Lunar New Year 2027 falls on 6 February, and the orders that decide your festive sales and Q1 cash are being placed right now, in early November, before you know your year-end results. Supplier cut-offs are arriving while Black Friday, 12.12, and Christmas are still selling, and anything not ordered in time won’t arrive in time.
The cause is rarely a lack of effort. Year-end and Lunar New Year planning sit in separate spreadsheets, sales and purchasing data don't talk to each other, and finance finds out about the damage in March. IHL Group's 2026 Inventory Distortion Study estimates that inventory distortion, the combined cost of stockouts and overstocks, costs retailers about US$1.7 trillion a year worldwide, equal to 6.2% of global retail sales.
This guide explains why this window is so hard, how AI demand planning works when it is connected to your ERP, and what to do in the next 14 days.
Why year-end and Lunar New Year are hard to plan
Year-end and Lunar New Year give you two peaks back-to-back with a decision deadline in the middle.
- The deadline arrives before the data. You commit capital for February based on incomplete November and December sales.
- The calendar moves. Lunar New Year shifts every year. A forecast built on "same period last year" can misalign gifting, grocery and apparel demand by one to two weeks, the difference between full price and a loss-making clearance.
- Promotions pollute your history. 11.11, Black Friday and 12.12 spikes reflect discounting as much as demand. Without clean promotion data, forecasts treat a 30%-off spike as organic growth and over-order next year.
- Seasonal stock has no second chance. Gift sets, red packets, snacks, decor and new-season apparel lose value fast after the festival. A bad call creates dead stock, not just slow stock.
The result is two-sided: fast sellers run out while cash sits trapped in slow ones, and the same planning process causes both.
The real problem: disconnected data
A failed Lunar New Year plan is usually blamed on bad forecasting, but the forecast is only a symptom. In a typical mid-market retailer, sales data sits in the POS or e-commerce platform, stock in a warehouse system or spreadsheet, purchase orders and lead times in email, the promotion calendar in a marketing slide deck, and budget with finance, who see commitments only after POs are approved. Each team builds its own view of how much to order.
An integrated ERP changes this. When sales, inventory, purchasing, warehouse and finance run on one platform, such as a retail ERP with POS, replenishment and multi-outlet management, the forecast reads from live data and writes back into purchasing and stock decisions. (For a broader look, see how ERP helps the retail industry.) That is what turns a forecast from a number into an action:
- Shared inputs: One system of record ends the "whose number is right?" debate between sales, purchasing and finance.
- Forecast to replenishment: Once demand is forecast at SKU-store level, a joined-up supply chain management module can turn it into suggested purchase orders and transfers, using current stock, open POs and supplier lead times.
- Multi-location visibility: A real-time inventory management system lets you move stock from an overstocked store to a fast-selling one instead of reordering.
- Landed cost and margin: With freight, currency and duties in the same system, planners can see the margin impact of ordering early versus late, or sea versus air.
- Financial control: Integrated financial management and management control show open-to-buy, working capital and cash flow in November, not March.
What AI demand planning does
McKinsey research indicates that AI-driven forecasting can reduce forecast errors by 20% to 50% compared with traditional methods, and cut lost sales from product unavailability by up to 65%. Results vary with data quality and implementation. In practice, AI does four things a spreadsheet can't:
- Uses more signals: It looks at sales by product, store, and channel, plus Lunar New Year's changing date, promotion size and length, price changes, weather, and whether a product was actually in stock, so days with no stock aren't mistaken for low demand. It also accounts for new products taking sales from existing ones.
- Works at SKU-store level: It can show that one store will sell out of festive gift sets in week two while another has too many, so you can buy and transfer stock with precision.
- Recalculates daily: If a product sells 40% faster than expected in the first week of 12.12, you still have time to reorder or move stock.
- Flags exceptions: It points out products selling above or below forecast, underperforming stores, and stockout risks, so planners spend their time making decisions instead of updating spreadsheets.
What AI doesn't know is the competitor opening next door, the supplier who is always a week late, or management's decision to push a range. That context comes from your people, so combine the model with structured human review for high-value and new items.
The Lunar New Year and year-end playbook
Work backwards from 6 February.
For each supplier, document the production cut-off, shipping time, customs clearance and receiving buffer, then set an internal last-order date and treat it as a hard deadline.
Segment your SKUs.
Core items are replenished continuously. Seasonal, one-shot items need conservative, scenario-based planning. Promo-driven items need promotion flags and no extrapolation from spike periods.
Plan seasonal buys in scenarios, with exit routes.
Build low, expected and high demand cases, and decide how much of the high case to commit up front versus hold for a second order. If your supplier can't reorder in time, order cautiously and accept some lost sales over dead stock. Before ordering, decide what happens to leftovers: bundle with core items, sell online, move to outlets, or hold for next year if the product doesn't date.
Set safety stock by velocity and variability.
A uniform buffer is one of retail's most expensive habits. Volatile fast sellers and unreliable suppliers need more; slow movers need little. McKinsey reports AI-driven planning can lower overall inventory by 20% to 30% in well-run deployments, largely by right-sizing buffers.
Load the promotion calendar and plan markdowns early.
Every campaign, discount depth and channel should be in the system so the forecast can separate organic demand from promotional lift. Agree when and how deeply to discount lagging seasonal items before the peak, not in the post-festival clearance when every competitor is discounting.
Rebalance weekly.
Review sell-through by store and raise transfer orders early. Rebalancing is often the cheapest way to cut stockouts.
| Period | Focus | Key actions |
| Early–mid November | Commit | Confirm supplier cut-offs, finalise Lunar New Year seasonal orders, review 11.11 data |
| Late November–December | Black Friday, 12.12, Christmas | Daily sell-through checks, weekly rebalancing, refresh Lunar New Year forecasts |
| Early January | Final top-ups | Use remaining supplier windows; air freight only for proven fast sellers |
| Mid-January to 6 February | Peak | Daily monitoring, store transfers, planned markdowns |
| After the festival | Clearance | Execute exit routes, review forecast accuracy, document lessons |
Common mistakes to avoid
- Poor data quality. Duplicate SKUs, inconsistent units, missing promotion flags and stockout days recorded as "zero sales" distort forecasts. Cleaning master data is unglamorous but often the highest-return step.
- Blind trust in the model. A forecast is a starting point; review top revenue items and products with no history.
- Forecasting at the wrong level. A category forecast looks tidy and hides everything that matters. Plan at SKU-store level.
- Ignoring supplier reality. A perfect forecast fails if the supplier can't deliver. IHL Group's research consistently ranks supplier issues among the biggest drivers of stockouts and overstocks.
- Waiting for the perfect system. Start with one category or a few stores, prove the value, then expand. Even moving up one maturity level before Lunar New Year, from spreadsheets to a connected system that links purchasing to stock visibility, will improve this season's outcome.
How to measure whether it's working
Review these weekly and compare them with last year's season: SKU-level forecast accuracy and bias, stockout rate on fast movers, seasonal sell-through, days of stock on hand, markdown percentage, working capital tied up versus plan, and supplier on-time delivery.
Your 14-day quick-start checklist
- Confirm supplier cut-offs and set internal order deadlines.
- Export two to three years of sales history, flagging promotions and stockout days.
- Clean the SKU master (duplicates, units, supplier links) and tag SKUs as core, seasonal or promo-driven.
- Run a SKU-level forecast for top categories and compare it with your current plan.
- Build low, expected and high scenarios for top seasonal items.
- Agree markdown and clearance rules before final orders.
- Share the promotion calendar with planning and purchasing.
- Set a weekly review of sell-through, rebalancing and forecast variance, and brief finance on expected commitments and cash timing.
Frequently Asked Questions
What is AI demand planning for retail?
→ It is the use of machine learning to forecast how much of each product a retailer will sell, by store and time period, using sales history plus promotions, seasonality, pricing and holidays. It updates as new data arrives.
How much can AI improve forecast accuracy?
→ McKinsey research indicates errors can fall by 20% to 50% and lost sales from unavailability by up to 65%. Actual results depend on data quality, tool fit and whether your team acts on the forecast.
Why does ERP integration matter?
→ An ERP holds sales, stock, purchasing and finance data in one place, so the forecast can turn directly into suggested purchase orders, transfers and cash-flow views.
When should I place Lunar New Year orders?
→ Work backwards from each supplier's production and shipping cut-off, plus a customs and receiving buffer. Many retailers must commit before year-end, so confirm exact dates now.
How much safety stock should I hold?
→ There is no universal figure. Set it per SKU: higher for volatile fast sellers and unreliable suppliers, lower for slow movers. For one-shot seasonal items, use low, expected and high scenarios instead of a flat buffer.
How much data do I need, and is this only for large retailers?
→ Two to three years of clean sales history is ideal, but patchy data is fine to start with top categories. Small and mid-sized retailers can begin with one category or a few stores.
Conclusion
Excess stock, missed sales and a cash crunch in March are not inevitable. They are the predictable result of disconnected data, coarse forecasts and late decisions. Connect sales, stock, purchasing and finance data, forecast where decisions actually happen, and commit early with clear scenarios and exit plans. AI demand planning is a powerful way to get there, but it works best on clean data and an integrated system, with experienced people reviewing what matters most.
Ready to get your Lunar New Year stock, right? Book a demo and see how AI-driven demand planning and inventory control work for your stores.
Sources
- IHL Group, 2026 Inventory Distortion Study: global inventory distortion of about US$1.7 trillion, equal to 6.2% of global retail sales.
- McKinsey & Company, "Succeeding in the AI supply-chain revolution": forecast error reductions of 20% to 50%, up to 65% fewer lost sales, and inventory reductions of 20% to 30%. Outcomes vary by business, data quality and implementation.
