Hypermarkets increase gross profit by 6,6% with AI and KVI pricing

CUSTOMER
PROFILE:

Location:
● Kyrgyzstan, Central Asia.

Industry:
● Grocery.

● 47 brick-and-mortar stores.

● 25,000 SKUs.

● Private label.

● Home delivery service.

● Mobile app.

Founded:
● 2011.

gipermarket.kg

FRUNZE is a leading grocery chain in Kyrgyzstan.

To maintain the leadership position and improve customer experience, the company was always ready to bring pioneering innovations: home delivery service, mobile app, self-service checkouts.

In July 2022, FRUNZE launched a pilot of dynamic pricing in three test hypermarkets, comparing their business results with six control hypermarkets.

Key results

In November 2022, the retailer's team compared sales performance of pilot hypermarkets with control hypermarkets performance.
The results of July — October 2022:

Revenue
growth

Gross profit
growth

CHALLENGES:

● Launch fully automated pricing.

● Strengthen the competitive position.

● Hit at least +2% growth in revenue and gross profit.

Description of the pilot

For the pilot, pricing experts chose 9 hypermarkets:

✓ Three pilot hypermarkets where they launched dynamic pricing with Imprice.
✓ Six control hypermarkets, whose sales dynamics for the last year were most similar to the dynamics of pilot stores.

During the pilot, FRUNZE pricing specialists were calculating prices for these control hypermarkets.

Step 1. Pilot and control stores selection

In June 2022, the system integration of FRUNZE's ERP with Imprice platform was complete.
Historical data on assortment, promos, sales, stocks, cost prices, goods supply and so on had been transferred to Imprice. New data began to load into the platform automatically.

Before the pilot, the retailer utilized competitive pricing.
The pricing team used the Top 300 (in terms of revenue) instead of KVIs list, and calculated the Top 300 prices considering competitors' prices.
In competitors' brick-and-mortar stores, prices for goods included in the top 300 were gathered manually. Online prices of key competitors were monitored automatically.

The retailer stored price monitoring data in its ERP; also this data began to load into Imprice automatically from June 2022.

Step 2. Integration

After the data was added into Imprice, machine learning algorithms calculated how each SKU impacted on total revenue and on sales of other items, and segmented all items by their roles in customer baskets. The result was the following segments:
– hard KVIs,
– soft KVIs,
– Foreground items,
– Long tail,
- ...

Later, pricing experts updated clustering once a month, running ML algorithms again. Such updates of goods' roles help:
a) Identify "hidden" KVIs that could not previously be revealed due to lack of data.
b) Detect goods that have changed their roles in the basket; for example, they have turned to KVIs or, overwise, have stopped being KVIs.

During the pilot, the algorithms found 65 Hard KVIs and about 1000 Soft KVIs, analyzing around 20,000 SKUs.

Step 3. Clustering

➜ Hard KVI is a product with ultra-high price sensitivity.
Often, only the "first minimum price" strategy is efficient for such products, that is, only the lowest price on the market helps maximize revenue.

➜ Soft KVI is an item with relatively high price elasticity of demand.
At the same time, the maximum revenue for such items can be hit at the third, fourth and even fifth market prices.

Imprice algorithms revealed that:

● Around 25% of the "top 300 goods by revenue" list were not KVIs. When the price of these items rose, demand remained relatively unchanged.
Thus, cutting prices of these SKUs to the competitors' prices level, the company was losing profits.

● On the other hand, around 75% of real KVIs were not in the "top 300" list. The prices of these goods were higher than in competitors' stores.
Due to that, FRUNZE hypermarkets were losing part of their sales and looked "expensive" compared to competitors.

After clustering, the retailer started to monitor "offline prices" of items from the KVIs list.

APPROACH:

● AI-powered Analytics.

● AI-driven assortment-level price optimization.

● KVI pricing.

Due to increasing the competitive pressure, the retailer's team first decided to test the most aggressive type of KVI pricing strategy.
The essence of the strategy: for both Hard KVIs and Soft KVIs, the Imprice platform began to automatically set the minimum price on the market, lower than the key competitors' prices (if this minimum price met restrictions of margin set by experts).

For other goods' roles, such as Foreground or Long tail, they launched ai-driven price optimization, based on demand sensing and chose the maximum of gross profit as an optimization criterion.

The Imprice platform managed prices of around 20,000 SKUs, which was 60% of the revenue and 65% of the gross profit of the pilot hypermarkets.

Step 4. Low season, "first minimum price" strategy for KVIs and demand sensing

For the pilot hypermarkets, FRUNZE pricing experts continued to manage promo prices and prices of a few categories, for example, cigarettes and ultrafresh products.

When evaluating the effectiveness of dynamic pricing, experts removed from the calculations sales data for the promos and "remained categories."

To keep within these daily limits of price tags, the retail chain used the Intelligent Queue, an analytical tool of Imprice that determines which prices have to be changed first and which price changes can wait.

On July 4, 2022, the Imprice system started to automatically calculate prices and send them to pilot hypermarkets.

Important detail.
Pricing experts of the grocery chain set upper limits on the quantity of price tags changing daily:

✓ Prices could be changed only on weekdays.
✓ On Tuesday, it was allowed to change <=700 price tags.
Tuesdays were days of suppliers' prices adjusting and cost prices changing.
✓ On other weekdays, it was allowed to change <=150 price tags only.

PILOT STEPS:

1 — Pilot and control stores selection

2 — Integration.

3 — Clustering.

4 — Low season, "first minimum price" strategy for KVIs and demand sensing.

5 — New pricing strategy for KVIs and price optimization for Long Tail and Foreground.

Buyers reacted quickly to the reduction of KVIs' prices: sales of the entire assortment rose sharply within the first 2 weeks. The number of cash receipts also increased.

Such a rapid reaction to prices is rare in grocery retail.
It usually takes several weeks to change the shopping behavior of a significant part of the consumers.

Pricing experts compared LFL of pilot with LFL of control stores. An aggressive competitive strategy boosted sales in pieces in pilot hypermarkets.
However, the increase in revenue was insignificant, and gross profit performance even slightly weakened.

Why it happened:

a) Significant decrease in the sales margin of KVIs.

b) Low season.

c) The first weeks of the pilot are the stage of the most active price exploration by AI algorithms.
If historical data does not contain enough price statistics for price optimization, then it requires an extended set of pricing experiments. Within a set price range, the algorithms change prices, measure the elasticity of demand, and calculate the optimal price — in this case, the price that ensures the maximum gross profit for the store.
Before the start of the pilot, FRUNZE rarely changed prices for most items of the assortment; therefore, the algorithms needed a significant amount of price experiments to collect statistics.
Commonly, business results slightly decrease within the price exploration period.

After increasing the number of cash receipts, that is, purchase frequency, pricing experts started testing a new hypothesis:
➜ Maintaining KVI prices lower than prices of key competitors is not required to retain customer loyalty.
➜ KVI prices are equal to competitors' prices, or even to the second or third minimum price of the market is an efficient pricing strategy to maintain high sales.

From the fourth week of the pilot, the retailer's experts slightly increased prices of KVIs.

The hypothesis was confirmed. With the new, less aggressive KVI strategy, sales in pieces and the number of cash receipts in pilot hypermarkets remained quite high. At the same time, pilot stores started to surpass control stores by both revenue and gross profit.

Step 5. New pricing strategy for KVIs and price optimization for Long Tail and Foreground

By this time, AI algorithms had completed their most active price exploration phase. Prices of non-KVI goods neared optimal ones. This caused an additional increase in profits and revenue.

SCALING:

➜ Across 47 hypermarkets and supermarkets of the chain

ROI of pricing automation.
Rollout

The pilot lasted 4 months, from July to October 2022.
Dynamic pricing results in pilot stores were as follows:

+5.5% — revenue growth

+6.6% — gross profit growth.

In monetary terms, the increase in gross profit in the three pilot stores was higher than the Imprice subscription fee. Thus, the launching of dynamic pricing paid off even at the pilot stage.

As the business targets of the pilot tests were surpassed, the retailer's team started the rollout proceeding, that is, dynamic pricing implementation in all stores of the FRUNZE grocery chain.

By December 2022, dynamic pricing with Imprice worked in almost all clusters of FRUNZE stores.

Inna Neverova
Head of Database and Pricing Department
FRUNZE grocery chain

"FRUNZE is an innovation-oriented company, and we're confident in our ability to improve stores' results with innovations.
Dynamic pricing also met our expectations: we strengthened our competitive position, as evidenced by the growth in the number of cash receipts, sales growth and +6.6% gross profit increasing.

In general, Imprice is the most comfortable team among all vendors we have worked with:

➜ Strong expertise.

➜ Always on touch. Almost every day we contact the Imprice team to discuss pricing strategies and their changes, the platform capabilities, and how we can implement what we want with the Impice platform.

➜ Stores always receive correct prices. If there are any problems with the data or their exchange, including from our side, the Imprice team immediately reacts and finds a solution.

➜ "Single window"; we don't have to communicate with ten managers, each of which is responsible only for his own question.

➜ Helping and advising. If we have a question, they answer us and do not force us to study the platform on our own and "read the documentation".

One of the reasons for our success in this project and the growth of our profits is that Imprice is a customer-centric company; it is a partner for true teamwork."