Hypermarket chain automates pricing by leveraging artificial intelligence to stay on top of competition

CUSTOMER
PROFILE:

Industry:
Grocery

Location:
Eastern Europe.

Hypermarket chain and supermarket chain.

47 brick-and-mortar stores in 12 cities.

6,000+ employees.

35,000 SKUs.

1,700 private label products.

Self-manufacture: 34 production halls produce 60 tonnes of food daily.

Founded:
1994.
Looking for new growth opportunities
In 2021, the retailer began to shift its pricing to a data-driven approach with the Imprice platform.

In terms of sales volume, profitability and remaining competitive position, the company's pricing was very efficient. However, the retailer's team saw an opportunity to enhance their profitability by leveraging innovations.
The key challenge:

The company's pricing processes were semi-manual and partially decentralized. Some processes were based on the expertise of category managers and store managers and difficult to standardize.

In fact, it would take months to describe the pricing system in detail.

On the other hand, launching a large pilot with a lack of pricing documentation could cause errors, a drastic drop in revenue and margins and the loss of customers.
The Imprice flexible solution empowered the retailer's team to start a price optimization journey in small steps:

Describe pricing processes in detail for specific product categories.
Add these described categories to the pilot.
Monitor results to find missed elements and factors of the pricing process and related errors.
Adjust pricing rules according to new information about business processes.
Continue to monitor pricing performance.
Describe pricing rules and details for other categories, add these categories to the pilot test, etcetera.
CHALLENGES:

Automate pricing processes, which, despite being intuition-based and partly decentralized, were ensuring a satisfactory high gross profit and competitive success.

Get more clarity in pricing and its results.

Estimate the potential effect of dynamic pricing implementation.
Key results
In a year, the price calculations processes for all product categories were fully automated.

The company successfully completed 3 pilot tests between April 2021 and March 2022.

With Imprice, the retailer's team standardized, automated and improved all processes that significantly affect price calculations and pricing performance evaluation.

Throughout the "pilot year", the team saw increases in profits and sales within the categories included in the pilots. The sales performance in the pilot stores exceeded the performance in the control stores.

The company gained +2% gross profit growth
in the last five months of the pilot,
when the Imprice platform automatically managed regular prices for most categories of goods
(the test stores' result compared to the control stores).


The company found out that the Imprice solution paid off after being implemented in one store, and expanding the implementation to more stores ensures an impressive ROI.

In April 2022, the retailer ran a rollout to leverage dynamic pricing across the entire chain.
28 Years Of Success
With 47 hypermarkets and supermarkets, the store chain is a retail division of a commercial real estate developer that builds and rents out shopping centres and store buildings.
The parent company owns all retail spaces of the chain.

In terms of frequency of store visits, the hypermarkets are close to convenience stores. A typical customer visits the store three to four times a week, because hypermarkets are located in residential areas.

The retailer has 34 self-manufacturing production halls, including 30 full-cycle ones. The company produces bakery, pastries and confectionery products, semi-finished meat products, salads and prepared meals.

Since 2017, competitive pressure from countrywide retailers significantly increased.

Despite large competitors' activities, the company managed to successfully maintain customer loyalty through private labels (the chain produces about 1,700 items), wise pricing strategies and excellent customer service based on efficient business processes.

Store managers spend up to 80 percent of their working time at the shopping area, not in their office. They execute a program called "600 points to check"; these are the parameters that the store manager inspects daily.
These points to check include the freshness of food, cleanliness, merchandising quality and so forth.
During the daily inspection, the store manager fills out a checklist and notes things to improve.
CUSTOMER
PROFILE:

Industry:
Grocery

Location:
Eastern Europe.

Hypermarket chain and supermarket chain.

47 brick-and-mortar stores in 12 cities.

6,000+ employees.

35,000 SKUs.

1,700 private label products.

Self-manufacture: 34 production halls produce 60 tonnes of food daily.

Founded:
1994.
Pricing was an essential part of competitive success.

Due to the skills, experience and dedication of the retailer's experts, the chain's pricing was efficient despite poor automation.
Over 28 years, the retailer's team developed a large, partially decentralized system of complex pricing rules, utilized following parts.

Category level and specific item level pricing matrixes which determined sets of vary markups.
Competitor price tracking.
KVI pricing.
A system of continuos promo actions, when the end date of the action depends on the promo performance and inventory balances.
Manual price adjustments at the category managers level.
Manual price adjustments at the store level, based on the competitor tracking results.

Partially automated, this pricing system ensured good sales and profits, and helped to effectively compete with larger retail chains.
The retailer's team and Imprice experts faced a serious challenge, to completely rebuild an effective pricing system without spoiling anything, and to find points of growth, while having an already "high base".
APPROACH:

AI-powered Analytics

AI-driven assortment-level price optimization

KVI pricing
The First Pilot. Non-Food Items With A Relatively Low Price Sensitivity
The pilot test steps were as follows:

1 - Pricing experts chose one test and one control hypermarkets.
The hypermarkets were located in the same city and had very similar sales performance dynamics.

2 - They decided to run the first pilot test with only non-food SKUs which didn't drive price perception, that is, foreground and background (long tail) groups.

3 - Then they ran the imprice clustering algorithm.
The algorithm is an AI-driven analytical module that accurately measures goods' cross-impact and identifies the role of each SKU in the consumer basket.

4 - The clustering algorithm automatically segmented the assortment by item roles in the consumer basket: hard KVI, soft KVI, Foreground, Long Tail, complementary goods, substitutes, and others.

5 - Then a pricing specialist created special segments in Imprice for non-food items from Foreground and Long Tail groups. These segments counted more than 10,000 SKUs. Imprice AI-driven, demand-based price optimization module began to manage prices of these segments.

6 - For one part of SKUs, experts set the price optimization goal as "to find prices at which the store will gain the maximum possible revenue." For the other part of SKUs, the optimization algorithms had to "calculate prices that would ensure the maximum possible gross profit."

The retailer's experts defined a price range permitted as
(X-10%; X+15%),
where X is the price calculated with the "old" semi-automated way of the store chain's pricing.

7 - Artificial intelligence found underpriced goods in the Foreground and Long Tail segments. In fact, consumers were willing to pay more for these items. Therefore, at the new, higher price, demand (sales volume) remained the same.
That is, at the "old" price, the store was losing profits. Algorithms have raised the prices of these underpriced goods to the optimum level.
Example of an underpriced item — cassettes Gillette from the Long Tail segment.
The algorithms identified that the demand for this item was inelastic. In other words, sales volume did not change while price changed, remaining within 0.3-.0.4 pieces per day. Therefore, by raising the price, the hypermarket increased revenue and gross profit, while maintaining the volume of sales in pieces.
More details are on the screenshots from the Imprice platform below.
Screenshot 1. Dependence between gross profit of Gillette cassette and its price, regression plot.
Screenshot 2. Dependence between sales volume in pieces of Gillette cassette and its price, regression plot.
Screenshot 3. Gillette cassette price changing plot.
Before the pilot, the price did not change for around a year.
The algorithm started price exploring on June 29, 2021.
First, it lowered the price, and then started to raise the price, before hitting the optimal price level two months later.
Note:
Exploring the prices of items from Long Tail and Foreground groups takes a long time. The reason is these items are rarely bought, therefore, each price has to be displayed for at least 3 weeks to accumulate sufficient statistics.
8 - Similarly, the algorithms identified overpriced items. Most consumers were not willing to buy at these prices.
The platform lowered such prices to the optimum level, and sales of these SKUs increased significantly.

9 - The AI-driven module optimized prices at the product category level. In particular, automatically considering if SKUs were substitutes, the algorithms prevented cannibalization — a situation when a decrease in the price of an item cuts down sales of its substitutes, and the store loses profits.

10 - Promotional pricing remained managed by category managers.
If managers included in the promo action an item from a piloting segment, the platform automatically fixed its price (set it equal to the promotional price) until the end of the promotion.

11 - The retailer's experts defined a constraint: the hypermarket staff could change no more than 500 price tags per day.

Therefore, they ran Imprice automatic schedule of price recalculation. The "intelligent queue" is an analytical tool that determines which prices have to be changed first and which price changes can wait.
There were also special days, when the platform did not send new prices to the test hypermarket. For example, it could be a day of promos changing.
On other days, the platform automatically sent a list of up to 500 items with new prices to the store.

12 - The first pilot was completed in 3 months.

The retailer's team assessed the test as successful and decided to expand the experiment by launching a second pilot.
APPROACH:

AI-powered Analytics

AI-driven assortment-level price optimization

KVI pricing
The Second Pilot. Food Items From Foreground and Long Tail Groups. Running Dynamic Pricing In A Supermarket
Generally, the second pilot test was similar to the first one, but now the retailer launched dynamic pricing for more categories and in two types of stores.

a) The team added other "Foreground" and "Background / Long Tail" categories to the experiment, including food items.
Artificial intelligence started to optimize prices of two food categories, confectionery and groceries, and the prices of household chemicals.

b) The Imprice platform now was optimizing prices in two different store types, a hypermarket and a supermarket.
The test supermarket's performance was compared with a control supermarket's results.





The second pilot test was completed in 3 months.
The retailer's team standardized and improved all processes related to pricing.

Compared with sales results of control stores, sales and profits performance of pilot stores showed the success of the second pilot.

Therefore, the retailer launched the third, final pilot.
APPROACH:

AI-powered Analytics

AI-driven assortment-level price optimization

KVI pricing
The Third Pilot. Launching Dynamic Pricing Across The Entire Assortment. KVI Pricing
With the first two pilots, the retailer step by step shifted to fully automated pricing without a risk of making crucial pricing mistakes.
During the implementation phase, the team standardized and strengthened their pricing approaches, studied the capabilities and advantages of the platform, and gained confidence that a data-driven approach ensured better results.

The final pilot aimed to scale automated pricing to the entire assortment, and to estimate accurately the effect of dynamic pricing implementation.

The team launched dynamic pricing for another 2,000 SKUs. Thus, the algorithms calculated prices of almost the entire assortment, except for fruits, vegetables, meats and self-manufactured meals. In particular, categories "canned food", "soft drinks", "frozen vegetables and fruits", "dairy products", "pet products" were switched to dynamic pricing.

With Imprice, they automated KVI pricing.
The platform automatically recalculated prices of KVIs and maintained the aimed competitive price position.
Since the Imprice team made a system integration of the platform and retailer's vendor, competitors' prices automatically uploaded into Imprice from the retailer's price monitoring platform.
In order to improve data quality and completeness, the retailer also launched additional price monitoring by Imprice.

The retailer's pricing experts handled the promotions and the prices of a few specific categories in the test hypermarket and supermarket.

The final pilot test was completed in 5 months.

Then the chain's specialists compared the performance of pilot stores that utilized dynamic pricing to the results of control stores that utilized an "old" semi-automated pricing approach.

Experts removed sales of promotions and sales of categories with "manual" pricing from the experiment results and cleaned data from the influence of other factors.

The key result:

+2% gross profit growth
in the last five months of the pilot, when the Imprice platform automatically managed regular prices for most categories of goods
(the test stores' result compared to the control stores).
SCALING:

Across 47 hypermarkets and supermarkets of the chain
Gross profit growth. Rollout
"We have been improving and automating our pricing for almost a year.
For example, standardizing data of promotions for fully automated uploading into the platform took five months, since before we manually determined and changed the range and periods of promo actions, both at the level of category managers and at store levels. So the task of setting up the process of exchanging data of promos with all these manual adjustments was very challenging.

With Imprice, our efforts led us to fast returns in efficiency and ROI.

+2% gross profit growth was a direct result of pricing automation.

Moreover, we got further enhancements and profit.

Now price calculations are really transparent and based on data, not on intuition. Advanced analytics tools and transparency of pricing rules empower us to make pricing decisions easier and accurately evaluate the results.

The platform calculates prices fully automatically, so our category managers don't have to spend time on routine operations. Automation and full transparency of pricing rules helped us to pass without stress and overwork a challenging period of supply instability with surging supply costs.

Advanced analytics tools help us identify business "blind spots" that are not related to prices directly, but crucial for sales.

For example, we replaced a few checkouts with self-service checkouts in the pilot hypermarket.
Thanks to Imprice analytics, we found out this replacement had a negative impact on the sales of chocolate and chewing gum, which were displayed in the checkout area. The problem was that self-checkouts did not have any area to place items.
The information helped us to solve the problem. We displayed chocolate and chewing gum close to self-checkouts and regained our sales of these categories.

With Imprice, we revealed On-shelf availability (OSA) problems. Some items were replenished on the shelf several days late, even though these items were in storage.
That is, the goods in the shopping area were sold out, but staff couldn't track all 35,000 SKUs, and to restock shelves on time. Obviously, if an item is unavailable on the shelf, its sales drop to zero, and the store loses profits.
We have fixed the problem with a special Imprice report titled "SKUs with suspiciously stopped sales", which is automatically emailed to the store manager every day.

The results of the pilots showed that we have quickly recouped the cost of our investment in automation.

In April 2022 we ran a rollout to implement dynamic pricing across the entire store chain."

— Business analyst of the store chain
Talk to Imprice pricing experts: