paint-brush
[Shower Thoughts] Why do we Need to Forecast Inventory and not Demand?by@elisavetta

[Shower Thoughts] Why do we Need to Forecast Inventory and not Demand?

by Leafio.aiOctober 3rd, 2019
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

Most companies use demand forecasting methods that are outdated 10-15 years ago: exponential smoothing, ARIMA, Moving Average, Holt-Winters method and others. They are not effective at solving the problem of inventory management for 94% of the product range and for almost all non-food products. With the help of probabilistic methods of forecasting the 4th generation, we can forecast the entire range of products up to Group Z. In the first part of this guide we will look at the following issues:Why do you need to predict inventory and not demand?
featured image - [Shower Thoughts] Why do we Need to Forecast Inventory and not Demand?
Leafio.ai HackerNoon profile picture

Most companies use demand forecasting methods that are outdated 10-15 years ago: exponential smoothing, ARIMA, Moving Average, Holt-Winters method and others. Not only are they morally outdated, but they are not effective at solving the problem of inventory management for 94% of the product range and for almost all non-food products, which is proved by many scientific studies (see scientific note).

As a result, the company loses from 20 to 40% of its working capital due to excess stock and up to 5-10% of its revenue due to deficit. The fact is that these methods offer a single number, which is often similar to guessing on the coffee grounds. The insurance stock is added either "by eye" or by standard distributions of demand, which simply do not correspond to these products.

Now we have access to multiply large computational resources which make it possible to use new methods of forecasting of the 4th generation.

In order for the 4th generation algorithms to be effective, they must work on correct historical data and be understandable to employees. Past sales statistics do not reflect the basic dynamics of demand and require cleaning from deficit periods, bursts due to marketing campaigns, one-time large sales, etc.

Evaluation of the accuracy of such methods also differs and is made in terms of economic losses of the company, rather than statistical errors MAPE or RMSE, which are incomprehensible to the business. This requires a restructuring of thinking not only of managers, but also of ordinary employees.

In the first part of this guide we will look at the following issues:

  • Why do we have to process our sales history before making forecasts?
  • How do we do this?
  • How can this affect the end result?

If you are already an experienced professional, you can go straight to the second part of the manual. In it we will tell you about it:

  • Why do you need to predict inventory and not demand?
  • What are the advantages of probabilistic forecasting methods?

In Europe, the use of probabilistic algorithms is minimal, although in Western practice they are considered to be the most advanced method for managing inventories. With the help of probabilistic methods of forecasting the 4th generation, we can forecast the entire range of products up to Group Z, not only calculate the optimal stock for each store and warehouse, taking into account the exact dates of receipt of orders from suppliers, but also take into account the shifts in timing of deliveries and their instability, the variability of marketing actions.

Moreover, these algorithms can suggest the optimal level of demand satisfaction for each position in each order. At the same time, a balance will be found between the deficit, the cost of inventory storage, the freezing of cash, the write-off of spoiled products and other factors.