Fulfillment scenario backtesting is the practical way to decide whether inventory should stay in China, move closer to demand, or use a hybrid model: replay past orders under the same SKU, cost, lead-time, and service rules before committing stock. It will not predict the future perfectly, but it makes the trade-offs visible before a warehouse move creates aged inventory, stockouts, or an unworkable returns path.
The expensive mistake is often not choosing China-only or destination-country stock. It is making that choice from a blended shipping quote or a single fast-moving month. For high-value, fragile, or complex-SKU goods, Bondjet 邦达捷 can help make the China-side part of the model more reliable through receiving, inspection, SKU control, protective packaging, consolidation, and international dispatch. A destination-country warehouse, however, must be modeled only after its operator, service scope, systems, fees, and returns process are confirmed.
Key Takeaways
- A useful fulfillment scenario backtest replays order-level history under fixed China-only, local-stock, hybrid, and staged-launch rules rather than comparing average parcel prices.
- Track eight outputs together: service-level attainment, contribution margin, landed cost, stockouts, inventory days, transfers, aged stock, and return cost.
- Build constraints into the model first: available inventory, replenishment lead time, transfer minimums, capacity, order-routing rules, and the real returns path.
- Test at least five sensitivities: demand error, peak demand, freight changes, return-rate changes, and delayed replenishment.
- Bondjet's documented services support China-side fulfillment controls; do not model a Bondjet-operated US warehouse or US returns center unless that specific arrangement is confirmed in writing.
What Fulfillment Scenario Backtesting Can Answer
Fulfillment scenario backtesting uses completed orders as a controlled history. For each order, the model asks: if this order had followed another inventory-placement rule, could the selected node have fulfilled it, how much would it have cost, what service promise could it have met, and what inventory would remain afterward?
This is more useful than asking whether a US warehouse is "cheaper." A local node may reduce final-mile time for a hero SKU while increasing transfer cost, storage exposure, and stranded stock for the rest of the catalog. A China-only model can preserve flexibility for new or long-tail products, but may not meet every delivery promise. The backtest forces both sides into the same definition.
Use it to answer four operational questions:
- Which SKUs have enough repeatable, market-specific demand to justify pre-stock?
- Which SKUs should remain available from China until demand is proven?
- At what demand error or replenishment delay does a local-stock decision stop working?
- Are the destination warehouse and returns workflows actually ready for the customer promise?
[Image suggestion: a timeline that replays one SKU's weekly orders, inventory receipts, transfers, stockouts, and returns under three inventory-placement scenarios.]
What it cannot prove
An inventory placement backtest is a decision aid, not a guarantee. Historical orders do not reveal an unlaunched promotion, a new competitor, a carrier disruption, a customs hold, or a warehouse partner's future execution. It also cannot turn an unconfirmed service into a capability.
Mark uncertain inputs clearly. If the local warehouse's inbound appointment rules, receiving SLA, domestic pick-and-pack charges, or returns disposition are unknown, the output should show a range or a blocked assumption. That is more credible than presenting a precise result built on a missing operating model.
Assemble Order-Level Data for an Inventory Placement Backtest
Start with the smallest data set that can replay a real order. Monthly averages are useful for a dashboard, but they hide the sequence that creates a stockout. A SKU allocation scenario model needs dates because inventory can only serve an order after it arrives and becomes available.
Use one row per shipped order or order line
For each historical order line, collect the fields below. Keep the raw data unchanged and create calculated columns in a separate worksheet or model tab.
| Field | Why the backtest needs it | Practical note |
|---|---|---|
| Order date and destination | Replays demand by market and day | Use the date the order became fulfillable, consistently |
| Order ID and SKU | Identifies the unit and prevents duplicate counting | Separate bundle components when they consume separate stock |
| Units, order value, weight, and dimensions | Calculates inventory drawdown and economic impact | Keep actual shipped dimensions when dimensional billing applies |
| Original fulfillment node and cost | Establishes the baseline | Separate product cost from fulfillment, freight, and duties where possible |
| Delivery promise and actual delivery | Tests service-level attainment | Record the agreed promise, not only transit time |
| Return flag, return reason, and disposition | Models reverse-logistics exposure | Distinguish restockable, damaged, and unsellable units |
| Opening stock, receipts, and stock adjustments | Determines whether a scenario could actually fulfill | Include write-offs and quality holds, not only purchase orders |
If the source systems cannot provide every field, do not invent it. Begin with the fields that are available, then list the gap. For example, if true return cost is missing, calculate a low, expected, and high range based on the known return rate and a documented handling-cost assumption.
Clean definitions before calculations
Use one definition for every scenario. "Landed cost" might include international transfer, inbound receiving, storage, domestic fulfillment, carrier charge, packaging, duty, and claims reserve. It should not include an assumed cost in one scenario and omit it in another.
The same discipline applies to service. Define whether a successful order means dispatched by the promise date, delivered by the promise date, or delivered within a stated range. The Council of Supply Chain Management Professionals glossary can help teams align common logistics terms before they compare providers or locations.
Define the Fulfillment What-If Analysis Scenarios
Build scenarios that represent a real decision, not a theoretical map of every warehouse in the world. Freeze the product cost, order history, customer geography, and measurement period. Change only the placement and operating rules being evaluated.
| Scenario | Inventory rule | Best use | Critical assumption to document |
|---|---|---|---|
| China-only | All sellable stock remains in the China node | New, volatile, long-tail, or high-variant SKUs | Route eligibility, order cutoff, cost, and realistic delivery promise |
| Destination-country-only | Eligible stock is transferred before demand occurs | Stable, high-velocity products in one market | Warehouse operator, inbound processing, local fulfillment, storage, and returns |
| Hybrid hero-SKU placement | A limited set of proven SKUs is local; the long tail stays in China | Brands balancing service and inventory flexibility | SKU selection rule, replenishment trigger, and order-routing logic |
| Staged new-SKU launch | New products ship from China until they meet a defined evidence threshold | Testing demand without pre-stocking the catalog | Threshold period, minimum orders, and maximum delay before review |
| Peak-season allocation | Temporary local buffer is added for a defined window | Seasonal launches and promotion periods | Demand uplift, cutoff date, aged-stock exit plan, and capacity |
For a warehouse strategy simulation, write the scenario rules in plain language before you write formulas. For example: "Allocate the top three US SKUs by trailing eight-week unit sales to the destination node, but only if the SKU has at least 35% contribution margin, fewer than 8% returns, a confirmed local returns process, and no more than 45 days of intended cover." The exact thresholds are business choices, not universal benchmarks.
Keep destination-country scope honest
Do not model a destination-country scenario as if a warehouse exists simply because a provider may introduce a partner. A usable node needs a named operator, address, intake process, inventory ownership, system connection, fulfillment charges, daily cutoff, exception process, and returns owner.
Bondjet's documented scope provides a strong China-side control point: receiving, inspection, photo records when required, SKU management, custom packaging, consolidation, international shipping coordination, and China-side dispatch. Learn more about that workflow through Bondjet's fulfillment services and the collectible seller case. The available materials do not establish a standard Bondjet-operated US warehouse, US local one-piece fulfillment, or US returns service. Keep those rows marked "provider capability to confirm" until the exact arrangement is agreed.
Model the Constraints That Create Real Outcomes
The backtest should process orders in date order. At each order, it checks eligible stock, assigns the order to a node, applies the relevant cost and service rule, then reduces inventory. This sequence is what distinguishes a decision model from a static cost comparison.
Inventory and replenishment constraints
At minimum, include opening inventory at each node, planned transfer quantity, transfer departure date, expected receiving date, receiving-to-available delay, replenishment lead time, and a stockout rule. A stockout rule may route an order from a local node back to China, allow a partial shipment, delay the order, or mark it lost. Choose the rule that reflects the actual customer experience.
Also include minimum transfer quantities and storage capacity. A calculation that moves ten units every day may look elegant but be impossible under the supplier's carton configuration, freight booking pattern, or warehouse receiving requirements.
Cost and service constraints
Use transparent components instead of one opaque rate. A simple order-level calculation can be expressed as:
Contribution margin = order revenue - product cost - transfer allocation - receiving - storage allocation - pick/pack - carrier charge - packaging - return reserve - claims reserve
This does not mean every cost belongs in every order. It means every material cost has a home and an owner. Storage can be allocated per unit-day, per cubic meter, or as a scenario-level total. Transfer cost can be allocated by weight, volume, units, or value, as long as the method stays consistent across scenarios.
For imports into the United States, inventory placement does not remove importer obligations. Model customs and entry responsibilities against the proposed route, and consult the U.S. Customs and Border Protection importing guidance for the relevant process. Final duty, admissibility, and entry treatment depend on the product and formal import arrangement.
Returns and condition constraints
Returns can reverse the apparent advantage of local stock, especially for high-value or condition-sensitive goods. Include a return probability by SKU or category, expected timing, restock rate, inspection cost, outbound return label cost, disposition cost, and write-down exposure.
For a collectible, bundle, or fragile item, a returned unit is not automatically sellable stock. The model should distinguish unopened, restockable, damaged, incomplete, and unsellable outcomes. Bondjet's China-side inspection, SKU verification, and protective packaging controls can help reduce avoidable condition and mismatch risk before dispatch, but return rules and claims terms still require route- and provider-specific confirmation.
Compare the Outputs That Matter to an Operator
Do not select the lowest simulated parcel cost by itself. A complete fulfillment scenario backtesting result shows the cost, service, inventory, and risk consequences together.
| Output | Calculation question | What a warning looks like |
|---|---|---|
| Service-level attainment | What share of orders met the stated promise? | Fast average delivery with many stockout-routed orders |
| Contribution margin | What remains after comparable fulfillment and return costs? | Local fulfillment improves service but removes most margin |
| Landed cost per order | What does each completed order cost under the same scope? | A quote omits receiving, storage, or transfer allocation |
| Stockouts and lost orders | How often was the selected node unable to fulfill? | Hero SKU inventory is insufficient during peak weeks |
| Inventory days and aging | How long does stock remain at each node? | Low-velocity products accumulate in the destination node |
| Transfers and replenishment | How many movements are needed to sustain the model? | Operationally impossible transfer frequency or minimums |
| Return cost and write-down exposure | What happens when units come back? | Returns have no owner, address, or disposition path |
| Exception rate | How many orders rely on manual rerouting or overrides? | The plan requires daily intervention to work |
Present results by SKU, market, and scenario, then show a catalog total. The total is useful for finance; the SKU view identifies what action to take. A hybrid model often succeeds because a small number of products carry the service opportunity, while the long tail remains flexible and avoids local aging risk.
[Image suggestion: a scenario dashboard with service level, contribution margin, inventory days, and stockouts by SKU, with clear assumption labels.]
Stress-Test the Warehouse Strategy Simulation
Base-case results are only a starting point. A useful backtest changes a small number of assumptions at a time to expose the breakpoints where the recommended placement is no longer attractive.
Run these five sensitivity cases:
- Demand error: Raise and lower forecast demand while holding the transfer plan constant. This reveals local-stock shortage and aged-stock exposure.
- Peak demand: Concentrate historical demand into a shorter promotion window. Test whether the node has capacity and whether replenishment arrives in time.
- Freight or carrier change: Increase transfer or parcel cost, or change the mix of weights and dimensions. This tests whether the cost advantage is resilient.
- Return-rate change: Increase returns for products with sizing, fit, condition, or buyer-expectation risk. Include a lower restock rate for condition-sensitive goods.
- Delayed replenishment: Add days to international transfer, receiving, or stock-availability time. This reveals how quickly a local buffer becomes a local stockout.
Show the result as a range, such as "hybrid remains preferable only while demand is within this band and the destination node receives replenishment within this lead-time window." Avoid claiming that a model guarantees savings. The point is to identify the conditions under which the decision remains sound.
Use This Reproducible SKU Allocation Scenario Model
Create one input tab, one assumptions tab, and one output tab. The structure below is enough to start in a spreadsheet and can later move into a planning system.
Input tab: order and inventory history
| Column group | Required fields |
|---|---|
| Order | order date, order ID, destination country/region, SKU, units, revenue |
| Product | weight, dimensions, product cost, handling class, return-risk group |
| Original fulfillment | original node, fulfillment cost, dispatch date, delivery date, promise date |
| Returns | return flag, return date, reason, disposition, recovered value |
| Inventory events | opening stock, receipt date, adjustment, transfer departure, transfer receipt |
Assumptions tab: scenario rules
Include a scenario name, SKU eligibility rule, starting inventory by node, transfer quantity, transfer and receiving lead time, capacity, handling and carrier rates, storage method, service promise, stockout fallback, return treatment, and provider capability status. Add a source or owner beside every assumption so a reviewer can trace it.
Output tab: decision table
For each scenario and SKU, show eligible orders, fulfilled orders, promise attainment, stockout-routed orders, contribution margin, landed cost, ending inventory, inventory days, aged units, returns cost, and manual exceptions. Add one final decision field: keep in China, test locally, pre-stock with guardrails, or do not move until capability is confirmed.
This worksheet can support a monthly review. Use recent demand data, but do not react to one promotional spike. Compare the prior decision against actual outcomes, revise the SKU rule when evidence changes, and preserve the assumption history. That turns the spreadsheet into an operating cadence instead of a one-time deck.
FAQ: Fulfillment Scenario Backtesting for Hybrid Inventory
How much order history do I need for an inventory placement backtest?
Use as much clean, comparable history as you have, with enough dates to capture normal demand and a relevant peak period. If a SKU is new, do not manufacture confidence from a short sample. Start it from China, set a clear evidence threshold, and run a staged-launch scenario before allocating destination stock.
Should every fast-selling SKU be placed near demand?
No. A fast-selling SKU still needs sufficient margin, predictable replenishment, compatible dimensions, a defined returns path, and a confirmed local operating model. The best candidate is not merely fast; it can sustain local service without creating unacceptable stockout or aging risk.
Can Bondjet run a US-local scenario in my backtest?
Bondjet can help define the China-side receiving, inspection, SKU, packaging, consolidation, and international-dispatch assumptions that feed the model. A US-local scenario needs the exact destination warehouse operator, inbound process, WMS, local fulfillment terms, and returns arrangement. Treat those elements as unconfirmed until they are documented for the proposed program.
Conclusion: Move Inventory Only After the Model Survives Reality
The best fulfillment strategy is rarely a catalog-wide answer. Fulfillment scenario backtesting gives operators a repeatable way to replay real demand, compare China-only, local-stock, hybrid, and staged options, and see the service, margin, stockout, and aging trade-offs before inventory moves.
Start with clean order lines, explicit definitions, dated inventory events, and a limited set of scenarios. Stress-test demand, freight, returns, and replenishment delays. Then use the output to keep uncertain or long-tail SKUs flexible, test the most promising hero SKUs, and expand only when the operating model has real owners.
For high-value or complex products, contact Bondjet with your SKU list, recent order history, packaging requirements, destination markets, and proposed inventory nodes. Bondjet can help assess the China-side fulfillment controls needed for a credible backtest, including receiving, inspection, SKU management, packaging, and international dispatch, while the destination-country scope is confirmed on its actual terms.