A new SKU can look like a growth opportunity right up until it ties up six months of cash or runs out two weeks after launch. To forecast demand for new products, Amazon sellers cannot rely on a sales history that does not exist. They need a disciplined way to combine comparable products, launch assumptions, supply constraints, and early performance signals into a buying decision.
The goal is not to produce one perfect number. It is to make a defensible initial purchase order, know what would change that decision, and react before a stockout or overbuy becomes expensive.
Why new-product forecasts fail
Most bad launch forecasts have less to do with math than with bad inputs and false certainty. A team may copy the sales volume of its best seller, assume every variation will perform equally, or let an optimistic advertising plan become the forecast. The result is often excess units on a slow launch and an out-of-stock event on the product customers actually want.
New items also have a cold-start problem. A mature SKU has order history, seasonality, promotion effects, and repeatable sales velocity. A new product has none of that. It may have a similar parent product, but it can still appeal to a different customer, price point, keyword set, or use case.
Amazon adds another complication: sales are not always the same as demand. If a SKU went out of stock during its launch, its recorded sales understate what customers would have bought. If inventory was constrained in FBA or ads were paused because coverage was low, the sales curve may be artificially capped. Treat stock-constrained sales as incomplete evidence, not proof of weak demand.
Forecast demand for new products from the closest evidence
The most reliable starting point is an analog forecast. An analog is not simply a product in the same catalog. It is the existing SKU or group of SKUs that most closely matches the new item’s demand drivers.
For an Amazon brand, the best analog typically shares several characteristics: customer type, price band, category, season of launch, product purpose, and expected advertising support. A refill size may resemble the original product’s customer behavior. A premium version may need a lower conversion assumption but a higher average order value. A new color can follow the demand pattern of the parent item, while a new bundle may behave more like a separate offer.
Use more than one analog when possible. One comparable product can carry an odd promotion, stockout, review spike, or pricing change that makes it misleading. A small set of comparable SKUs gives you a range and forces a more honest conversation about uncertainty.
Start with weekly unit demand rather than monthly totals. Weekly planning exposes the timing that matters for replenishment: when demand may accelerate, how fast coverage is being consumed, and whether supplier lead time leaves room to react.
Adjust the analog instead of copying it
An analog establishes the baseline. Then adjust that baseline for the differences that matter. If the new product launches at a 20% higher price, do not assume it will sell the same number of units. If it enters a highly seasonal period, account for the lift. If the launch will receive less ad spend or fewer reviews than the analog had, reduce the initial expectation.
Keep those adjustments visible. A forecast built from clear assumptions is easier to challenge and improve than a spreadsheet number no one can explain. The planning team should be able to say, in plain language, why the base case is 80 units per week instead of 140.
Plan a range, not a single launch number
For a mature SKU, a forecast may narrow after enough history accumulates. For a new product, uncertainty is part of the model. Build three demand cases: conservative, base, and upside.
The conservative case should represent a slower-than-expected launch, not a disaster scenario. The base case reflects the most likely outcome based on analogs and the planned go-to-market effort. The upside case reflects stronger conversion, better ad efficiency, or faster organic traction.
These cases should drive inventory actions, not sit in a planning deck. For example, the initial order may cover the base case through the next replenishment opportunity, while an earlier supplier commitment or conditional follow-on PO covers the upside case. This reduces exposure if demand is soft without forcing the business to wait too long if the product moves quickly.
The right approach depends on supplier economics. A long lead time, high minimum order quantity, and slow purchase cadence may require more initial coverage. A supplier that can replenish quickly in smaller quantities allows a leaner opening position. Do not treat safety stock as a fixed percentage. It should reflect demand uncertainty, lead-time variability, and the financial cost of being wrong.
Put supply constraints into the forecast decision
Demand forecasting alone does not answer how much to buy. The purchase decision must account for available inventory, units already in transit, open purchase orders, supplier lead time, minimum order quantities, and the date inventory can actually support sales.
A common mistake is to forecast 12 months of demand and then place a 12-month order. A long planning horizon is useful because it shows future cash requirements and potential stock gaps. It does not mean every unit should be purchased immediately.
Instead, calculate coverage against each demand scenario. Ask when the current order would run out under base demand and upside demand. Then work backward from the required in-stock date using production, transit, and receiving buffers. If the reorder date has already passed in the upside scenario, the launch plan needs a decision now: increase the opening buy, secure capacity with the supplier, or adjust advertising to protect availability.
This is where a purpose-built planning system earns its place. Inventory Optimizer can bring sales velocity, inventory, in-transit POs, and multichannel demand into one forecast so teams are not manually stitching together Amazon, Shopify, and supplier spreadsheets every Monday.
Use early launch data without overreacting
The first few days of sales are useful, but they are noisy. A launch can spike from an email campaign, early review activity, a coupon, or a brief period of aggressive advertising. Ordering against that first spike can create expensive overstock.
Set a review cadence before launch. In the first two weeks, monitor daily signals to catch availability issues and major surprises. For replenishment changes, favor a rolling seven-day or 14-day view unless daily demand is consistently stable. By weeks three through six, compare actual sales against each forecast scenario and begin replacing analog assumptions with the product’s own trend.
Watch conversion rate, ad-driven orders, organic orders, price changes, and inventory availability alongside unit sales. Sales velocity rising while ad spend rises may be expected. Sales velocity rising after ad efficiency improves may justify an upside revision. Sales velocity falling while inventory is unavailable tells you very little about real demand.
Avoid revising the forecast every time a daily number moves. Define triggers in advance. For example, a review may be required when two consecutive weeks exceed the base case by 25%, when sell-through falls below the conservative case, or when projected coverage falls inside supplier lead time.
Connect inventory and advertising decisions
Advertising should accelerate products that can stay in stock. It should not create a demand spike that the replenishment plan cannot support. When projected inventory coverage becomes thin, slowing ad spend can preserve organic rank, protect customer experience, and prevent a stockout before the next inbound inventory arrives.
That does not mean turning ads off at the first sign of risk. It means matching spend to coverage and margin. A product with a reliable inbound PO arriving soon may support continued acquisition. A SKU facing a long supply gap needs a more defensive plan. The forecast, reorder plan, and ad strategy must use the same view of available and incoming inventory.
Build a repeatable launch forecast workflow
The strongest teams make new-product forecasting a standard operating process. Before the first PO, select analogs, document assumptions, build conservative, base, and upside demand cases, and map every case to inventory coverage and reorder timing. After launch, compare actual performance on a scheduled cadence and record what the original assumptions got right or wrong.
Over time, this creates a library of launch patterns by product type, price tier, season, and channel. That history becomes more valuable than any one forecast because it improves the next buying decision.
A new product will always carry uncertainty. The commercial advantage comes from seeing that uncertainty early, sizing inventory deliberately, and having a clear action ready when real demand starts to show itself.
