Facebook Ads
TikTok Ads
Shopify
When off-the-shelf systems did not fit the factory, the factory built its own WMS
How a Romanian furniture manufacturer united stock, made-to-order production, Shopify stores, warehousing and delivery in a system built around its exact workflow.
Operational data extracted on 22 August 2026 from the production database and codebase. It does not represent revenue, profit or estimated savings.
Executive overview
A Romanian furniture manufacturer operated with both warehouse stock and made-to-order products. Several Shopify stores, physical locations, the workshop, invoicing and delivery had different rules, but needed to work from the same stock and orders.
Instead of forcing the company into a standard ERP, the team built a custom WMS around the actual workflow. In approximately two months of intensive development, the system grew to manage 1,774 products, 1,042 orders and 19,199 Shopify events and validated shipping label issuance for three courier services.
The result verified at this point is not a slogan about "digital transformation." It is an operation with a single source of truth for stock and orders, production tracked by work area, and repetitive steps automated between the store, warehouse, workshop and delivery.
Warehouse goods and made-to-order production in the same orders
The company sells both in-stock products and furniture made to order in its own workshop. A warehouse product must be located, reserved, scanned and shipped. A made-to-order product must enter production, be broken down into components, pass through workshop areas and be tracked through completion.
This mixed model also involves several Shopify stores selling from the same physical stock, sales at physical locations and a team spanning office, warehouse, production and sales roles. Each role needs different information and permissions, but every operation must reach the same source of truth.
The problem was not the absence of software. It was the absence of a system that understood the factory.
One physical inventory, several online storefronts
The same products could be sold through different stores without a single operational source of truth.
Parallel inventory in the e-commerce platform
A forgotten secondary location could retain quantities that made stock figures difficult to reconcile.
Manually prepared shipping labels
Incomplete addresses, missing postcodes and inconsistently written place names required checks and intervention.
Invoicing, the warehouse and orders did not speak the same language
An action completed in one system did not automatically update the rest of the workflow.
Production lacked a single trace
The question "where is the item from order X now?" depended on phone calls, messages and people's memories.
The company did not change its factory to fit the software. It changed the software to fit the factory.
Standard solutions usually treat inventory, production and e-commerce as separate modules. Here, those boundaries disappear within a single order: one item leaves the warehouse, another enters production, both may come from an online store, and the customer needs the correct documents and tracking.
The custom system allowed company rules to become software rules. Stock-tracked products follow the warehouse workflow. Made-to-order products initiate the workshop workflow. Local invoicing and courier integrations were connected directly to operations, without exports and manual steps between applications.
The advantage is not that the software has more features. It is that every feature corresponds to a real decision in the factory.
One system, from order to workshop and delivery
The solution, shown as an order's journey through the system:
| Stage | What the system does | Available evidence |
|---|---|---|
| Multi-store orders | brings orders and edits from stores into one place | 1,042 orders, 1,852 lines |
| Unified inventory | updates the correct quantity in connected stores | 861 stock pushes |
| Pick & Pack | physically verifies products through scanning | 834 scans |
| Multi-courier shipping labels | validates data before issuing the label | 257 shipping labels since 3 August 2026 |
| E-commerce fulfillment | after issuing the shipping label, updates the order and tracking | automated workflow implemented |
| Production and bills of materials | consumes materials according to the ordered product and variant | 31 active bills of materials |
| Workshop areas | tracks each item through the work stages | 7 modeled work areas |
| Returns | identifies the return by shipping label and routes it to the correct workflow | workflow implemented |
| Dispatch | groups delivery runs and operational loads | module implemented |
| Role-based access | shows each user type only the modules they need | 12 active users at extraction |
The numbers describe system usage and coverage at the extraction date. They do not independently quantify financial impact or productivity.
From a hand-drawn diagram to the workshop's rules engine
One of the central features did not start in a specification document. The owner drew the real workshop flow on paper: two work branches that advance separately and meet before the final stage.
The diagram became operational rules. Each item in an order receives its own route sheet. People confirm completion of their stage on a tablet. The shared stage opens only when both branches are ready. The system can then show where each item is and which work area has held it up.
Fast development, with safety nets
The system entered use early and evolved through operational feedback. Major features were prepared separately before moving into production, and the test environment could not issue real documents or orders to external services.
Every important change had a backup and automated checks, and every resolved bug became a permanent test. AI assisted development, but decisions, rules and approvals remained with people.
Challenges that became permanent rules
Stock in the forgotten location
A secondary location in the e-commerce platform retained quantities that distorted the stock picture. After remediation, checking locations became mandatory when connecting each new store.
Two simultaneous operations on the same product
Concurrent updates could produce incorrect deductions. Operations were serialized at database level, and the scenario became a permanent automated test.
An order edited after a shipping label was issued
If an item is removed from the order after the label is printed, the system does not automatically delete the logistics record. It raises an exception and leaves the decision to a person.
What we can demonstrate at this point
| Operational metric | Value verified in the source material |
|---|---|
| Orders in the system | 1,042 |
| Order lines | 1,852 |
| Products managed | 1,774 |
| Shopify events processed | 19,199 |
| Shipping labels issued since 3 August 2026 | 257 |
| Stock pushes to stores | 861 |
| Pick & Pack scans | 834 |
| Production bills of materials | 31 |
| Active users | 12 |
| Automated tests | 355 |
| Lines of code | approximately 21,000 |
What we are not yet publishing as a result
The following claims are absent from this study because no validated before/after measurement exists yet. We will publish them only with the metric definition, period, source, analyzed volume, validation date and publication consent:
- hours saved per day;
- reduced time for shipping label issuance and invoicing;
- percentage or number of errors eliminated;
- "zero overselling";
- higher workshop productivity;
- lower operating costs;
- return on investment;
- quotes attributed to employees.
The system is not a closed project. It is the platform the operation grows on.
The next store can be connected through an established process. The next workshop modules start from the same orders, products and rules. The data gathered on stock, materials and stage durations can later support planning and forecasting.
The value of a custom system lies not only in what it solved at launch, but in the fact that the next problem no longer requires another isolated application.
The right software does not force the factory to work differently. It learns how the factory works.
"This project did not start with a feature list. It started with the question: what actually needs to happen from the moment an order arrives until the product reaches the workshop, warehouse or truck? The answer became a WMS uniting multiple stores, physical stock, production and delivery. On 22 August 2026, the system managed 1,774 products, more than 1,000 orders and almost 20,000 Shopify events, backed by 355 automated tests. This is not a promise that every company needs custom software. It is evidence that when an operation does not fit an off-the-shelf solution properly, a system tailored to the real workflow can close the gap between the factory and its data."
Frequently asked questions
What type of company is featured?
A Romanian furniture manufacturer with its own workshop, Shopify stores and warehousing and delivery operations. The company's name and location are not published.
Why was a custom WMS needed?
The company combines in-stock products with made-to-order products. They follow different workflows but appear in the same orders and share e-commerce, invoicing and delivery operations.
How long did implementation take?
The source material indicates approximately two months of intensive development and gradual adoption in July–August 2026. This is not a guaranteed timeline for other projects.
Which results are confirmed?
Operational volumes in the system on 22 August 2026 are confirmed: products, orders, Shopify events, shipping labels, inventory syncs, scans and automated tests. Time savings, error reductions and financial impact are not yet published as measured results.
Is AI-assisted development the main story of the project?
No. AI accelerated technical execution, but processes, rules, validation and decisions remained people's responsibility. The main story is how well the system fits the factory's operation.
Have a process that does not fit your current software?
Talk directly with the ALLSoft Agency team about your workflow, integrations and what is worth automating.