Continuous Improvements of Price Promotion KPIs via artificial intelligence

The retailer:

The retailer is one of the leading German grocery discounters which has over five billion Euro yearly revenue and more than 2.000 stores. The retailer has no previous loyalty card in place.


This retailer had previously focused on non-personalized mass marketing and some targeted couponing based on sociodemographics and simple heuristics. Said activities saw low redemptions, limited margins, and frequently replaced regular sales.

The retailer wanted to establish an attractive new loyalty card program – offering highly efficient, personalized promotions without having to take into account any extensive data protection issues.


SO1 set up an anonymous loyalty card which was issued at the checkout without registration. Most importantly, SO1’s Optimized Discounts solution provided automated, personalized and optimized promotions in real-time to each loyalty card holder. Powered by SO1’s advanced AI for retail promotions, customers received relevant offers with individually adjusted discounts when scanning their loyalty card at a kiosk at the store entrance. Discounts were granted in % Euro and directly subtracted from the retail sales price when redeemed at the POS.

Primary Pilot Setup and Results:

SO1 was launched in 60 stores across Germany. A 12 week pilot phase was preceded by a joint integration phase where data transmissions and preparations, including the training of SO1’s algorithms on the retailer’s historical data (one year), took place.

SO1’s personalized promotions were well accepted by customers (23% basket penetration, of which 52% regularly printed their individual offers at a kiosk in the store entrance). In the first week, SO1 increased redemptions from 5% to 42% thanks to its upfront training, while simultaneously reducing the discounts from 25% to 22%. After 12 weeks the redemption rate had risen from 42% to 51%, and the discount had dropped to 18%, thanks to SO1’s self-learning algorithms and integrated features.

Fig. 1: Continuous improvements in terms of redemption rate and avg. discounts applied, thanks to SO1’s artificial intelligence and its self-learning algorithms

Pilot Extension:

The pilot was extended to 40 weeks and expanded to 100 stores to explore the interdependence of applied discounts and customer acceptance. For this A/B test, participating customers were randomly assigned to different cohorts. The first optimized exclusively for customer loyalty (high offer redemption rate), the second for profit (low discounts), to show which effect one KPI would have on the other.

Figure 2: Redemption rate and avg. discount for two main cohorts (Cohort 2 maximized redemption rate, Cohort 4 focussed on reducing the average discount)


The cohort test (see Figure 2) showed that average discounts could be reduced to 8%, though redemption rates would drop to 55% as a consequence (cohort 4). Vice versa, redemptions could be increased to 63%, while requiring average discounts of 15%.

Overall (see Figure 1), SO1’s artificial intelligence was able to continuously learn about consumer behavior and thus optimize financial KPIs. On average, the redemption rate was improved from 42% to 60%. The discount was reduced from 22% to 13% – thus saving 48% off promotional expenditure compared to previously applied discounts in the retailer’s weekly leaflet. With SO1, this grocer achieved an incremental sales increase of 13% per targeted basket and a growth of 1.6% at the store level.