Amazon Inventory Forecasting: Demand Planning for FBA and Multi-Channel Sellers

Most Amazon sellers forecast inventory the same way: look at last month’s sales, assume next month will be similar, and reorder roughly the same quantity. This works until it does not. A Lightning Deal doubles your sales for 48 hours and blows through your safety stock. A competitor runs out, your organic rank jumps, and demand increases 40% for three weeks. PPC budget changes shift your sales velocity overnight. Seasonal demand ramps faster than you expected.

Amazon inventory forecasting is fundamentally different from forecasting for your own website or a brick-and-mortar store. The marketplace environment introduces demand volatility that does not exist in channels you control. A 2024 Feedvisor analysis of 5,000 Amazon sellers found that sellers who relied on simple trailing averages for restock decisions experienced stockout rates 2.3x higher than sellers using weighted or seasonal forecasting methods.

This guide covers the forecasting approaches that work for Amazon, how to estimate FBA-specific lead times accurately, and how to handle the demand spikes that make Amazon different. For the operational framework around acting on your forecasts, see the Amazon inventory management hub.

Why Amazon forecasting is different

BSR volatility and competitive dynamics

Your Best Seller Rank (BSR) on Amazon is relative to every other product in your category. When a competitor stocks out, your organic visibility increases and sales spike without any action on your part. When they come back in stock, your sales drop. These competitive-driven fluctuations are invisible in your own sales data unless you are tracking competitor availability.

A product selling a steady 30 units per day can jump to 50 when two competitors stock out during a shipping delay from China, then drop back to 25 when they return with aggressive pricing. A forecast based on last month’s average of 35 units per day misses both the spike and the correction.

PPC-driven demand shifts

For many Amazon sellers, 30 to 60% of total sales come through Sponsored Products and Sponsored Brands advertising. Changes to your PPC budget, bid strategy, or keyword targeting directly change your sales velocity. If you increase PPC spend by 40% to push a product launch, your demand forecast needs to reflect that planned change, not just historical data.

The reverse is also true: reducing PPC spend or pausing campaigns during a cash flow crunch immediately reduces sales, which means your existing inventory will last longer than your forecast predicts. Overstocking follows.

Seasonal surges are steeper on Amazon

Amazon’s Q4 peak is more pronounced than most other channels. According to Amazon’s 2024 holiday season recap, FBA order volume during the week of Black Friday and Cyber Monday was 3.2x the average weekly volume from the prior 90 days. For individual sellers in gift-oriented categories, the multiplier can be 5x to 8x.

The ramp-up is also faster on Amazon than in physical retail. Sales might increase 20% in the first week of November, then 60% in the second week, then 200% during Cyber Week. A linear seasonal adjustment misses this hockey-stick pattern.

Forecasting methods that work for Amazon

Simple moving average

Take the average daily sales over a set lookback period (typically 30, 60, or 90 days).

Formula: Forecast daily demand = Sum of units sold over N days / N

When it works: Stable products with low seasonality and consistent PPC spend. Products that have been selling for 6+ months with low variance.

When it breaks: Any time demand has shifted recently. A 90-day average includes stale data that dilutes recent trends. If your sales jumped from 20 to 35 units per day over the past two weeks, a 90-day average might still show 24.

Weighted moving average

Assign higher weights to recent periods and lower weights to older periods.

Formula: Forecast daily demand = (W1 x recent period sales) + (W2 x prior period sales) + (W3 x oldest period sales)

A common weighting for Amazon sellers:

• Last 7 days: 50% weight
• Days 8 to 30: 30% weight
• Days 31 to 90: 20% weight

Example: If a product sold an average of 40 units/day last week, 30 units/day in the prior 3 weeks, and 25 units/day in the two months before that:

Forecast = (0.50 x 40) + (0.30 x 30) + (0.20 x 25) = 20 + 9 + 5 = 34 units per day

This captures the recent upward trend that a simple average would underweight. Weighted averages are the best default method for most Amazon sellers because they balance responsiveness to recent changes with enough historical data to smooth out single-week anomalies.

Seasonal decomposition

For products with predictable annual patterns, layer a seasonal index on top of your base forecast.

How to build a seasonal index:

1. Calculate average monthly sales for the past 12 to 24 months
2. Calculate the overall monthly average across the entire period
3. Divide each month’s average by the overall average to get a seasonal index

Month	Avg units sold	Overall monthly avg	Seasonal index
January	450	600	0.75
February	420	600	0.70
March	510	600	0.85
April	540	600	0.90
May	570	600	0.95
June	600	600	1.00
July	630	600	1.05
August	660	600	1.10
September	600	600	1.00
October	720	600	1.20
November	900	600	1.50
December	600	600	1.00

Application: Multiply your base forecast (from weighted moving average) by the seasonal index for the target month. If your weighted average says 34 units/day and you are forecasting for November (index 1.50), your adjusted forecast is 51 units/day.

Seasonal decomposition requires at least 12 months of sales data to be reliable. For newer products, you need to estimate seasonality from category-level trends or comparable ASINs.

Estimating FBA lead time accurately

Your forecast only matters if you act on it early enough. FBA lead time is the gap between deciding to restock and having units available for sale at Amazon, and it is longer than most sellers realize.

FBA lead time components

Component	Typical range	Variable factors
Supplier production time	5 to 45 days	Product complexity, supplier location, order size
Shipping to your location	1 to 30 days	Domestic vs. international, shipping method
Prep and labeling	1 to 3 days	Volume, whether you prep yourself or use a prep center
Transit to Amazon FC	2 to 7 days	Carrier, distance, shipping method (SPD vs LTL)
Amazon receiving and processing	3 to 21 days	Season, FC congestion, shipment type

Total lead time range: 12 to 106 days

The most variable component is Amazon’s receiving time. During Q4 or Prime Day prep periods, Amazon’s fulfillment centers get backed up and receiving can stretch to 3 weeks. Outside peak periods, receiving typically takes 5 to 7 business days for standard shipments.

Use your actual lead time data, not averages. Pull your last 6 to 8 inbound shipments from Seller Central and note the time from shipment creation to “Available” status for each. Use the 80th percentile (not the average) as your lead time estimate. This means 80% of your shipments will arrive within your planned window, giving you a realistic buffer.

Reorder point formula

Reorder point = (daily demand forecast x total lead time in days) + safety stock

If your forecast says 34 units per day, your total lead time is 28 days, and your safety stock is 7 days of cover:

Reorder point = (34 x 28) + (34 x 7) = 952 + 238 = 1,190 units

When your FBA available quantity drops to 1,190 units, you need to trigger a new order. For the full breakdown on safety stock calculations and how to prevent stockouts, see the stockout prevention guide.

Factoring in promotions and Lightning Deals

Promotions create demand spikes that should not be blended into your baseline forecast. Treat them as discrete events with their own demand estimates.

Lightning Deals

Amazon Lightning Deals typically increase unit velocity by 3x to 10x during the deal window (usually 4 to 12 hours). A product selling 30 units per day might sell 100 to 200 units during a Lightning Deal.

Before scheduling a Lightning Deal, calculate the inventory impact:

• Expected deal volume: (Normal daily units / hours in a day) x deal duration in hours x velocity multiplier
• Post-deal halo: Sales typically run 10 to 20% above baseline for 3 to 5 days after a successful deal due to improved BSR and organic visibility

Make sure you have enough FBA inventory to cover the deal volume plus the post-deal halo plus your normal ongoing demand. Running out of stock during a Lightning Deal wastes the deal fee and trains the algorithm that your deal performance is poor.

Coupons and price promotions

Coupons and percentage-off promotions create more moderate but longer-lasting demand increases. A 20% coupon on a competitive product typically increases daily velocity by 40 to 80%, depending on the category and the size of the discount relative to competitors.

The key mistake sellers make is running a promotion without adjusting their restock plan. You end up with a successful promotion that drives a stockout two weeks later because you sold through inventory faster than planned.

Data sources for Amazon forecasting

Amazon Seller Central reports

• Business Reports (Detail Page Sales and Traffic): Daily and weekly unit sales, sessions, and conversion rates by ASIN. This is your primary data source for demand history.
• Inventory Dashboard: Current FBA quantities, inbound shipments, and Amazon’s own restock recommendations (useful as a sanity check, not a primary input).
• FBA Inventory Age report: Shows inventory age distribution, helping you identify slow movers before surcharges hit.

Third-party tools

Tools like Jungle Scout, Helium 10, and Keepa provide competitor sales estimates, BSR history, and category-level demand trends. The accuracy of competitor sales estimates varies, but they are useful for understanding whether demand shifts are specific to your ASINs or category-wide.

Your own operational data

The most underused data source is your own sales data across all channels. If you sell the same product on Amazon, Shopify, and eBay, your total demand signal is stronger than any single channel. An Amazon inventory management platform that connects all your channels gives you a unified demand picture that produces better forecasts than Amazon-only data.

Common forecasting mistakes

• Using a single lookback period for all SKUs: Fast-moving SKUs need shorter lookback periods (14 to 30 days) to stay responsive. Slow-moving SKUs need longer periods (60 to 90 days) for statistical reliability.
• Not separating organic and PPC-driven sales: When you pause PPC, only your organic baseline remains. Know the split so you can forecast the impact of ad spend changes.
• Forecasting in monthly buckets instead of weekly: Monthly forecasts miss within-month demand patterns. A product that sells 70% of its monthly volume in the first two weeks needs a different restock cadence than one that sells evenly throughout the month.
• Ignoring Amazon’s fee calendar in restock decisions: Your forecast might say to send 90 days of inventory, but sending that much in September means paying Q4 storage rates on all of it. Factor storage cost timing into your shipment sizing.
• Treating new product launches like established SKUs: New products have no demand history. Use category benchmarks and conservative estimates for the first 60 to 90 days, then switch to data-driven forecasting once you have reliable sell-through data.

Quick Reference

• Sellers using weighted averages experience 2.3x fewer stockouts than those using simple trailing averages (Feedvisor 2024)
• Recommended weighting: 50% last 7 days, 30% days 8 to 30, 20% days 31 to 90
• FBA total lead time ranges from 12 to 106 days depending on supplier location and Amazon receiving speed
• Use the 80th percentile of your actual receiving times, not the average, for lead time planning
• Amazon Q4 peak volume is 3.2x the average weekly volume from the prior 90 days (Amazon 2024)
• Lightning Deals typically increase velocity 3x to 10x during the deal window
• Reorder point = (daily demand x lead time) + safety stock
• Forecast weekly, not monthly, to catch intra-month demand patterns
• Separate organic and PPC sales in your data to model ad spend changes accurately
• New products need 60 to 90 days of sales data before data-driven forecasting is reliable

Forecasting method	Best for	Lookback period
Simple moving average	Stable products, low variance	30 to 90 days
Weighted moving average	Most Amazon SKUs (default)	7/30/90 days with weighting
Seasonal decomposition	Products with annual patterns	12 to 24 months
Promotion overlay	Lightning Deals, coupons	Event-specific estimates