Routing was tribal knowledge
Territories, priorities, and stock-aware failover lived inside a macro and the person who understood it.
FIELD NOTES / SUPPLY CHAIN & FULFILLMENT
Fulfillment.
Finally governed.
RouteStock turns marketplace order files into stock-aware warehouse decisions, carton-level shipment records, and warehouse-ready outputs—without leaving critical logic inside one person’s spreadsheet.
SUPPLY CHAIN CASE STUDY
RouteStock
Multi-warehouse routing.
Fulfillment execution.
RouteStock / Fulfillment operationsPRODUCT EXPERIENCE
ONE OPERATIONAL VIEWEvery run has an outcome.
Processed orders, configuration exceptions, stock gaps, and warehouse load stay visible.
Follow the workflow ↗Channel ordersValidate & normalizeRoute & commitWarehouse files
DOMAIN
Multi-warehouse fulfillment
WHAT WE BUILT
Routing + allocation engine
OUR ROLE
Product & full-stack engineering
01 / THE OPERATIONAL CHALLENGE
An order arrives.
Four decisions follow.
Which warehouse should ship it? Is stock truly available? How should it be packed? What file does the warehouse need?
In a multi-site network, every order line combines channel-specific data, ZIP territories, live availability, SKU conventions, packaging rules, and downstream formats. A spreadsheet macro can hide that complexity, but it cannot govern it.
RouteStock turns those decisions into explicit configuration, an explainable engine, and a daily workflow any authorized operator can run.
Available was not the same as on hand
Concurrent order runs could compete for the same units without a reservation made at commitment.
Physical execution was still manual
Carton splits and per-warehouse file formats created repetitive work after allocation was complete.
02 / INSIDE THE PRODUCT
From order file.
To warehouse-ready work.
Explore the supplied screens behind daily processing, inventory control, packaging configuration, and operational visibility.
The daily operation, not just the data.
Recent batches, exceptions, warehouse load, and stock attention sit together so operations can see what moved and what needs intervention.
THE FULFILLMENT CYCLE
Import→Validate→Normalize→Route→Reserve→Cartonize→Export
03 / AUTOMATION WITH OPERATIONAL GUARDRAILS
Automate the decision.
Explain the exception.
The engine commits work only when configuration and stock support it. Every uncommitted line carries a specific reason, and a processed batch stays reversible only until warehouse files leave the system.
Look under the hood ↘01
Normalize the inbound order
Amazon TSV · Walmart XLSX/CSV · Manual template
Commit and cartonize
Choose the strongest stocked site, reserve inventory, and emit physical boxes.
Classify the exception
Name the missing packaging rule or unavailable stock—per order line.
02
Render for each warehouse
Template-shaped XLSX or CSV, one file per destination site.
03
Reverse only while safe
Undo releases commitments before export; after export, the interlock protects physical execution.
04 / THE TECHNOLOGY & THE REASON
Configurable rules.
Transactional guarantees.
The stack separates operator experience, API authorization, business logic, and governed data—so a new channel or output format can be added without changing the fulfillment engine.
OPERATOR EXPERIENCE
React + Redux Toolkit
Permission-aware workflows make processing state, configuration, exceptions, and site scope visible.
Why this combination?
Complex operational state stays predictable while long-running batches can report live progress.
SERVICE BOUNDARY
FastAPI + Pydantic
Typed request and response contracts sit in front of 13 documented routers and a separate service layer.
Why separate the service layer?
Routing and allocation rules remain testable business logic rather than behavior buried in endpoints.
GOVERNED DATA
PostgreSQL + SQLAlchemy
Network rules, inventory, orders, shipments, users, and audits share one transactional record.
Why a relational core?
Row locks prevent double-commitment, constraints protect history, and transactions keep a batch coherent.
FILES IN / FILES OUT
openpyxl + tolerant parsers
Channel exports and configuration workbooks normalize into one model; stored templates shape warehouse output.
Why template-driven output?
New downstream formats become configuration rather than changes to the allocation engine.
ACCESS & SCOPE
JWT + bcrypt + server-side RBAC
Role baselines, user overrides, and warehouse assignments are enforced on each relevant request.
Why enforce scope in the API?
Hiding an interface control is not a security boundary; refusing an unauthorized request is.
CHANGE CONTROL
Alembic + audit snapshots
Versioned migrations evolve the schema while before/after records support attributable reversal.
Why retain prior state?
A bad configuration edit or import can be restored without reconstructing values from memory.
Architecture visuals reproduce the supplied project documentation. Demo screen values illustrate product behavior and are not presented as measured client outcomes.
Explore the complete documented stack
The documented implementation covers the web client, API, business services, relational data, file handling, identity, and deployment.
- Client
- React 18 · Vite 6 · Redux Toolkit · guarded routes · batch progress polling
- API & contracts
- FastAPI · Uvicorn · Pydantic 2 · JSON + multipart file exchange
- Business services
- ingestion · ZIP routing · stock commitment · bulk conversion · carton splitting · output rendering · reversal
- Data
- PostgreSQL 15 · SQLAlchemy 2.0 · Alembic · row locks · decimal quantities · soft deletion
- Files & identity
- openpyxl · XLSX/CSV/TSV parsers · stored templates · JWT · bcrypt · RBAC
- Platform
- Docker Compose · composed database/API/web services · health and diagnostics endpoints
Technology names identify the documented implementation, not partnerships or endorsements.
A CONVERSATION ABOUT YOUR FULFILLMENT WORKFLOW
Which routing decision
still lives in a spreadsheet?
Bring one order-routing, inventory-allocation, or warehouse-output workflow. In 20 minutes, we’ll map the decision rules, operational risks, and a practical automation boundary.
20 minutes · Your operation, constraints, and next step
20 minutes · Your operation, constraints, and next step