Data-Driven Inventory Optimization - Part 1: driving down Out-of-Stock

When it comes to inventory, companies are often exposed to a double negative effect, with shortages on the best-selling products and excess on the bad ones. In this article in 2 parts, we will explore optimizations for limiting both out-of-stock and excessive inventory.

We have all as consumers experienced this frustrating situation of facing an empty shelf when looking for our product in a store.

Out-of-stock is a critical problem for many industries:

• Retailers lose sales with the risk of losing the whole recurring customer in case of repetitive out-of-stock
• Suppliers pay penalties to their clients for delays caused by out-of-stock?
• Airlines ground aircrafts due to critical spare-parts out-of-stock

If getting rid of out-of-stock is a major challenge, it is also a great lever to grow sales, increase customer satisfaction and save money.?

In this post, we will see how to put out-of-stock under control. We will focus on the Retail industry but most of the concepts apply to other industries.

1) Assessing the true cost of out-of-stock

Studies reveal that on average, every time a shopper enters a grocery store, one in 12 items on his list is out-of-stock, i.e.: an average out-of-stock rate of 8.3%.

Shopper response to out-of-stock leads to a sale loss in 40% of the cases for retailers and 35% for manufacturers according to a worldwide study.

Retailers lose on average 3% of their annual sales due to out-of-stock.

The sale loss is not the only consequence of out-of-stock, in the long run, repetitive out-of-stock leads to permanent customer loss. Study reported that when a consumer faces an out-of-stock in a planned purchase category, the shopper will permanently switch store after an average 2.4 such experiences.

Based on the historical sales and inventory data, a simple tool can be built to size the cost of out-of-stock.

2) Putting out-of-stock under?control

There are 4 key processes for driving down out-of-stock.

1. Assortment Planning

The first step for preventing out-of-stock is to have a well-defined assortment plan. What products need to be in each store at the different periods of the year??

• Strategic products and Key Value Items (KVIs) must be available in all stores
• Niche and specialty products should be available in large stores with high shelf space
• Some seasonal products must be available only within a certain period of the year
• Some products are must have only in specific regions because of climate or cultural reasons

Assortment Planning is a key enabler for managing on-shelf availability and inventory. Analytics and AI can be leveraged to strategically plan assortment to get the most value out of store square footage. This will be the subject of a future article.

2. Demand Planning

Demand Planning is a critical process for better inventory planning and out-of-stock prevention.

The objective of demand planning is to provide the most accurate possible demand estimates to be used by supply to effectively support replenishment decisions.

Several forecasting techniques can be used to predict demand. They can be classified in 3 families:?

• Na?ve methods that predict future values based on simple mathematical operations applied to past observations. While these methods can be surprisingly effective, especially for stable or, at contrary, very chaotic signals, they find their limits with products with complex seasonality or sensitive to promotions, holidays or weather. These methods are often used as benchmark to assess the value added of more complex methods.
• Judgmental methods that rely on experience, intuition, and educated guesses.?Judgmental methods are either used when there are no available data, so that statistical methods are not applicable or when data are available, statistical forecasts are generated and then adjusted using judgement. These methods are more effective to respond to sudden demand shifts where statistical models are slower to respond like the sudden return of high inflation in western countries, the pandemic or new labeling driven by legislation.
• Advanced methods that rely on Statistics, Data Science, Machine Learning and Artificial Intelligence. There are a wide range of techniques, each has its own performance, complexity and data requirements.

Retailers need granular demand forecasting at the product and store level on a daily or weekly basis. This represents a tremendous number of series to forecast. Fortunately, modern technology and computing power bring significant capabilities for forecasting large number of series more accurately.

3. Supply Planning

Supply Planning is the process of anticipating the demand for products and planning their fulfillment. It takes the demand plan provided by demand planning, inventory on-hand and supply constraints like lead time and minimum order quantities to determine re-order points and quantities.?

One of the most important parameters to optimize out-of-stock is Safety Stock, an extra quantity of inventory used to prevent a shortage situation. Safety stock is very important to prevent either higher demand than forecasted or lead times delays.

Many methods exist for calculating safety stock.

With a demand and supply variabilities following the normal statistical distribution, with no safety stock, the cycle stock carries enough inventory to meet demand in 50% of replenishment cycles. By adding a safety stock equal to 1.65 standard deviations of demand and supply variabilities, demand would be satisfied in 95% of cycles, meaning 5 out of 100 cycles will know an out-of-stock.

Although the classic King’s formula uses the standard deviation of demand as the proxy for demand variability, a better estimate is the forecast error. Either Mean Absolute Deviation (MAD) or Root-Mean-Square Deviation (RMSE) metrics can be used.

Supply variability can measured by the total lead time standard deviation.

For strategic products safety stock can be increased by applying a Z-factor of 2.05 to reach 98% of service level that correspond to an out-of-stock probability of 2%.

Review period and up-to level are 2 other important parameters for out-of-stock optimization. The longer the review period is, the higher is the risk. If orders are made every Monday and an out-of-stock happens on Thursday, there will need to wait 4 days before making the order. The review period must consider the supplier schedule, if the supplier process orders or schedule deliveries once a month, this must be accounted for in the review period analysis.

The up-to or max level is the target inventory quantity at the beginning of the cycle, it has to be calculated to cover the demand over the total lead time (from order to receive) + the safety stock.

Shelf space allocation also plays a big role for out-of-stock optimization, especially for stores with limited or no stockroom. The ideal number of facings can be calculated from the up-to level.

4. Availability Monitoring

I’ve seen so many retailers blind on their on-shelf availability and at the exception of some mature players, a lot still rely on manual review by store employees to check availability of the products on the shelf to replenish.

To effectively put out-of-stock under control, on-shelf availability must be a KPI tracked at the top of the sales and store operations organisation and cascaded throughout the organisation to regions, stores and departments within the stores.

With proper Analytics, availability can easily be tracked and monitored at different level of the organization. The inventory data in the ERP might give only a partial view of the on-shelf availability as it often does not distinguish whether the inventory is on the shelf or in the stockroom. A cost-effective way to detect potential out-of-shelf with inventory in the store is to use a statistical algorithm of sales pattern that analyzes products with zero sales for a certain period vs their usual sales frequency. This works best with Fast-Moving products which are the most important for out-of-stock prevention.

Other solutions use cameras and computer vision artificial intelligence for out-of-shelf detection.

What’s next?

In the second part of the article, we will see how to optimize excessive inventory.

About the?author

Nassim Hartani is a senior business strategy advisor with a track record of leveraging the power of advanced analytics and data science to drive profitable growth and operational excellence.