When e-commerce surged from 2% to 15% of Kroger's sales, they faced a crisis. Pickers were walking 4+ miles per shift, criss-crossing stores inefficiently. Pick lists were sorted alphabetically—not by aisle. And when items were out of stock, substitution logic often suggested items customers didn't want.
4.2 mi
Walking per shift
34%
Sub rejection rate
23%
Late deliveries
Average Pick Path
Before
4.2 miles/shift
After
2.1 miles/shift
Pick Time (30 items)
Before
45 minutes
After
24 minutes
Items per Hour
Before
40 items
After
75 items
Wrong Item Rate
Before
3.2%
After
0.8%
Rules-based substitution picked the cheapest alternative. AI substitution understands individual customer preferences.
Requested: Organic whole milk, 1 gal
2% milk, 1 gal (cheaper)
Organic 2% milk, 1 gal (same brand, organic)
Customer's purchase history shows organic preference is more important than fat content
Requested: Honeycrisp apples, 3 lb bag
Red Delicious apples, 3 lb bag
Honeycrisp apples, loose (same variety, different packaging)
Customer consistently buys Honeycrisp, previously rejected Red Delicious sub
24 min
Pick Time (30 items)
down from 45 min
94%
On-Time Delivery
up from 77%
+81%
Orders/Picker/Hour
3.2 to 5.8
-42%
Fulfillment Cost
$8.40 to $4.90
"We didn't build new fulfillment centers—we made our existing stores smart. Pickers are happier walking half the distance. Customers get substitutions they actually want. And we cut fulfillment costs by 42% while improving customer satisfaction."
Tom Harrison
VP of E-Commerce Operations, Kroger
We help retailers build AI that makes existing infrastructure smarter— without massive capital investment.
Talk to our AI systems architects about your specific challenges. Get a personalized roadmap for implementation.