Jump to content

How Panini Tessuti eliminated errors in order preparation and cut warehouse time by 30%

Blog
Every morning, Panini Tessuti's warehouse printed the orders and sorted them by hand, department by department. Today a Shopware plugin generates a single PDF with critical information highlighted in color. Here we explain why we ruled out OCR and AI and opted for a solution built around the way the warehouse actually works.
Ongoing projects  ·  02/09/2026
In this article

Panini Tessuti sells fabrics by the meter, haberdashery and sewing accessories to two kinds of customers: private individuals and professionals, meaning tailors' workshops, clothing shops with in-house tailoring, and curtain and home-textile shops. Both order on Shopware: private customers from the e-commerce store, professionals from a dedicated B2B site.

Every morning the warehouse, which serves both channels, printed the orders to be fulfilled and sorted them by hand by department, taking two people about 30 to 60 minutes each, twice a day.

Today a plugin for Shopware that we developed specifically for Panini Tessuti generates, in a few minutes, a single PDF already divided by picking department, with visual highlights for the product characteristics that need checking.

The goal: eliminate errors during order preparation and make better use of the warehouse team's time.

How the warehouse sorted orders before

Panini Tessuti sells fabrics by the meter, haberdashery by the piece, children's items with personalized embroidery options, curtains and faux leather. Each department follows a different picking logic (picking means taking products off the shelves to prepare the order): a wrong length or a wrong zipper size is only discovered once the parcel has already left.

Warehouse and operations are the same for the two channels, B2B and B2C: orders from tailors' workshops and from private customers come off the same shelves, pass through the same hands and end up in the same print batch. For a professional customer, though, a wrong length cannot be fixed with a return, because it affects their own customer.

Before the project, the warehouse team did the following:

  • Printed all the orders to be fulfilled

  • Sorted them by hand by department (haberdashery, fabrics, mixed, children's items, etc.)

  • Highlighted lengths, special quantities and customer notes by hand

  • Handed the batches of orders to the relevant departments

Two people, 30 to 60 minutes, twice a day: up to 4 hours a day spent on repetitive work, exposed to human error, that added nothing to the product or to the customer.

The trigger that sped up the solution was technical: when many orders were printed together, the server ran out of memory (out-of-memory) and interrupted the generation of the individual order PDFs. So the problem was not only the warehouse staff's time: at those volumes, the IT system could not cope.

Why we ruled out OCR and AI and read the orders directly from Shopware

There were two options. The first: use OCR (optical character recognition) to analyze the PDFs that Shopware can already produce, extract category, length, quantity and notes, and split the output. The tools evaluated were n8n with OCR and custom-built AI solutions. The second: develop a custom plugin able to read the orders directly from Shopware and assign them to groups using rules defined and written in code.

After several calls with the Panini Tessuti team, we decided to go with the second option, for two reasons:

  • The data was already there: category, length, quantity and notes are structured fields in the Shopware database. Extracting them from a PDF would mean rebuilding with OCR information that already exists in a clean, structured form.

  • Errors must be excluded, not reduced: an OCR that is 98% accurate gets 1 order in 50 wrong, and in a process that runs twice a day, error becomes routine. With fixed rules, the same order always ends up in the same group, without errors.

The plugin reads orders directly from Shopware, applies the rules through an algorithm (so without the probabilistic component of AI) and generates an optimized PDF. It has no external dependencies and no variable costs, and it lives in the admin area of the e-commerce platform that the Panini Tessuti team already knows and uses well.

What the plugin does when the warehouse starts the generation

The plugin we designed and developed adds a section to the Shopware admin area, the one the team uses in its daily work:

  • Background generation: clicking a dedicated button starts the operation on its own, and the button stays disabled until it finishes, to prevent duplicate submissions. A large PDF can take several minutes: generation is asynchronous and runs in the background, so warehouse operators can close the browser or get on with something else in the meantime.

  • Order selection: only paid orders that are ready for fulfillment are included, filtered by order, payment and delivery status.

  • 39 groups with a cover page: orders are organized in a hierarchical tree, and each group opens with a page showing its name, to make life easier for the warehouse staff.

  • Summary email: when generation is complete, a configurable email arrives with the list of processed orders and the link to download the PDF.

  • Status change: orders included in the PDF automatically move to the "in progress" status.

  • Color highlights: the product characteristics to check while preparing the order are highlighted with colors that the Panini Tessuti team can choose and customize on its own from the Shopware admin panel, with no technical intervention.

How we built the custom plugin

Step 1 - It starts with the people who sort the orders every day

We did not ask management how the process worked: we talked to the people who ran it every day. A call with the warehouse operations team and a review of internal documentation led us to a classification tree with 39 groups, built on two main dimensions:

  • Product type: fabrics by the meter, haberdashery, children's items, curtains, faux leather, remnants.

  • Shipping method: standard, express, in-store pickup.

Weight, volume and the presence of products sold by the meter also influence the classification. Each order is assigned to the first group whose criteria it matches: the same order always ends up in the same group, without ambiguity. B2B orders can be told apart from private ones, because they come from the dedicated B2B site.

Step 2 - PDF generation and optimization

The out-of-memory error was caused by the product images included in the Shopware print template: PDFs reached hundreds of megabytes, and printing them in large volumes became technically impossible.

We optimized the compression and resolution of the images in the print template, down to manageable sizes without losing legibility. Background generation removes the risk of timeouts whatever the size of the document.

The images in the print template were the cause of the heavy PDFs in this project: good practice is to check their characteristics if you intend to rely on PDF generation, rather than after the first out-of-memory error.

Step 3 - A cover page for every group

Each group in the PDF opens with a page showing the group's name. Once printed, the Panini Tessuti team can split the batch faster into blocks to hand to the various departments, without wasting time leafing through it.

Step 4 - The eight highlighting rules

This was the part that took the most work together with the Panini Tessuti team. Highlighting rules that look obvious in theory change in practice after the first day of real use, so they were checked and improved following tests on real orders.

Today there are eight highlighting rules:

  • Order notes: notes entered by customers (B2B and B2C) are always highlighted, so they are not overlooked during picking.

  • Non-standard lengths: if an order has several fabric lines with the same length except for one or two exceptions, the exceptions are flagged and highlighted.

  • Haberdashery at risk of error: for zippers, bias tape, aida cloth, cords and ribbons, the variant or size and the quantity are highlighted, if different from 1.

  • Remnants and offcuts: for products classified as remnants, the Panini code and words in the name such as "ritaglio" (remnant) or "scampolo" (offcut) are highlighted.

  • Highlighted words in the title: a configurable list (for example "pannello" (panel), "trapunta" (quilt), "grande" (large), "piccolo" (small)) is highlighted in any product of the order where it appears.

  • Personalized children's items: for products with embroidery, the chosen options (color, type) and the quantity are highlighted, if different from 1.

  • Express shipments: lines with an express shipping method have a dedicated color, so they can be handled with top priority.

  • Orders from the same customer: when the PDF contains several orders from the same customer, each is highlighted and cross-references the others.

Colors, product categories, words to highlight and shipping methods can all be changed from the plugin's admin panel, without touching the code.

What changed in the warehouse

After a few weeks of daily use:

  • Manual sorting eliminated: the 39 groups arrive already separated by cover pages in the PDF generated by Shopware.

  • Picking errors: the number has dropped to zero. The highlights are built around the cases the team says are most at risk: zipper size, anomalous lengths, embroidery options. If new ones are needed, they are added.

  • Order preparation time: the Panini Tessuti team estimates a time saving of more than 30% across all the orders processed in a day.

  • Out-of-memory solved: the problem has not occurred again since the images were optimized and generation moved to the background.

  • Rules managed independently: the Panini Tessuti team changes rules and parameters from the admin panel, without calling the developers for small operational changes.

“Mornings used to start with two of us hunched over sheets of paper, splitting orders by department. Now the PDF is generated without interrupting our work and arrives already divided: everyone takes their block and gets going. Lengths and customer notes stand out right away, and with B2B customers, who can't afford a return, having no errors makes all the difference.”

— Chiara M. - Supply Chain Manager, Panini Tessuti

Checklist to automate order fulfillment on your e-commerce platform

It doesn't matter whether your platform is Shopware, Adobe Commerce (Magento), Sylius or something else: if you are designing a similar system for an e-commerce business with complex operational flows, where B2B and B2C orders are fulfilled from the same warehouse, these are the points to check before you start:

  • Map the current manual process, involving the people who run it every day, not just those who manage it.

  • Check that the data you need is already available through the platform API before considering parsing or OCR solutions.

  • Design PDF generation as asynchronous from the start (HTTP timeouts and out-of-memory errors are predictable and preventable).

  • Check the images in the print template: in this project they were the cause of the heavy PDFs.

  • Define the classification rules as a documented, versioned decision tree.

  • Make categories, words to highlight and colors configurable from the admin, instead of writing them into the code.

  • Add configuration validation with explicit error notifications, not silent failure.

  • Plan iterations and tests with end users on real cases: highlighting rules change after the first day of use in production.

Frequently asked questions

Does this plugin work on all versions of Shopware?

It was developed and tested on Shopware 6. Compatibility with other versions depends on the internal API used: before replicating it on another shop, check the installed version and any API changes between releases.

How long does it take to generate a PDF with a lot of orders?

On days when order volumes peak, generation can take up to 10 minutes. During this time the admin area stays usable while the PDF is being prepared.

Can the highlighting rules be changed without involving the technical team?

Yes. Product categories, words to highlight, colors and express shipping methods can be changed from the plugin's admin panel, and the operations team updates the rules independently.

Can the classification tree be adapted to a catalog other than Panini Tessuti's?

The 39-group tree comes from the departments of Panini Tessuti. For another warehouse it has to be redesigned starting from the way that warehouse works, as we did with the Panini Tessuti team. The dimensions it rests on (product type, shipping method, weight and volume) apply to any shop that needs a different picking flow for each department.

Talk to the team that developed the plugin

Would you like to explore this further with the people who developed the solution for Panini Tessuti? Let's take an hour to talk it through. Contact our team to book a call.