top of page
Anushka Kher
SENIOR TECHNICAL PRODUCT MANAGER, AMAZON
01 // LOGISTICS & SYSTEM
Grocery Order Reschedule
AMAZON FRESH . SHOPPING. FULFILLMENT. LOGISTICS. CONSUMER EXPERIENCE.
The Problem
SECTION 01 // PROBLEM
Every year, Amazon Fresh auto-cancels hundreds of thousands of grocery orders that customers still want.
On Amazon's side, when a fulfillment center hits capacity constraints: a winter storm knocking out power, a printer failure halting label production, a network-wide outage overwhelming picking capacity, the system's only response is blunt: cancel the order.
On the customer's side, when their own plans change after checkout: a late meeting, not being home for the window, there's no way to move the delivery themselves. Their only option is to cancel and start the entire order over from scratch.
The delivery promise is rigid on both sides, groceries are still intact.
The customer can't keep the promise when their own plans change after checkout, a late meeting or not being home, with no way to move the delivery themselves, only to cancel the order and go through entire process again to reorder.
NUMBERS THAT DEMANDED ACTION
1.64%
VANISHING ORDER
Auto cancelled by Amazon every year
$55.4MM
INVENTORY & OPERATIONAL LOSS
wasted yearly to auto cancellations
#3
CAN I MOVE MY SLOT?
3rd largest customer service complaint
1.5%
SELF CANCELLED
1.5% of all grocery orders, because slot couldn't move
Persona and User Quote
SECTION 02 // PERSONAS
CHARACTERISTICS
-
Monitors site health dashboards across the network in real-time
-
Makes rapid triage decisions during large-scale events (LSEs) affecting thousands of orders
-
Coordinates between site associates, drivers, and central planning
-
Has deep knowledge of which orders can still be salvaged vs. which have passed the point of no return
FRUSTRATION
When an LSE hits, the only tool available is mass cancellation. There is no mechanism to selectively save orders. He watches deliverable orders get cancelled, knowing the groceries are fine and just need a new window
Carlos
Grocery Fulfillment Operations Lead
Role: Central Operations Manager, Amazon Fresh
Context: Manages order flow across 14+ fulfillment sites. First responder when capacity constraints hit.
CORE USER STORY
"The snowstorm knocked out our fulfillment and delivery systems, but the packed orders are still good, sitting on the shelves and in the freezers, cold and safe. I just need a way to get them to customers once we're back up — and to tell those customers their delivery is moving, not gone."
CHARACTERISTICS
-
Get her weekly groceries reliably, around a schedule that shifts without warning
-
Keep temperature-sensitive items (frozen, chilled, dairy) safe, never left on the porch
-
Checks order status on her commute. Stay in control of household routine
FRUSTRATION
A late meeting means she can no longer be home for her 5–7 PM delivery window and has no way to change the slot herself. Her order has perishables that can't sit on the porch until she's back. Her only options are to cancel the existing order and rebuild it from scratch to get a new slot, or call customer support (CS) and hope an agent can find another way.
Maya
Working Parent and Amazon Fresh Subscriber
Role: Weekly grocery customer, Amazon Fresh
Context: Orders every Sunday for Monday evening delivery. Relies on Amazon Fresh for perishables her schedule won't let her shop for
CORE USER STORY
"I've got people coming for dinner tonight and everything I need is in this order - the chicken, the fresh broccoli, the ice cream for dessert. My delivery's set for 5, but my meeting just got moved and I won't be home till 8. I don't want to cancel and lose it all. I just need it to come a few hours later."
The Solution: Amazon-initiated Re-promise
SECTION 03A // PROCESS
1
STEP 1
Trigger Detection
Fulfillment Center hits site capacity and labor constraints due to an outage, resulting in site capping and delay in processing orders.
2
STEP 2
Identify Salvageable Orders
Site Ops and Central Operations identify impacted orders and initiate same-day re-promise for orders to a future window selected based on driver availability. This covers both pre-pack orders (not yet picked) and post-pack orders (packed, on staging wall).
3
STEP 3
Apply Guardrails
Order re-promise system filters out ineligible orders: Orders that have left the staging wall and are in delivery handoff or Orders past food safety chill-chain thresholds (chill/frozen items left temperature-controlled zone) or Orders where packages have been batched for pickup
4
STEP 4
Find New Window and Replan Capacity
Re-promise orchestration finds a slot within time range provided by operations, then replans capacity for that window: reserving pick/pack, batching, driver routing so the order is deliverable. Order contract updated. Site notified about the new delivery promise.
5
STEP 5
Preserve Fulfillment State
For already-packed orders: System does not replan pick/pack (order is already packaged). Same packages reused for delivery against new window. Existing labels reused (no reprint required, saving labor during capacity-constrained events).
6
STEP 6
Notify Customer
Customer informed of new slot via notifications (SMS, push, email). Order Details page displays new slot as banner. Customer can cancel through CS or self-service if new slot is inconvenient.
7
STEP 7
Deliver
Order batched in new window, handed off to driver, delivered to customer.
WAREHOUSE STORAGE SHELVES
GROCERIES
The Solution: Customer-initiated Slot Modification
SECTION 03B // PROCESS 2
1
STEP 1
Customer Requests Change
Customer opens Order Details in app or web. Sees "Change delivery time" option available before shopper starts working on order.
2
STEP 2
Display Available Windows
Customers can choose from available slots in real-time. Filters to same fee or lower delivery options only. Shows only windows the fulfillment center can realistically execute.
3
STEP 3
Find New Window and Replan Capacity
System releases the customer's original delivery slot capacity, making it available for new orders and simultaneously reserves pick/pack, batching and driver routing capacity in the newly selected window. Order contract updated to reflect the new promise.
4
STEP 4
Confirm
Confirmation notification sent. Order Details page updated. Order progress tracker reflects new slot. Customer can still add items or make further modifications until new pick-start cutoff.
USER ACTION
Two taps, slot moved
OUTCOME
Home in time, groceries in hand
Product Evolution and Roadmap
SECTION 04 // ROADMAP
PHASE 1 - MVP
Scope: 3-week controlled experiment at 2 Fresh fulfillment centers.
What we shipped:
-
When a fulfillment center or weather outage forces site operations to shut, rescheduling order to future window and notify the customers. Experiment in two pilot fulfillment sites.
RESULTS
70%
customers accepted the next-day slot
65%
Fewer cancellations
per event.
PHASE 2 - FULL LAUNCH
Scope: Auto-reschedule as the standard outage-mitigation response for pre-pack orders, scaled from a single-market launch to worldwide.
What we shipped:
• Launched auto-reschedule across all US fulfillment centers as the default response to outages.
• Scaled worldwide across North America, United Kingdom, Japan, and India.
• Evolved capability into a self-service tool for the operations team
• Scaled the capacity replanning layer across all types of grocery businesses (Amazon Fresh, WFM, 3P grocers).
RESULTS
15K
orders auto rescheduled in
Year 1
17%
Reduction
in cancellation
$1MM
Annualize savings
in operational costs
PHASE 3 - EXTEND CAPABILITY TO POST PACK ORDERS
Scope: Extend reschedule capability from pre-pack orders to orders that are already picked, packed, and waiting on the staging wall, validated through experimentation.
What we shipped:
• Launched post-pack rescheduling for packed orders, reusing the existing packages and labels
• Added food-safety guardrails that exclude any order whose chilled or frozen items have left temperature-controlled staging to guarantee chill-chain integrity
RESULTS
$31MM
Annual Sales
$6.33MM
Contributing Profit in Inventory Savings
PHASE 4 - CUSTOMER FLEXIBILITY
Scope: Launch self-service delivery window change for customers post-checkout.
What we shipped: Launched self-service window modification in the Amazon experience. Post-checkout slot picker shows slots the fulfillment center can actually execute, backed by real-time release of the original slot's capacity and reservation in the new one. Customers keep their items, pricing, and payment transactions.
RESULTS
$84MM
Annual Sales
$3.7MM
Contributing Profit in Concessions
$1.2MM
Contributing Profit in CS Contact Reduction
bottom of page