Inventory management is a task that is directly linked to business not only for EC sites but also for retailers and manufacturers. If the appropriate inventory is always managed, it is possible to reduce the loss in terms of profits due to useless discount sales and disposal due to excess inventory, and to reduce the opportunity loss in terms of sales due to shortages. I can.
Recently, many products such as automatic management tools and ordering systems using AI and machine learning have been released, and the number of companies that have introduced them is increasing. Compared to the situation where inventory management relied solely on manpower, there is a possibility that a certain effect can be expected by reducing unmanaged areas due to personal mistakes and resource shortages.
However, introducing an AI system does not mean that all issues will be resolved. Even after introduction, there are many situations where human judgment and flexibility are required in the actual operation site.
In this article, as one of the themes of the [EC business growth x AI] series, we will introduce the points of inventory management using AI demand forecasting in EC site management.
Author profile
CEO of Colabofact / CDO of Bsidefunny Co., Ltd.
Mr. Yusuke Komatsu Yusuke Komatsu
Joined Members Co., Ltd. after graduating from Hosei University. Six and a half years of experience in website construction / operation direction and project management for major domestic financial institutions and manufacturers. Engaged as manager of the company. Joined Amazon Japan GK since 2015. After two years of experience as a site merchandiser in digital marketing, from strategy to execution and analysis, as a buyer, experienced business strategy, P/L management, demand forecasting, and product planning in the category in charge. Participated in Principle Co., Ltd. since March 2019. Launched a dedicated sales department, and since October 2020, has been in charge of launching the company's DX business department and responsible for business operations. In May 2021, he joined Bsidefunny Co., Ltd. as a partner and CDO. Independent as a sole proprietor from June 2021.
[EC Business Growth x AI] Series No. 3 INDEX
- Why do we need inventory?
- How much inventory should I keep?
- How do you forecast demand?
- Is it possible to entrust inventory management to AI?
------------------------------------------------
▼[EC Business Growth x AI] Click here for previous articles in the series
[Part 1] What are the specific measures to grow your EC business? Explain the points to keep winning in the expanding EC market!
[Part 2] The success or failure of ASKUL Corporation, which handles more than 10 million items, uses AI behind the EC
------------------------------------------------
1. Why do we need inventory?
When considering how to manage inventory using AI, I would like to first consider why inventory is necessary.
To put it simply, the reason why inventory is necessary is because the timing of supply and demand is different. If it is a product that can be manufactured and provided immediately when the customer wants it, it may not be necessary to have inventory.
However, many products take several weeks to manufacture, and several months including the transportation time from the manufacturing base to the sales base. In other words, if we start manufacturing when the customer wants it, it will be several months before we can deliver it to the customer. In order to be able to deliver the product immediately when the customer wants it, it is necessary to anticipate the future demand of the customer and prepare the necessary amount in advance.
This necessary amount is generally called the appropriate inventory amount.
2. How much should I keep in stock?
So, how should we determine the appropriate inventory amount? The answer to this question needs to be determined based on various constraints and conditions such as marketability, unit price, and product/material supply situation. There are several ways to think about appropriate inventory levels.
Adequate inventory = safety stock + number of demand over a period of time
Safety stock is stock that is always kept to a minimum to prevent shortages due to sudden fluctuations in demand. Again, there is no absolute level, and there are fluctuations depending on the company and product characteristics, but we maintain the number of demand expected in the period from placing an order to the manufacturing base until the product arrives (production lead time) It is an impression that there are many cases to keep.
What should be noted is the number of demand over a period of time. It's a matter of course, but if you know in advance when and how much the customer will purchase, you can secure the necessary and sufficient inventory to meet the demand, and neither shortage nor excess inventory will occur. . In addition, as mentioned above, the level of safety stock is also based on the number of demand for a certain period.
This is also natural, but since it is difficult to perfectly predict this demand in the actual business field, no matter how much you consider the outlook, more or less excess or shortage of inventory will occur. To meet this challenge, a process called inventory management is necessary.
3. How do you forecast demand?
So how can we forecast demand? There are two main methods of demand forecasting. Demand Forecast and Demand Planning.
A demand forecast is a forecast of future demand based on actual data such as past sales. For example, for a product that sold 11 units last month and 9 units last month, it is predicted that around 10 units will likely be sold this month as well. In recent years, it has become possible to make advanced predictions by utilizing various statistical models and using AI to calculate not only past sales results but also various correlation data. Demand forecasting is a method of forecasting the future based on various past performance data, and in general, this method is often referred to as demand forecasting.
On the other hand, while introducing and operating demand forecasting in actual business, there are cases where demand forecasting alone is not enough. This may be due to the fact that the calculation accuracy of the demand forecast itself is low, but there are also cases where forecasts are off due to factors other than calculation accuracy. For example, in the case of a product that sold 1050 units the year before last and 950 units last year, we would expect to sell around 1000 units this year if there are no special reasons. However, what will happen if this product is a focus product in this year's business strategy and the sales plan is for 5,000 units? Aside from whether we can sell 5000 units as planned, there is a high probability that we will miss the forecast of at least 1000 units, so we need to prepare an inventory close to the planned quantity.
In this way, there are cases in which the number of demand trends are an extension of past performance, and there are cases in which it is not expected to be an extension of past performance due to internal and external factors. In order to deal with such cases, the method of forecasting with some kind of"judgment and will of people"is called "demand planning". In order to perform inventory management based on demand forecasts in actual business sites, we believe that it is important to use these two demand forecasting methods properly or in combination depending on the situation.
Four. Is it possible to entrust inventory management to AI?
The performance of AI models and platforms that perform calculation processing is improving year by year, and it is possible to predict demand at a certain level by adjusting various learning data and explanatory variables, and we will make it a level that can be used in practice. I think it is possible. In fact, in the past, I used to manage inventory and promote business based on AI demand forecast results. However, from my experience, I don't think it's possible to completely leave it to AI. There are two reasons.
- No matter how much the forecast accuracy is increased, it is impossible to make a 100% accurate forecast, so it is necessary to continuously manage excess and shortage of inventory
- Due to internal and external factors, there are cases where the number of demand is not an extension of the past, and in the end there are cases where it is necessary to incorporate human intentions and judgments.
No matter how much data AI learns, demand forecasts will never be 100% correct. No matter how far you make predictions, ultimately human business decisions are required. Specifically, the main decision is whether to take excess inventory risk or shortage risk. The judgment criteria differ depending on the business environment, timing, and product characteristics. For example, if the product is a hot-selling product with a certain inventory turnover rate, rather than losing opportunities due to shortages, it is likely that even a little excess inventory can be quickly resolved, so the decision is made to take the risk of excess inventory. I guess. On the other hand, for products that do not sell very often, sell irregularly, and have a high unit price, the business impact of unsold items is large. It's like deciding to keep it.
However, this does not mean that AI demand forecasting cannot be used in operations at all. When people do all the demand forecasting and inventory management from 0 to 100, especially on sites with many SKUs, problems such as lack of human resources and individualization arise, which becomes a big bottleneck in business growth.
The key is hybrid operation of AI and human resources. As I mentioned earlier, the final business decisions are made by people, but I believe that hybrid operations, such as entrusting other parts to AI and systems, are important in order to achieve a balance between quantity and quality. .
The catalog below introduces an inventory optimization service that utilizes demand forecasting AI to ensure the success of this hybrid inventory management operation.
▼ Click here to download the catalog (free)
Also, please check the detailed explanation of the contents of this blog and the seminar video that includes a demonstration that can be used in actual business.
▼ Click here to watch the video (free)
In the fourth installment, we plan to discuss the management and optimization of "price", which is one of the important points for the continuous growth of EC sites and gaining the trust of customers.
looking forward to!
============================
▼ Product management support AI CrowdANALYTIX for EC to support expansion of EC site product lineup
Click here for details
Latest Information
Case Study
Blog
Document List