How Lenta creates an effective assortment based on data

Hello Habr! My name is Katya, I am a BigData product manager at Lenta Tech, responsible for the development of digital products in the “Assortment” and “Pricing” blocks. In this article I’ll tell you about the path to data-based assortment management and our flagship – Deli app – a workplace in which the manager analyzes the matrix of his categories and makes changes to it based on the recommendations of the algorithm.

Data-driven decision-making culture

Making management decisions based on data is not a new trend; the foundations of this principle were formulated at the beginning of the 20th century. Terabytes of data, user logs and any digital footprint become the source of business decisions every day. Therefore, it is unwise to ignore big data, but it will not be possible to process it with one conditional manager at the speed dictated by the market.

BigData divisions take companies to a new level, enriching the experience and expertise of their employees artificial intelligence, machine learning and big data processing algorithms. Business processes are accelerated, their automation minimizes the number of errors associated with manual data processing.

There are still disagreements in professional communities regarding the definition of data informed and data driven methods; the topic has been raised more than once in many articles.

From our point of view, the key difference by which a company can be considered data driven is this is maximum data purity, high priority for the company’s technological development, as well as the presence of a business model in which numbers have more weight in decision-making than management intuition.

At the stage of low or medium level of trust in the results of testing machine algorithms, the data-informed approach to decision making is considered optimal (in comparison with data-driven):

• it is quite flexible, but based on logic, patterns and statistical analysis;

• frees up team resources through automated data aggregation and primary analysis;

• has virtually no blind spots, leaving room for creativity, context, nuance, experience, anomalies and exceptions to the rules.

Deli is part of Lenta’s data-informed approach

The foundation of retail is the assortment of goods on the shelf in a store. To simplify greatly, the range should: meet the needs of the buyer and bring profit to the business owner.

If the assortment is not a narrow niche, but 99% of the goods are FMCG, as in Lenta, then fulfilling the conditions is mandatory for each of the hundreds of product groups. The same condition is met for businesses with a wide geography. In Lenta, each SKU on the shelf in a separate section of each store in each city must be profitable, high turnover, in demand by customers, and preferably cheaper than its competitor.

Back in 2019, our team launched the first pilots to localize the assortment: we tried to fill the shelves, taking into account the needs of customers of a particular store. To manage a matrix of products with such granularity, a transparent methodology for identifying outsider products and top positions, aggregation of a huge amount of information and, of course, a single decision-making window were required.

We developed the Top-Flop product rating logic and used it in the recommendation algorithm for each store. This algorithm grew to become the first MVP of the application and showed in A/B testing that using the new approach gives a significant increase in turnover in pilot stores.

Once we were convinced of the usefulness and ROI of the solution, we began developing Deli (de – delisting/output, li – listing/input) – an analytical tool and recommendation system for assortment management.

This was a big step towards the introduction of a data-informed approach in Lenta and the movement towards data-driven: preprocessing and analysis of big data, searching for solutions and assessing their impact on the business are no longer limited to a specific manager, and the speed of decision-making increases.

Top Deli Users

Deli is intended for employees responsible for creating Lenta’s assortment:

• federal category managers,

• category managers in divisions,

• category management directors and commercial directors,

• analysts.

We have more 120 employees are active users of Deli on a monthly basis, and we expect this number to only grow as the product evolves. The tool has become an integral part of the multi-stage business process of assortment management, which helped us significantly optimize it

Multitasking Deli: Its Level of Capability

Deli collects and processes data from various sources and makes recommendations for changing the assortment matrix. The Deli input receives a huge amount of information:

• matrix with data on sales categories at various levels of geography (federation, division, city);

• more than 30 key indicators for 12 months from more than 10 data marts (we update them daily);

• 20 attributes for each product;

• analytical sales forecast;

• merchandise structure and matryoshka;

• Nielsen market data;

• monitoring prices and assortment of competitors;

• TOP-FLOP methodology for ranking products;

• the effect of the changes.

The user has at his disposal the main KPIs of goods and their dynamics, price positioning “Lenta” vs Competitors with GAP analysis, algorithm recommendations for the input/output/rotation of goods, the current and future assortment structure.

Based on this data, the user evaluates the achievement of target indicators and makes an informed decision on matrix optimization. The user can use suggestions from Deli's algorithms or make manual changes based on their own expertise, and also evaluate the future economic impact of their actions.

Since the beginning of 2024, more than 10,000 sessions have been created to analyze the effectiveness of different product categories. We are gradually increasing the conversion of users to the tool due to new functionality, expanding the coverage of the product range, and adding new indicators and attributes.

Case #1: We provide a solid basis for the withdrawal of ineffective SKUs

The reasons for removing a product from the range or replacing it can be different: from discontinuation of a product and non-compliance with quality requirements to a low rating and exceeding quotas. For 8 months 2024 Almost 7000 SKU – a colossal amount that demonstrates flexibility and speed in changing the store shelf.

30% of these products (more than 2000) are derived based on Deli data and recommendations:

– low FLOP rating – we refuse ineffective or unprofitable SKUs, directly affecting the company’s turnover;

– over-quotes (i.e. exceeding the planned quantity of a product) – we respect the shelf capacity, so each product has a place and the opportunity to win the heart of the buyer and, perhaps, become the locomotive of the entire category;

– low level of performance – we monitor whether new products are gaining momentum quickly enough and how the mature assortment is slowing them down, we react before this can affect the overall indicators of the category.

Case #2: we quickly respond to KPIs of goods in each city

Since the beginning of 2024, the Volga division has introduced regular analysis of the range of sausage products based on Deli. The manager creates more than 100 sessions and analyzes the matrix of goods separately in each of 20 cities of the divisionhaving at your disposal:

1. calculated rating for ~1200 SKU,

2. checking the rules for filling the matrix and compliance with quotas,

3. recommendations for changes and their effects,

4. Competitors' prices

5. Performance of top market products.

Category managers note that the main convenience is that you do not need to independently and through a bunch of VLOOKUP reports analyze what is displayed and what is in the active matrix.

It’s very easy to look at the results of the month to see what’s going strong in terms of sales or, conversely, write-offs. Flop-SKUs are immediately visible and can be derived from the matrix. Lots of useful numbers in one place: both the market and competitors’ prices – you can react quickly.

How we plan to develop the tool

We continue to develop the product and enrich it with new features. We are gradually transforming Deli into Assortment Management Center and we plan to implement:

• generation of artifacts and documents, their exchange (for example, input-output card);

• coordination of decisions to change the assortment within the application;

• integration with internal accounting systems and other Lenta products;

• separate modules for regular category reviews (with functionality transferred from existing dashboards);

• improved algorithms and additional logic to manage categories that require special approach or rules.

Together with Deli users, we have already covered an impressive part of the path to data-driven: from idea and testing to a high level of trust in recommendation algorithms and implementation of the tool in the business process.

With Deli, we, as a team, and Lenta will continue to increase the speed of decision-making in assortment management, enhance accuracy, rely on data and the harmonious logic of algorithms, and increase business efficiency and profitability.

Colleagues, share your experience in automating decision making using data in the comments. Do you consider your company data-driven?