# Demand Planning Data Readiness Checklist: Is Your History Fit to Forecast? > Source: https://netray.co/tools/demand-planning-data-readiness-checklist > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Free Tools ## Short answer Work through the data checks that decide whether a statistical forecast can be trusted, from history length to unit of measure and product hierarchy. ## Category supply-chain ## Audience - Demand planner - ERP data steward - Supply chain analyst ## Tool **Kind:** checklist **Groups:** #### History coverage **Items:** **Text:** At least 24 months of history exist for items with seasonality **Critical:** true At least 36 months exist for items where you want seasonal indices to be stable History is stored at the same time bucket you will forecast in **Text:** Months with no sales are present as zero, not missing **Critical:** true The first and last period of each item are verified against sales reports Items launched or discontinued inside the window are flagged **Text:** History totals reconcile to finance within an agreed tolerance **Critical:** true #### What counts as demand **Items:** **Text:** You have decided between orders by requested date and shipments by ship date **Critical:** true **Text:** Returns and credit notes are netted or excluded by a stated rule **Critical:** true Cancelled order lines are removed Intercompany and inter-plant transfers are excluded from external demand Free of charge, sample and warranty shipments are tagged and handled Lost sales from stock-outs are flagged or estimated Backorders are counted once, in the period of the original request #### Outliers and events **Items:** **Text:** Large one-off orders are reviewed and marked as non-recurring **Critical:** true Promotions and price changes are recorded as events with dates Plant shutdowns, strikes and supply disruptions are marked Pandemic, tariff or customer loss periods have an agreed treatment **Text:** Outlier cleansing rules are written down and applied consistently **Critical:** true Original and cleansed history are both retained #### Master data and hierarchy **Items:** **Text:** Every item has one base unit of measure and conversion factors are correct **Critical:** true **Text:** Superseded part numbers are linked so history carries forward **Critical:** true Product family, brand and category hierarchy has no orphans Customer and region hierarchy is defined for aggregation Item status (active, new, phase-out, obsolete) is maintained Lead times and planning bucket settings exist for every forecasted item #### Process and ownership **Items:** A named owner is accountable for history quality **Text:** The forecast lag used for accuracy reporting is agreed **Critical:** true Every forecast version is stored so accuracy can be measured later The data refresh schedule fits the planning calendar A baseline statistical forecast exists to measure manual overrides against Exceptions found in data checks have a route to correction at source **Scoring Note:** Fix all critical items before trusting any statistical forecast. Non-critical gaps reduce accuracy but rarely invalidate the result on their own. ## Content **Intro:** Forecast models only reflect the history they are given. This checklist covers the data checks that matter most before you fit a model or load a planning tool: coverage, definition of demand, outliers, master data and ownership. **Sections:** #### How to use the checklist **Body:** Go through the groups with your planning lead and an ERP data owner. Mark each item as done, partly done or not done, and record who will fix gaps. Treat the critical items as gates: if history does not reconcile to finance or demand is not defined, stop and fix that before modelling. Export the completed list as evidence for a planning tool implementation or an audit of your forecast process. #### What good demand data looks like **Body:** Good history is complete, consistent and explainable. Each period has a value, including zero. The definition of demand is written down, so everyone knows whether returns, cancellations and transfers are in. Unusual spikes have a recorded reason. Items that changed part number keep their history. When these hold, simple methods perform well. When they do not, even advanced models fail. #### Handling outliers and lost sales **Body:** Do not delete unusual points silently. Mark them, state a rule such as replacing a month with the median of surrounding months when an order exceeds a set multiple of normal, and keep the original. Stock-outs hide true demand in shipment history. Flag those periods and consider uplifting them, or use order history instead of shipments to avoid the problem. **Bullets:** - Keep original and cleansed series - Write the cleansing rule down - Review the largest items by hand - Flag stock-out periods #### Common mistakes **Body:** Teams forecast on shipments and then wonder why the model repeats past shortages. Others forget unit of measure conversions, so cases and pieces are added together. A third problem is superseded part numbers that split history across two items and make both look like poor candidates for modelling. Fix these at source in master data, not in the forecast sheet. **Cta:** If data checks like these are run manually before each cycle, Netray can run them against your ERP tables and list the failures with the query used. **Faq:** #### How much history do I need for demand planning? **Answer:** Three years of monthly data is a good target for seasonal items, with two years as a practical minimum. Stable non-seasonal items can be forecast with less, but accuracy statistics need enough periods to be meaningful. #### Should I forecast from orders or shipments? **Answer:** Orders by requested date are closer to true demand because shipments reflect what you could supply. Shipments are acceptable if you flag stock-outs. Whichever you pick, document it and keep it consistent. #### How do I treat a large one-off order? **Answer:** Review it with sales. If it will not recur, mark it as non-recurring and replace or smooth it in the cleansed history, keeping the original. If similar orders recur, leave it in and model it as an event. #### Who should own demand history quality? **Answer:** Assign one named person, typically a demand planning lead working with an ERP data steward. Quality problems found in the forecast should be corrected at source in the ERP, otherwise they return every cycle. ## Search terms demand planning data quality checklist, forecast data cleansing checklist, demand history cleaning, data readiness for demand forecasting, forecast hierarchy setup checklist, demand planning implementation checklist, clean sales history for forecasting, outlier handling demand planning --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json