Custom POS System Development Using WPF

Client: Fashion Retailer with 45 Locations in Germany and Austria

Challenge: Legacy POS systems are Cloud-Based, Slow and Lack Resiliency. Staff have difficulty processing returns, matching inventory with sales, and no access to real time sales analytics during the sales transaction.

Solution: WPF Based POS System with Offline Billing Functionality, Integration for Peripherals and AI and IoT Based Control of Cash Drawers.

Despite the hype around web apps, physical retail still relies on reliable, fast, peripheral-connected desktop systems.

WPF (Windows Presentation Foundation) offered:

The Windows Presentation Foundation (WPF) provides three main benefits over alternative options like Electron/React.

Hardware Abstraction

We used WPF to provide seamless access to essential hardware like thermal printers, cash drawers, barcode scanners, and the customer display.

Offline-First Model

Our solution uses an offline-first approach with a local SQLite database that's periodically synced with a central Azure SQL back-end to ensure accuracy and reliability.

Complex UI for both touch and mouse

WPF provides large buttons, quick navigation through categories, and easy compliance with accessibility regulations (Data pertaining to the EU GDPR and WCAG).

The client chose to use WPF over Electron/React because of issues with high memory usage (Electron) and speed (React) when using USB peripherals. The client's use of WPF and .NET 8 provided the native performance they needed.

Billing and Text Support

• Itemized billing including VAT separations (19% and 7% for German customers)
• Textual assistance through intelligent searching based on Pro name, SKU, or barcode, with instant recommendations and fuzzy logic.
• Partial Payment can be made by either cash, credit card, gift card or via a combination of methods, all rounded to the nearest Euro.

Printer & Cash Drawer Integration

• Used Microsoft POS for .NET and raw ESC/POS commands.
• Epson TM-T20III thermal printer: Receipts, return slips, and end-of-day reports.
• IoT-enabled cash drawer – Raspberry Pi Pico W connected to a legacy drawer. The POS sends an MQTT command over local Wi-Fi to pop the drawer. No USB cable mess.
• Fallback: If IoT broker is down, fallback to manual unlock with audit log.

Offline Mode

• Critical for Alpine stores with unstable internet.
• All transactions cached locally in SQLite, encrypted with SQLCipher.
• When online, background sync pushes sales, updates inventory, and fetches price changes.
• Conflict resolution: Last-write-wins with store-preference for pricing.

Sales Analytics (Built-in)

In order to better assess sales and product performance, the following information has been created for use:

A real-time reporting dashboard that provides:

• Sales activity by hour (heat-map style)
• Top ten SKU's by volume
• Average transaction values
• Return rates for product categories

Product/transaction data is stored locally, and aggregated to cloud-based systems for access by Regional Managers.

XLSX files can be exported with a single click and will be formatted according to the DATEV standard for EU accounting requirements.

After on-site workshops, we added:

Following the onsite workshops we completed the following items:

Display for the customer (2nd Screen)

Displays item name and price, loyalty points on VGA/HDMI connection.

Loyalty system:

Uses QR Code scanning & AI-based prediction for future purchases to notify user if low stock of their preferred items.

Multi-language / tax-free transactions

UI in German, English & French. Allows printing of tourist tax free forms.

Role based access to system

Cashier, Supervisor & Manager. Includes option for using biometric fingerprint (IoT Fingerprint Reader).

Automated inventory adjustments

When item is sold its stock total is immediately deducted from inventory. When available stock is below set threshold an automatic purchase order suggestion is created & emailed to appropriate party.

Shift management

Cashier logs in to start their shift, enters starting security box amount, ends their shift and produces difference report if applicable.

Hardware health monitoring

Hardware sensors use IoT technology to alert user should there be an issue (low printer paper, draw jam, etc). Alerts user locally and generates ticket to cloud based system for further action.

In our WPF application, we built a number of lightweight ONNX models that perform local inference on the PC:

• Product Recommendations ie "Customers that purchased this also purchased....." we trained on the company's own transaction history
• Fraud Detection we can flag unusual refund patterns ie, items that were refunded multiple times in under 5 minutes before they have been processed
• Demand Forecasting predicts the peak hour for the next day and will make staffing recommendations for the shifts

All AI is performed locally on the POS CPU (no cloud latency) using .NET ML solution.

• Temperature & humidity sensor in store backroom (ESP8266). If too high for chocolate/cosmetics, POS shows warning before sale.
• People counter (PIR sensor) at entrance – merges with POS transaction count to calculate conversion rate.
• Energy monitoring – Power plug with IoT reports if POS terminal or printer remains on after store closed → remote shutdown command.

Team: 1 WPF lead, 1 backend dev (Azure Functions), 1 IoT engineer.

Timeline: 10 weeks (MVP in 8 weeks).

• Two-week sprints with store manager feedback.
• GitHub + Azure DevOps – CI/CD deploys MSI installers through Octopus.
• Zero-touch update – Background service checks for new version, downloads, and updates overnight.

Testing:

• Hardware simulation (virtual printer, drawer, scanner) for CI.
• Live pilot in 2 stores for 3 weeks.

• Transaction speed – 1.8 seconds per item (vs. 4.2 seconds old system).
• Offline success – Zero lost transactions during 3 regional internet outages.
• Cash variance – Reduced by 73% due to automated shift counting and IoT drawer triggers.
• Inventory accuracy – Improved from 82% to 96% because real-time sync eliminated double-selling.
• Staff training – Under 20 minutes, thanks to familiar WPF UI with keyboard shortcuts.

Cash Drawer Compatibility:

24V RJ11 Cash Drawers are very common in Europe (UK) therefore we made a small relay circuit with an ESP32 rather than replacing all the hardware.

GDPR Compliance for AI:

No customer Data is stored locally and only use anonymised vectors

Printer Character Sets:

German language characters (ä, ö and ü) require encoding in Code Page 850 format.

In conclusion, this case clearly demonstrates that WPF continues to be the best solution for retail POS in Europe, where offline performance, reliability and control of peripherals outweigh the significance of web hype.

In addition, by leveraging AI for intelligent decision-making and IoT technology to provide hardware control, WPF maintains data sovereignty (i.e., retention of sensitive data at local locations) within legacy retail requirements while incorporating modern smart retail technology.