Automated Zinu store

Automated store. How to organize the supply chain?

An automated store – Zinu Shop – was built in Kraków, in the M1 shopping center. How is it different from the solutions presented by retail chains so far?

First of all, it is a Polish initiative. In contrast to the concepts tested earlier by Lewiatan, Carrefour, and Żabka, the operator is a domestic company with capital ties to the owner of the Delfin supermarket chain.

Poles are definitely not behind and they have their first automated store. The question of whether this type of concept will really catch on in our market should be left for the customers to answer.

Below, however, I will explain how to organize a supply chain for a chain of automated stores offering groceries and fresh products. I will discuss the most important business measures and issues that require attention in planning the flow of goods.

In many aspects, the logistics of such a business resembles the so-called e-grocery, i.e. online grocery stores. So what is the challenge?

Customer service

Whatever you do, especially in business, it’s worth considering why you do it. The question recommended by Simon Sinek will therefore be the starting point of the analysis.

As a standard, the supply chain is accounted for in three groups of KPIs. Service offered to the customer, quality, and logistics costs. Depending on the organization, a selected group comes to the fore.

The Zinu automated store operates in the convenience category. The promises made to the client are based on proximity and constant availability.

Zinu Shop is a fully automated grocery store, open 24 hours a day, 365 days a year, including Sundays and holidays, which gives us a significant advantage over traditional trade[…]

— Artur Kowalski, CEO of Zinu, in an interview for

In this case, service is the most important goal of logistics. But how to quantify and measure customer service?

Product availability

In the retail chain, supply chain service means the availability of the product in the store, literally on the shelf. It’s all about On Shelf Availability or OSA, in short.

The distinction between in-store and on-shelf availability is important in traditional stores that have a sales floor and a warehouse. Contrary to appearances, this is not a trivial distinction. Many retail chains invest in the technology of radio-frequency identification of products, i.e. RFID, in order to know where the goods are physically located.

However, in the case of Zinu, due to the modularity of the store, it makes sense to stick to the traditional measure of availability in the store. In this case, OSA should be expressed in relation to the required stock and, at least, a single piece of goods.

Fulfillment of the vision presented by Artur can be achieved through the appropriate level of frequency of accessibility measurement.

When I started managing the supply chain in trade, we still operated on daily data. The IT system at the point of sale sent data to the central system only after the sale was completed.

Currently, inventory-level data can be synchronized almost continuously. This way, information about availability in the store can be used not only to plan deliveries but also to react to unforeseen shortages of goods.

However, if Zinu really focuses on accessibility for the customer, taking the right perspective when reporting is as important as choosing a specific measure. Most retail companies use a reporting scheme when a product is available.

Source: Own study. Sample data.

However, as you can see in the first presented chart, in most cases the goods are actually in the store. So we have an excessive sense of control over the situation.

Of course, you can work on the perspective of the presented data by selecting the minimum axis start. But then, we come to the issues raised in the short and classic book – How to lie with statistics.

I admit that I used the availability reporting model mentioned above for years. It was only the responsibility for logistics operations for e-commerce that taught me a different approach.

Currently, I believe that a business that puts service as its main goal should look at the lack of availability. In such a situation a simple transformation of previously used data gives a completely different perspective. You can clearly see the difference between a missing 1% and 2% or 4%.

Automated store. Stock shortage

Source: Own study. Sample data

This view on availability, or rather the level of order fulfillment, is characteristic of, for example, InPost. You can find out more about this in a conversation between Maciej Wieczorek and Rafał Brzoska, which is available on Youtube.

On-time delivery

But let’s get back to the Zinu Shop logistics. Another measure that we should follow when running an automated store is the timeliness of deliveries in relation to the assumed schedule.

Timeliness, sometimes also referred to as deliverability, belongs to the category of leading metrics. By itself, it is usually not in the center of interest of the business, let alone the client. However, if we start to have problems with timeliness in a given location or, even worse, in the whole retail chain, the next step is the lack of goods.

Therefore, it is worth tracking the timeliness of deliveries. Best broken down into individual points. Alternatively, you can break them down into routes in which the points are grouped. I have prepared an example of such a report below. I recommend using it in the logistics of any retail company that manages physical locations. Regardless of whether they are shops or  Pick-Up Drop-Off Points (PUDO points).

Automated store. On time delivery

Source: Own study. Sample data.

The selected degree of aggregation over time depends on the number of deliveries per week. In the example used, I assumed no more than one delivery per day according to the information I obtained from the company that manages Zinu. Thus the report presents the perspective within a week.

It is only in this perspective that we can see in which locations we may encounter problems. More importantly, you can see it even before problems with the availability of goods occur.

It is also worth making timeliness of deliveries one of the measures tracked on an ongoing basis if the automated store is also to function as a PUDO. In one of the comments under the interview mentioned above on, there was an idea for a mobile application that would allow you to pick up a previously placed order.

Automated store. Development idea

Source: Comment under the article on in which one of the users suggests creating an app that would enhance the shopping experience.

A natural development of the described concept would be the supply of products that are not permanently available at a given point of sale. If Zinu Shop provides such a service over time, the timeliness of deliveries will become, next to the availability of goods in the automated store, the basic measure of the service.

Then it will also be a good idea to analyze and report the reasons for delays in deliveries to the store. You can do this using several classes of reasons such as delays on the route, delays in leaving the distribution center, or arriving at the center for goods.

Finally, when the rotation at individual points of sale increases, it will become important to replenish store stocks along dynamic, and not fixed routes. Thanks to this, you can counteract the impact of traffic jams on the time-consumption of deliveries.

Sounds complicated? Well, it can be, but convenience obliges.

Accuracy of demand forecasts

One of the key measures that ultimately translate into the availability offered to the customer is also the accuracy of the demand forecasts. But how to express accuracy in a retail network in operational terms?

You have a group of numerous metrics to choose from. Most of them are variations of the following 3:

  • forecast bias
  • mean absolute deviation (MAD)
  • mean absolute percentage error (MAPE)

An automated store focusing on product availability can be planned based on the MAPE variant. As long as the forecast bias remains at a reasonable level and we assume a day as the measurement period.

For manufacturing companies, or even non-food retail chains that use weekly scheduling cycles, you can select longer periods.

However, if we analyze a store in the convenience category, characterized by a clear cyclicality within a week and high sales per product, we must go to the lowest possible level.

Weighted mape formula

It is also worthwhile to take sales volume as a weight. In this way, the measure of forecast accuracy supports the first of the supply chain objectives, i.e. product availability at the point of sale.

The value of the weighted forecast error should of course be as low as possible. However, in order to really reflect the situation in the store, the forecast must be unbiased.

In practice, the weighted MAPE is directly interpretable when the forecast bias remains within +/- 10% at the chain level. For a broader discussion of this type of dependency, see How to Best Understand Forecast Bias.

Forecast bias formula

Finally, when analyzing both of the aforementioned measures, it is important to remember that the forecasting process is always only a means, not an end in itself. The forecast is only relevant to the extent that it enables the supply chain to meet other priorities.

Higher stock availability, lower damage levels, or lower operating costs. That is why we are looking for methods for better planning.

Besides, while forecasting is an important part of supply chain responsibility, it still represents just one step in the process. A number of other factors may have a greater impact on the availability and value of capital invested in the stock.

For example, the construction of the planning process itself, which I wrote about some time ago in the article – How to increase profits with Sales & Operations Planning?

Quality and control

One of the assumptions taken for granted in any supply chain is quality. In the case of the supply chain of a retail chain, and this is what the Zinu automated store aspires to become, this means control over best-before dates and the lowest possible loss of stock.

The control of best-before dates begins at the stage of accepting the goods at the distribution center. The Warehouse Management System used should enable storing information about the best-before dates of each batch of goods received.

With a rigorous approach to the organization of the process, the received goods are physically marked with an additional code. In some industries, for example in pharmacy, product expiration dates are managed this way.

However, for the assortment from the convenience category, you can use the system tag as long as the goods from different deliveries are stored in the central warehouse in different storage locations. This is how I would organize the process in Zinu Shop.

Automated store. Expiry date control

Source: Twitter post. Accessed: March 2021. Personal information is intentionally hidden.

As a result, it would be possible to stock the store in accordance with the FEFO (First Expiry First Out) principle. This would lead to the reduction of losses resulting from the expiration of goods in the warehouse and the elimination of errors similar to those seen on Twitter by an employee of a retail chain in England.

Losses in the supply chain can also result from theft and damage on the sales floor, damage during receipt or delivery from the logistics center, or damage in transit. The automated shop is practically devoid of the former.

However, it is still necessary to monitor the value of losses in the other two areas. Preferably not only as overall statistics expressed in terms of value but also in breakdowns by:

  • delivery routes (or stores),
  • product categories,
  • suppliers,
  • SKU.

Always in relation to the volume processed by the chain in the analyzed period. Only then can you see what could be causing the problem.

In practice, I learned about this once when one of my suppliers changed the type of cap on the bottles. Only the analysis in cross-section by routes and SKU in these deliveries showed what was the cause of the abrupt increase in transport losses.

Logistics costs

The third group of KPIs, on which the supply chain works, are costs. On the one hand, the operating costs of the logistics itself. On the other hand, the cost of the capital that we need to invest to keep the stock at a safe level.

I have already written about the level of supply chain costs in retail chains, among others, in the article – How much logistics costs you? In analyzing the possibilities of Zinu Shop, I will focus mainly on the costs of maintaining the stock.

Regarding operating costs, I will only mention that in fast-growing companies it is worth tracking costs in relation to the turnover achieved. In absolute terms, they will generally be higher every year.

However, you should also monitor costs in absolute terms per unit of handled goods. You should control the average costs of transport, storage, or receive. This creates an analytical balance of looking at logistics costs.

Capital costs

Compared to a traditional store, an automated store has some limitations. There is no possibility to insert additional seasonal or promotional product displays. This changes the way the customer’s cart is built. It also makes it difficult to obtain additional income, which retail chains derive from presenting the offer of their suppliers.

From the perspective of supply chain management, however, it has some advantages. One of the non-obvious ones is controlling the level of inventory turns.

The stock of chain stores that do not have a sales floor is always located in a strictly defined place. On the shelf, in a transport container, or in a storage location in the central warehouse.

An automated store simply cannot accept more goods than there is space in the chambers inside. Unlike a traditional store, which “always” will fit something in the storage in the back room.

In a way, it enforces rational inventory management. It does not allow for situations that you can sometimes notice when looking at the back of traditional stores.

Back of a store. Warehouse stock replenishment
Source: Mailboxde

But even in an automated environment, it’s worth keeping an eye on stock levels. Both per store, product category, and supplier in the chain stores.

Stock that does not rotate means costs. First, the costs of frozen capital. Ultimately, the costs of returning the goods to the supplier or disposal. So how to check if the rotation of goods remains at the right level?

In the context of a retail store, it is best to analyze the coverage of sales with stock, the so-called stock cover. It is enough to relate the time-averaged value of the stock to the sales of the goods in the same period, which we will express in its purchase prices.

In this way, we obtain a measure that can be converted by individual stores, product categories, or suppliers within one category. A measure that determines for how many days (on average) we have enough stock.

Stock cover formula

Of course, for each product that the automated store has included in the offer, we must have stock. However, it is important that it is not excessive.

If deliveries to points of sale are made several times a week, there is no reason to keep stock for more than 7 days in each of them.

All with the assumption that the safety stock itself does not have to exceed this level. If our sales are characterized by significant variability, higher stock levels may naturally be required.

In practice, however, retail chains have a greater problem with the stock remaining after promotional campaigns or unsuccessful product implementations than with the stock required to cover fluctuations in regular sales.

Therefore, you should regularly analyze stock cover in every retail chain. The Zinu Shop supply chain team will have less of a problem with this than similar teams in traditional chains. However, they must also pay attention to their impact on the cost of capital in the organization.

The automated store of the future

Organizing the logistics of an automated chain of stores is a complex task. However, it is not impossible. If we take care of the issues described above, the supply chain created by us will be up to the task.

It will fully control its key parameters. Then you can peacefully work on the rest of the business.

As in a synchronized orchestra, all areas should play at a similar level in any well-functioning company. They should adapt to each other. Otherwise, even with a well-organized supply chain, the organization may fail.

I wrote more about this earlier in the post – Logistics is not everything.

I hope reading my thoughts on organizing an analytical supply chain of an automated store will inspire you to do something. Maybe you can use some of it in your own business?

This time I have not described how to physically organize the logistics of a retail chain. I have focused on the planning process and key metrics. I will cover this topic in one of the next articles.

Next post coming soon!

Leave a Reply

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