What to Expect in the Walmart Global Tech System Design Interview

Walmart Global Tech's system design interview draws from the largest physical-plus-digital commerce operation ever built: e-commerce at Black Friday scale (the burst benchmark of the industry), inventory truth across 10,000-plus stores and a digital catalog simultaneously (the omnichannel consistency problem in its hardest form), and supply-chain systems moving more goods than any network on earth. The rounds split by altitude (low-level design of retail components; high-level design of distributed commerce systems), and the register rewards scale arithmetic anchored in retail's real rhythms: the holiday curve, the store-day cycle, and the physical-world constraints (trucks, shelves, pickers) that pure-digital candidates forget.

The Question Territory

• Design omnichannel inventory. The signature problem: a single inventory truth serving stores (POS decrements), e-commerce (promise-to-order), and pickup/ship-from-store flows: per-store-per-SKU state at hundreds-of-millions-of-pairs scale, reservation semantics across channels (the last unit, wanted online and in-aisle simultaneously), and the accuracy problem (physical counts drift; the system must promise conservatively and reconcile continuously).
• Design cart and checkout for Black Friday. Burst engineering's classic: traffic at 20-to-50x baseline with doorbuster spikes on top: pre-scaled capacity on the known calendar, queue-based absorption with honest waiting rooms, inventory contention on hot items (the hot-key playbook applied to televisions), and the degradation ladder (recommendations dim, checkout never does).
• Design order management and fulfillment routing. An order's life across fulfillment options (warehouse, store-ship, pickup): sourcing optimization (which node fulfills, weighing distance, stock, and labor), split shipments, and the state-machine discipline of orders that touch physical operations.
• Design supply-chain visibility or replenishment. Forecast-driven replenishment per store-SKU, truck and DC scheduling, and event ingestion from the physical chain (scans, receipts, exceptions) at national scale.
• LLD variants: an inventory-reservation component, a pricing engine with rollback-priced items, a store-pickup slot scheduler: object modeling with retail texture.

What Interviewers Are Probing

1. Inventory-truth honesty. The domain's core insight: store inventory is physically uncertain (theft, misplacement, breakage), so designs that promise from buffered availability (on-hand minus safety threshold), reconcile against scan events continuously, and handle the oversell gracefully (substitute, refund, apologize: as product design) demonstrate retail fluency generic e-commerce answers lack.
2. Burst arithmetic on the known calendar. Black Friday is scheduled: pre-scaling with load-tested headroom, waiting-room queues with honest positions, and hot-SKU contention handled (per-item inventory sharding or token allocation for doorbusters) with the reasoning stated: the Intuit tax-season register at retail rhythm.
3. Physical-digital seam realism. Ship-from-store means a picker walking aisles against live shoppers; pickup slots mean labor capacity, not just database rows. Designs acknowledging the physical constraint layer (and feeding it back into digital promises) read as omnichannel-serious.
4. Consistency budgets by money-proximity. Checkout and payment strongly consistent; browse availability eventually consistent with staleness bounds; the partition reasoned explicitly: the enterprise commerce canon.
5. Scale numbers with retail texture. SKU counts in the hundreds of millions of store-item pairs, order volumes at holiday peak, event streams from every register in America: candidates who estimate in the domain's actual units stand out.

Walkthrough Sketch: Omnichannel Inventory with Buy-Online-Pickup-In-Store

Requirements first: per-store, per-SKU availability serving three consumers: in-store POS (decrements at scan speed), e-commerce browse and promise (reads at national scale, promises that must survive to pickup), and store operations (receiving, counts, corrections): with the honest constraint stated up front: physical on-hand is an estimate, so the system's promises must be conservative and self-correcting.

The data model: per store-SKU, an availability record: on-hand estimate, reserved (promised to orders), and a safety buffer tuned by SKU volatility (high-shrink categories buffer more): with sellable = on-hand minus reserved minus buffer as the only number e-commerce ever sees. Writes: POS scans and operational events stream through per-store event logs into the availability service (regional sharding by store; a store's inventory is naturally partitioned and local-write-heavy), while e-commerce reservations run as atomic sellable-decrements with order-scoped holds that expire if checkout stalls: the reservation semantics doing the cross-channel arbitration. The BOPIS flow proves the design: a pickup order reserves at a specific store, store systems receive the pick task, and the seam handling is explicit: if the picker cannot find the item (physical truth disagreeing with digital), the exception flow triggers substitution or refund plus an inventory correction event that tightens that SKU's buffer: the system learning its own uncertainty. Browse-scale reads: sellable projections cached regionally with bounded staleness (seconds, not minutes, for pickup-eligible views: staleness bounds by use case), and hot-item invalidation on threshold crossings (the last-few-units state pushes rather than waits). Black Friday hardening: doorbuster SKUs get token-allocated inventory (a fixed sellable count distributed as claim tokens through the waiting room: contention converted to allocation), and the degradation ladder protects reserve-and-checkout above all. Reconciliation closes the loop: nightly counts and continuous scan-versus-estimate drift detection feed buffer tuning per store-SKU, and the accuracy metric (promise-kept rate at pickup) is the system's real grade: because in omnichannel retail, the database being right means a family's pickup order is actually on the shelf.

How to Prepare

• Foundations and depth: Grokking the System Design Interview and Grokking System Design Fundamentals for the method and blocks; Advanced System Design Interview, Volume II for the sharding, consistency, and burst depth these rounds reach.
• Rehearse the two house designs: omnichannel inventory (buffers, reservations, the physical seam) and Black Friday checkout (waiting rooms, hot-SKU allocation, degradation ladders).
• Learn retail's rhythms: the holiday curve, store-day cycles, and shrink reality: an evening of domain reading that converts generic scale answers into retail-fluent ones.
• LLD reps with retail texture: an inventory-reservation component modeled cleanly, twice.

For the full loop, see What is the Walmart Global Tech interview process like?, and prepare the values dimension with Top Walmart Global Tech behavioral interview questions and your answer to "Why Walmart Global Tech?"