A Primer on Probabilistic Forecasting in Amazon Inventory

With Amazon, you will find that even the smallest details can determine whether you make a profit or lose money. As a result of the limited ability to look further ahead, many Amazon vendors operate in short-term bursts. Making actual sales is more appealing than forecasting demand, so it’s often ignored. 

Even small businesses need to anticipate their needs in advance. As part of its Amazon Brand Analytics, Amazon has launched Probability-Level Demand Forecasting to help their vendors and sellers make well-informed decisions about their inventory demand. As a result, vendors must make critical decisions to meet customer demands in a short period of time. 

I’ll review Amazon’s Probability-Level Demand Forecasting, its rules of thumb, what it means for vendors, how to overcome the challenges and how analytics can help close the gaps. Want to learn more about forecasting techniques with inventory forecasting software Consult with our experts at Inventooly. 

How does Amazon’s Probability-Level Demand Forecasting work? 

Probability Level Demand Forecasting enables vendors to view their inventory levels in the near future by formulating a forecast. Based on Amazon’s calculations, it is the probability of customers buying a certain item every week.

The forecast includes four probabilities, including: 

  • P70. A 70% likelihood of consumer demand being equal to or below the units shown is estimated by Amazon. Customers may purchase more units by 30%. 
  • P80. Amazon is likely to purchase (or less than) the level of Demand indicated (80%) and has a 20% probability of purchasing more. 
  • P90. 90% of consumers will demand the amount indicated each week, and 10% of Amazon will order more. 
  • Mean. Suppliers who import or produce overseas are given a waiting period of five to six weeks to deliver products. The likelihood percentage is usually in the range of 50 to 60%. 

There are several factors to consider when choosing a P-Level for your business, including the nature of the company and the products sold. In the case of a P70, you would have some protection against stockouts, but chargebacks might also be a worry. In addition to the P90 forecast, you can also keep too much inventory to ensure orders are always fulfilled. Although it has its own benefits, it is not without its own cons, such as expiring products and storage costs.

Are these probabilities influencing sales in any way?

Online shopping is increasingly popular in the US, with two-thirds of Americans doing so. These consumers flocked to Amazon to purchase household staples after the COVID-19 pandemic pushed them out of their comfort zones. Amazon is scrambling to resolve its disrupted supply chain with amazon warehouse system as many sellers didn’t have enough stock and missed out on sales.

Many retailers and producers in China have turned to e-commerce platforms in response to hoarding and stockouts. After the outbreak, their supply chains recovered by 50% only a few weeks later. Profitability won’t be boosted by predicting demand using empty shelves. Companies can better manage risks and allocate resources if they predict what, where, and when demand will occur.

Vendor impact of P-level demand forecasting

It’s probably not the biggest, but it’s probably the only disclaimer about P-level demand forecasts: they’re just estimates, not guarantees. Vendors should be aware of the risks involved. Amazon does not promise to purchase any items based on a demand forecast. 

All the forecasts are based on 26-week rolling estimates. In an effort to increase efficiency, Amazon intentionally stopped making safety stock forecasts. The company has also stopped including safety stocks in the demand forecast. A vendor’s own needs can be used to estimate their lead times (VLTs).

The challenges associated with implementing probabilistic demand forecasting 

Vendor-driven actions mean Amazon has more freedom to use the tools and technologies available to them at their disposal. It also means a greater amount of responsibility.  

Despite their excellent start, Amazon’s forecasts can be shortsighted. For a couple of days or even weeks, you probably don’t need to worry about how many products you must carry to meet your sales. Consumer demand is what Amazon forecasts for its P-Level Demand. Advertising, promotions, and seasonality that directly affect your inventory are not covered.  

Likewise, the demand forecast is the best estimate for a week or two at most. When you’re a bigger retailer or a brand with operations and manufacturing plans far in advance, it will be difficult to compile long-term estimates and year-over-year comparisons. Getting data into action can also be a challenge due to integration issues. Some of Amazon’s major forecasting tools are not compatible with existing technologies. 

Vendors may not be able to see or know all the variables that Amazon uses in its formula. You may encounter difficulties integrating Amazon’s data with your own tools and systems, as these variables may be crucial but cannot be accounted for by your systems or tools.

Analyzing inventory data can help make better decisions. 

Most companies have access to the data they need to generate more accurate and better forecasts of demand. There is a problem with your own data: You may know it is gold, but it is cumbersome and confusing to sort through, let alone use it.

Analytical tools can help vendors prepare for shorter product life cycles, changing market conditions, and volatile demand. Vendors on Amazon can use analytics to determine what probability level is most appropriate for their needs and also forecast their sales on Amazon. With a 70% close rate, analytics can help keep the remaining 30% under tight observation and eliminate guesswork.  

Analyzing data will help you manage the P-Level forecasts on Amazon and understand your responsibilities as a vendor. Therefore, you can meet customer expectations while preventing problems from arising in the future.


Amazon launched its probabilistic demand forecasting to help sellers gain a better understanding of the demand from their customers. This article addresses both the pros and cons of the program. This program is only intended for Amazon sellers who want forecasts, not actual results. You don’t want to be faced with an overstocked or an understocked inventory, so be sensible when making decisions based on these forecasts. 

We hope this article helps you in understanding what probabilistic forecasting means for Amazon sellers and how they can leverage it to their advantage. If you are looking for inventory management Amazon services for your Amazon seller account, check out Inventooly. 

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button