2.5k
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill openclaw/skills --skill inventory-demand-planning- _meta.json304 B
- SKILL.md22.2 KB
Overview
This skill codifies 15+ years of demand planning expertise for multi-location retailers covering demand forecasting, safety stock optimization, replenishment planning, and promotional lift estimation. It guides method selection (moving averages, exponential smoothing, ML), ABC/XYZ segmentation, seasonal buy and markdown timing, and vendor negotiation frameworks. Use it to turn commercial plans into executable PO recommendations that balance service levels and inventory investment.
How this skill works
The skill inspects SKU-level history, seasonality, promotions, lead-time variability, and vendor constraints to recommend forecasting methods, safety stock, reorder logic, and review cadences. It applies rule-based method selection (e.g., Holt-Winters for seasonal, Croston for intermittent), computes service-level–driven safety stock (including lead-time variability), and layers promotional lift and post-promo dip adjustments on top of baselines. Outputs are actionable reorder points, order-up-to targets, and cadence recommendations by vendor tier.
When to use it
- Forecasting demand for steady, trending, seasonal, intermittent, or promotion-driven SKUs
- Setting safety stock accounting for lead-time variability and target service levels
- Designing replenishment logic (ROP, Min/Max, EOQ, periodic review) and vendor review cadence
- Estimating promotional lift, cannibalization, and post-promo dips for PO planning
- Managing seasonal buys, markdown timing, and end-of-season liquidation planning
Best practices
- Classify SKUs with ABC (value) and XYZ (predictability) on de-seasonalized demand before policy-setting
- Validate model parameters and service-level choices on out-of-time holdout periods, not on training data
- Optimize alpha/beta/gamma for exponential smoothing via holdout and monitor tracking signal (±4 trigger)
- Adjust safety stock for lead-time variability using demand and lead-time variance; bootstrap for intermittent demand
- Segment vendors by spend tier and align review frequency (weekly for A, bi-weekly for B, monthly for C)
Example use cases
- Automating weekly replenishment for AX items with tight safety stock and vendor-case-pack rounding
- Forecasting seasonal launches with initial 60–70% buy and reserve 30–40% for early-season reorders
- Estimating lift and post-promo dip for TPR + display promotions to size purchase orders and avoid overstocks
- Applying Croston or SBA for low-velocity SKUs to prevent inflated safety stock from normal-distribution formulas
- Selecting ML (LightGBM/XGBoost) when >1,000 SKUs and external regressors exist, with quarterly retraining
FAQ
Use ML when you have large cross-SKU history (1000+ SKUs × 2+ years), reliable external regressors, and engineering support; otherwise prefer simpler, interpretable methods with robust holdout validation.
How do I handle new products with no history?
Profile analogous items (3–5 similar SKUs), apply their demand variability as a proxy, and add a 20–30% buffer for the first 8 weeks while collecting own-data.