Drogas. Forecasting and replenishment done by 4 people

MySales and Drogas integration project in Latvia and Lithuania

MySales helps retailers with producing more accurate and automated sales forecasts leveraging its deep learning technology.
Drogas, the leading home and beauty retail chain in Latvia implemented MySales two and a half years ago.
This year (2022) MySales was launched at Drogas in Lithuania.

In the interview Artyom Usakov (Supply Chain Manager) and Igor Scebrin (Commercial Director) from Drogas talk about goals that forecasting sales automation helped achieve, what were their criterias of choice for the forecasting tech provider and why did they decide to go with MySales.




— Please tell the readers more about Drogas.

—Drogas operates in Latvia and Lithuania, part of A.S. Watson Group, which has over 16,000 stores in Europe and Asia. A.S. Watson Group operates in 35 markets and in each market is either a leader or one of the leaders in the segment of health, beauty and home products. Drogas operates in the hygiene, cosmetics and home goods segment. We have 95 stores in the Latvian market and 55 in Lithuania. Drogas is the leading retail chain in the health and beauty segment in Baltics.

Who are your customers?

— Our product is targeted for all age and price audiences in the mass market. Typically, our customer is a woman between 18 and 65.

— How is the store assortment being determined?

— The assortment depends on the size of the store. There are locations with an area of 100 sq.m and 400 sq.m., but this is rather an exception. Our standard store area is 250-300 sq.m.

Tell us about your online store, how important is it to your business?

— We do not consider a website only as an online store. It aggregates all information about the products, our projects, the company, tips&tricks. Customers can check the reviews written by other clients. We see many customers familiarizing themselves with the products online and heading towards the store to make a purchase. It doesn't really matter for us if they buy online or offline. We consider ourselves an O+O (offline plus online) retailer. For a relatively small Latvia, we have a fairly high coverage of physical stores. If you take 15-min through Riga city center, you’ll encounter 13-15 Drogas stores. In Latvia half of the population lives and/or works in the capital and we have half of our stores located here. For stores located within walking distance, it is easier to pick the product on your own than order online. Traffic jams in Riga are not really a big enough problem to disencourage from visiting stores. We don't have huge malls where it takes 30 minutes to walk from the parking lot to the store. Coronavirus restrictions in offline stores have done the best marketing for online shopping. As soon as they were removed, sales in physical stores immediately increased.

— Your audience is fairly large and diverse.  There are a number of motives, factors, preferences for its various segments. Which factors influence the sales the most for your business?

— There are a lot of factors and they are relevant not only for our business, but for the entire retail industry. These are seasonality, local holidays, weather, tourism, consumer preferences, trends, prices and promotions. For example, recently Korean cosmetics have been especially popular throughout Eastern Europe. It is almost impossible for a human being to analyze everything, so help from algorithms is needed. In the past, store managers ordered goods in our company. It was ineffective. The quality of ordering depended on people whose expertise did not really lie in ordering goods. They could see their store's sales, but not the general trends. We were dependent on a person in case of personnel changes. Will the new manager be able to order goods efficiently? And the quality, shall we say, was “average”. Our colleagues from other countries told us that they did tests and the best store manager ordered the goods at the automatic order level. Instead of dealing with an order, store managers can focus on improving store operations by serving customers, preparing the store for work, preparing goods for presentation, etc. This is why we started looking for a solution to automate the ordering process. There were already operational solutions for that within the other chains of A.S. Watson Group. Our shortlist consisted of two vendors: MySales, which was already used in one of the A.S. Watson businesses that time and another well-known system operating in another retail chain in the group. We spoke to our colleagues from various departments of the company that used MySales. In the end we came to the conclusion to go ahead with MySales sales forecasting module and stay with our present auto-ordering tool. So far we did not decide to move to MySales automated replenishment module due to significant effort needed for conducting such a migration. We load MySales forecasts into the solution we’ve used before.

— Why did you choose MySales?

— It was important for us to keep the number of people involved in working with the system at the minimum and how autonomous it is.

An alternative solution involved a significantly larger number of human effort..

It is worth noting that our Latvian and Lithuanian business use two different system instances since they have separate databases. The same product from the same supplier for different markets has different article numbers and SKUs. Latvian and Lithuanian markets in terms of sales are radically different. In Lithuania, our stores are more focused on beauty. In addition to beauty, personal and home care products are in demand in Latvia. With various sales patterns, different store specifics and other top products, two auto-ordering systems is a must. We saw that a team of three people is able to comfortably control orders in around 90 stores in Latvia. Currently a team of four controls ordering in two parallel systems for two countries. According to our estimates, for an alternative system we’d need to involve 15-20 people. I can't really say if MySales is more accurate than the other solution we looked into. Certainly there was less automation there. For us it was crucial that MySales allows for a significant process automaton and keeps resources needed to maintain it at a reasonable level.

— Which departments within your company are involved in working with MySales sales forecasting module?

—  It is a commercial department encompassing logistics, online store, planograms, analytics, category managers, and even some marketing specialists. Only HR and Finance are not involved.

— What difficulties did you encounter during the implementation phase?

— An unexpected difficulty turned out to be that it was difficult for store directors to step away from the ordering process. There was a question of trust. Previously they kept control over it and now it is being taken away. It was a psychologically difficult moment for them.

We also had to rebuild a lot of processes. The department of orders was created from scratch with new procedures. System requires filling in new parameters. In all planograms, we had to set the x, y, z indicators for the product and we needed to improve the quality of order schedules from suppliers. A lot of processes have been revamped to fit the new system and make it work. During the implementation process, we saw anomalies, that is, a discrepancy between what we expected and what actually happened. We contacted the MySales team, who studied the issue and quickly implemented the required changes. The system is self-learning and every month it gets better. We encounter fewer and fewer anomalies. At some point, 2,5 years ago, something broke, but MySales quickly resolved everything. After that, there were no major breakdowns.


— MySales stresses the importance of the existence of a history of sales and all activities for the system to work properly.

— Indeed, before the launch, we spent a lot of time preparing historical data on sales, promotions and marketing activities, so the system had a basis for forecasting. Before the launch, we discussed in detail with the MySales team all the nuances, including how certain factors will be taken into account.

—  How many factors are taken into account in the forecast?

— Definitely more than 10 factors. Among them are weather, seasonality, and competitor activities. Keep in mind that MySales is different for each market. After relaunching, we did certain optimizations in the system setup. By trial and error, with both us and MySales involved, we got a good result. Another important point, there are significant differences between Drogas and  other retail chain from A.S. Watson that has already used MySales. They have their own warehouse and we don't. This significantly affects the process and orders. We have a 3PL partner who provides logistics services. They have their own system, which is not integrated with ours. Therefore, MySales focuses only on forecasting at the store level. It works with deliveries to stores but not to a centralized warehouse.

— For how long do you build a forecast and how often is it recalculated?

— The forecast is recalculated every day. Forecast date range in the future covers all the active promotions and marketing activities. As soon as we receive information about the next promotions, it is added to the system and forecast recalculation takes place. In case of changes in input parameters, we can manually start the recalculation at any time. For example, promo rules were changed and the forecast recalculation was requested. With fresh data, the system may interpret the data differently based on the latest sales. For example, 6 months have passed since the last similar promo. The same SKU which was sold at a 50% discount may now perform differently. There are various factors that affect product sales at any given time.

Results of Drogas and MySales cooperation

— Have you achieved the results that you set as goals before the start of the project?

— Without any doubt, yes. The system is now working well and ensures optimal inventory levels in our stores. We use several KPIs: inventory qualitative indicators and inventory availability (i.e. Slow Moving Inventory, Weeks of Cover). There are certain budgetary frameworks that we need to fulfill and we can see that everything is in order.

We separately monitor the best stores and top products to ensure their availability for the nearest dates. An important KPI is On-Shelf Availability. This indicator is kept at a good level. And if it decreases somewhere, then as a result of the analysis, we can identify the reasons. Recently this is the impact of Covid, the logistics and energy crises. In most cases, this is not related to MySales.

— To what extent do forecasts and actual sales match?

—We didn't check for it specifically, but colleagues from MySales provided us with a report that the accuracy of promo forecasts is at the level of 80%+, and regular at 90%+. If the forecast is lower than actual sales, then very quickly we will receive feedback from the store that there is not enough product or, conversely, there is too much of it. We can track it by such queries. For some time we do not receive such feedback, which means that the system does the job.

— How does MySales deal with promotional sales? This is an important area for a mass-market retailer. 

— We sometimes had questions about promo forecast output. We asked questions to MySales representatives and they explained the rationale behind such a forecast output. We also worked on our bugs.

We apply some safety stock management factors that allow us to slightly offset the effect in case MySales is wrong. We apply an increased coefficient on top products and on top stores. This is a certain insurance. We developed and independently added these coefficients to the MySales settings. The system allows you to add custom parameters for every store and product.

— Could you share any other business results after the MySales implementation?

— Definitely saving store resources. In each store we freed one person to work directly at the outlet rather than work with ordering. We have optimal store inventory levels and keep a control over it - if 90 people work with it and each of them does it his/her own way, you can’t really call it a good level of control. I’d like to also stress the simplicity of the system. This was a must-have for us. Currently, sales forecasting for our two businesses in two countries is done by MySales with only 4 people involved in the process.

— How much time did it take from choosing MySales to having the first forecast delivered?

— In Latvia it took us more time, because we worked on it from scratch. Procedures had to be redesigned, information exchange and collaboration between departments needed to be improved. We did a lot of tests: how data is transmitted, how providers receive it. And when we did all the preparatory work, there were still certain concerns. The system is new, but what will happen? We started with migrating small suppliers, then medium ones. We were connecting big suppliers one by one, checking results in each case. From a technical point of view, the process took up to 6 months for us, and 2-3 months for MySales. In Lithuania, the process has already gone faster. From the MySales side, everything happened quite simply and quickly.


— What would you recommend to top managers who are now on the way to implementing an automated sales forecasting system in their stores?

— I would recommend not being afraid and trying something new. If we talk about MySales, then on a 10-point scale, I would give 8-9 points. Having learned how to work with it with little human effort involved, you can achieve good results.