# Demand Pattern Classification Reference: ADI, CV2 and Forecast Method > Source: https://netray.co/tools/demand-pattern-adi-cv2-classification-reference > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Free Tools ## Short answer Look up the four demand classes from ADI and CV2, with the usual forecasting method and stocking approach for each. ## Category supply-chain ## Audience - Demand planner - Materials planner - Inventory analyst ## Tool **Kind:** worksheet **Mode:** reference **Columns:** #### Item **Key:** item **Type:** text #### Definition or threshold **Key:** definition **Type:** text #### Typical stocking approach **Key:** policy **Type:** text #### Suggested forecast method **Key:** method **Type:** text **Rows:** **Item:** ADI **Definition:** Average demand interval: periods divided by number of periods with demand **Policy:** Higher means demand is rarer **Method:** Used to classify **Item:** CV2 **Definition:** Squared coefficient of variation of non-zero demand sizes **Policy:** Higher means sizes vary more **Method:** Used to classify **Item:** ADI cut-off **Definition:** 1.32 periods **Policy:** Splits regular from intermittent **Method:** Syntetos and Boylan scheme **Item:** CV2 cut-off **Definition:** 0.49 **Policy:** Splits stable from variable sizes **Method:** Syntetos and Boylan scheme **Item:** Smooth **Definition:** ADI below 1.32 and CV2 below 0.49 **Policy:** Reorder point or MRP with safety stock from normal distribution **Method:** Moving average, exponential smoothing **Item:** Erratic **Definition:** ADI below 1.32 and CV2 at or above 0.49 **Policy:** Larger safety stock, review sizes of orders **Method:** Exponential smoothing with outlier control **Item:** Intermittent **Definition:** ADI at or above 1.32 and CV2 below 0.49 **Policy:** Low min and max, order on demand where lead time allows **Method:** Croston or SBA **Item:** Lumpy **Definition:** ADI at or above 1.32 and CV2 at or above 0.49 **Policy:** Avoid stocking by formula, consider make or buy to order, review by item **Method:** SBA or bootstrapping, judgment **Item:** Croston method **Definition:** Smooths demand size and interval separately, forecast = size / interval **Policy:** Applies to intermittent items **Method:** Known to be slightly biased upward **Item:** SBA **Definition:** Syntetos Boylan Approximation: Croston multiplied by 1 - alpha / 2 **Policy:** Reduces Croston bias **Method:** Common default for spare parts **Item:** Normal safety stock **Definition:** Z x sigma x sqrt(lead time) **Policy:** Reasonable for smooth items only **Method:** Poor fit for lumpy items **Item:** Fill rate target **Definition:** Share of units met from stock **Policy:** Better target than cycle service level for slow items **Method:** n/a **Item:** Time period choice **Definition:** Weekly buckets often make ADI larger than monthly buckets **Policy:** Classify at the bucket you plan in **Method:** n/a **Item:** Minimum history **Definition:** At least 24 periods is a common working rule **Policy:** Fewer periods give unstable classes **Method:** n/a **Notes:** Cut-offs of 1.32 and 0.49 come from published research by Syntetos, Boylan and Croston and are a working convention, not a law. Reclassify items regularly because patterns change. ## Content **Intro:** ABC and XYZ tell you value and variability, but they do not tell you whether demand is intermittent. The ADI and CV2 scheme separates smooth, erratic, intermittent and lumpy demand, and each needs a different stocking and forecasting approach. **Sections:** #### How to use the reference **Body:** Compute ADI and CV2 for each item from at least two years of history, then look up the class here. ADI is the number of periods divided by the number of periods that had demand. CV2 is the squared coefficient of variation of the demand sizes, using only periods with demand. Match the result against the table and note the suggested approach. **Bullets:** - ADI = total periods / periods with non-zero demand - CV2 = (std dev of non-zero demand / mean of non-zero demand) squared - Classify at the planning bucket you use in the ERP #### Why the classes matter **Body:** Normal-distribution safety stock formulas assume demand is reasonably regular. For intermittent and lumpy items, many periods have zero demand and a few have big spikes, so the standard deviation overstates or understates risk depending on the history. Croston style methods and fill rate targets give more stable results than a mean and sigma approach. #### Example **Body:** A bearing used in 6 of 36 months has ADI of 6. If the quantities in those months are 2, 3, 2, 4, 3 and 2 the mean is about 2.7 and the standard deviation about 0.82, so CV2 is about 0.09. That makes the item intermittent, so a SBA forecast and a small min and max are more appropriate than a reorder point from monthly sigma. #### Common mistakes **Body:** Calculating CV2 with zero periods included mixes two effects and moves everything toward erratic. Using one class cut-off for all bucket sizes is another. Finally, do not turn the class into a rule without business judgment: a critical spare may deserve stock whatever the class says. **Cta:** Netray can compute ADI and CV2 for every item in your ERP history and feed the results into planning parameters with human review. **Faq:** #### What are the ADI and CV2 cut-off values? **Answer:** The widely used values are 1.32 for ADI and 0.49 for CV2. They come from research on forecasting methods for intermittent demand and mark where Croston style methods start to beat simple smoothing. #### How is this different from XYZ analysis? **Answer:** XYZ ranks items by variability using coefficient of variation of all periods. ADI and CV2 separate how often demand occurs from how large it is, which matters for slow moving and spare parts. #### Should lumpy items be stocked? **Answer:** Only after a business review. Many lumpy items are better handled by ordering against a customer need, or by a small stock backed by supplier agreements, because formula-based safety stock is unreliable. #### How much history do I need? **Answer:** At least 24 periods is a practical minimum, and more for very slow items. With a short history, a single sale can change the class, so treat the result as provisional. ## Search terms adi cv2 demand classification, intermittent demand forecasting, lumpy demand definition, croston method, syntetos boylan classification, smooth erratic intermittent lumpy, spare parts demand pattern, xyz vs adi cv2 --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json