. Supplier transmission follows approval and returns an acknowledgement.
7. Measure forecast and inventory outcomes
Evaluate bias, absolute error, service level, stockouts, excess, emergency purchases, late supply, overrides and write-down exposure by item class. Do not optimise one metric alone: a low forecast error can coexist with poor service when lead time is wrong. Review whether accepted recommendations arrived and were consumed as expected. Segment results by stable, seasonal, promotional, new and intermittent items so a strong aggregate does not hide one product class that needs a different policy.
8. Treat new and intermittent items separately
A new product has no direct history, while an intermittent service part may have long periods of zero demand followed by an important requirement. Do not force the same forecast method used for regular sellers. Use an approved analogue, launch plan, customer commitment, service policy or manual minimum and label the assumption. Keep the item review-only until enough relevant evidence exists. For superseded products, connect demand to the approved replacement without copying history blindly; compatibility, price and customer adoption may differ.
Worked pilot
A wholesaler chooses forty regular items with twelve months of clean weekly history. The baseline forecast excludes a once-off project order and marks two stockout weeks. Planning nets firm sales, available stock and confirmed purchase orders, applies supplier lead time and case multiples, and proposes draft orders. Three volatile items are review-only. Buyers record why they change a quantity. After eight weeks, the team compares forecast error, supplier delay, overrides, stockouts and excess before adding more products or automating release.
Sources checked
- Microsoft Learn: Sales and Inventory Forecast extension
- Microsoft Learn: About planning functionality in Business Central
- Odoo: Reordering rules
Reviewed by
Mitrend Digital editorial team
2026-07-17
Evidence used for this page
Reviewed against current Microsoft Business Central sales-and-inventory forecasting and supply-planning documentation plus Odoo reordering rules. Includes an original forecast-to-order workflow and exception test.
Turn the guide into a practical next step
This resource provides general implementation guidance. Verify platform settings, tax, legal, payment and operational requirements against the current business context before making a live change.