Forecast demand through seasonal swings
Peak demand isn't a mystery. It's last year's curve times this year's growth. The move: build a per-month forecast from your own history, work backwards from lead times to set order deadlines, size safety stock by how bumpy each SKU is, and plan the post-season sell-down.
AI can turn a sales export into the forecast. You supply the growth rate and the lead times, because only you know them.
Fastest path: one prompt, end to end
🤖 AI prompt — paste into ChatGPT / Claude
You are a demand-planning analyst. Use MY data only.
I will paste my last 12 to 24 months of sales by month (and by SKU if I have it).
This year's growth rate vs last year: [e.g. plus 30 percent]
Supplier lead time (order to arrival): [e.g. 45 days]
Cash or storage constraints: [any limits]
Do this:
1. Build a month-by-month forecast: take last year's monthly shape (the seasonal curve) and scale it by my growth rate. Show the peak month(s) and the units expected.
2. Work backwards from lead time to give me the ORDER-BY date for peak stock, so it arrives before demand hits, not during.
3. Size safety stock per SKU by its variability: steady sellers get a thin buffer, spiky or unpredictable ones get a larger one. Explain the logic simply.
4. Give a post-season sell-down plan: how much to hold for the tail, and when to start marking down leftovers so I don't carry them into next season as dead stock.
Do not invent numbers, use only what I paste. If something is missing, ask.
Output: monthly forecast table + order-by dates + per-SKU safety stock + sell-down plan.
Or do it in 4 steps
- Start from last year's shape, then scale. Export monthly sales for the last 12 to 24 months, that curve IS your seasonality. Multiply each month by this year's growth rate to get the forecast. This beats a flat 'order more for Q4' because it tells you which months and how much.
- Work backwards from lead time to an order deadline. If peak lands in December and your supplier needs 45 days, your order-by date is mid-October, not December. Missing the deadline means stocking out at the worst possible time. Set the deadline first, then commit the buy.
- Size safety stock by variability, not a flat percent. A steady seller needs a small buffer; a spiky, hard-to-predict SKU needs a bigger one to avoid a stockout. Give your erratic items more cover and your predictable ones less, so cash isn't tied up evenly across the board.
- Plan the sell-down before the season ends. Decide up front how much tail stock to hold and when to start marking it down. Leftover seasonal stock that misses its window turns into dead stock, so schedule the markdown while demand still exists rather than discovering the pile in January.
Worked example (labeled): last November sold 400 units, December 900, January 300; this year is tracking plus 30 percent. Forecast: Nov 520, Dec 1,170, Jan 390.
Supplier lead time is 45 days, so the peak order must be placed by mid-October to land before December. A steady core SKU gets a 10 percent buffer; a trendy variant that swung wildly last year gets 25 percent.
Hold 15 percent of December's stock for the January tail, then start markdowns January 10 so nothing carries into spring.
Anchor the forecast to your own curve, set the order-by date from lead time, and schedule the sell-down early; use your inventory-planning note as the base method this builds on.