Why fixing product sync boosted customer onboarding by 40%

In the world of retail tech, groundbreaking engineering often starts with solving everyday inefficiencies—not building the next distributed system or AI model.

At Casa Retail AI, a small but pivotal automation project is still celebrated by leadership years later. The work centered on a seemingly simple goal: keeping product data consistent across multiple retail platforms. The result? A 40% reduction in onboarding time and a scalable foundation for future growth.

The backbone of product data management: Commerce Connect

Casa Retail AI operates an internal platform called Commerce Connect (CC), which serves as the central Product Information Management (PIM) system and the single source of truth for product data.

Built on a customized version of the open-source e-commerce platform Spree Commerce, Commerce Connect supports multi-vendor and multi-tenant environments. Its core responsibility is to ingest product information from various retail ecosystems and distribute it to downstream applications within the Casa ecosystem.

Once a product enters CC, its data is synchronized across multiple systems—ensuring consistency across customer-facing tools, analytics dashboards, and external integrations.

Where product data drives decisions—and why consistency matters

Product data powers far more than just inventory tracking. It fuels critical business functions across the company:

• Customer interactions: Lead management systems use product details to enrich conversations, while ticketing platforms link customer issues to specific items.

• Digital experiences: Receipts, catalogs, and B2B portals all rely on accurate product names, images, and specifications.

• Analytics & reporting: Retailers depend on product data to track performance—identifying top-selling categories, high-traffic products, and recurring customer complaints.

• Third-party integrations: Some tenants require product synchronization with external platforms like B2B portals, digital flipbooks, or custom APIs.

With so many systems depending on the same data, maintaining a single, accurate source is essential. Discrepancies can lead to delayed onboarding, reporting errors, and frustrated customers.

The original onboarding bottleneck: manual CSV uploads

When a new tenant joined Casa Retail AI, the onboarding process began with a tedious workflow:

1. The Customer Success (CS) team requested product exports from the tenant.
2. Tenants provided CSV files containing their product catalog.
3. The CS team manually mapped CSV columns to Commerce Connect fields.
4. The CSV was uploaded into CC, triggering a batch processing job.
5. Background workers created products in CC.
6. Additional sync jobs distributed updates to PostgreSQL, ClickHouse, B2B apps, and other integrations.

Visually, the flow looked like this:

POS Export → CSV File → Manual Mapping → Commerce Connect → Batch Processing → PG / ClickHouse / B2B Sync

At first, this process appeared manageable. But as Casa scaled, the limitations became clear.

Three major pain points emerged

1\. Onboarding delays tied to manual data entry

A tenant couldn’t fully use the platform until their entire product catalog was loaded into CC. Delays weren’t caused by engineering—they stemmed from waiting on tenants to provide, format, and upload CSVs.

Sometimes files arrived late. Sometimes they were incomplete. Sometimes mappings were incorrect. Each issue added days to the onboarding timeline.

2\. Constant updates required repetitive manual work

Retailers frequently update their product catalogs—adding new items, modifying existing ones, or deactivating old products. Every change required:

• A new CSV export from the tenant.
• Re-uploading the entire catalog into CC.
• Reprocessing thousands of products.

Keeping data in sync through spreadsheets became unsustainable at scale.

3\. Manual configurations per POS provider

Different POS systems generated different CSV formats. Each time a new tenant onboarded or a new POS provider was added, the CS team had to reconfigure mappings—an error-prone and time-consuming task.

The breakthrough insight: why not pull data directly?

The product data already existed inside the tenants’ POS systems. So why were humans manually exporting spreadsheets and uploading them into CC?

The answer was simple: because APIs weren’t being used.

The team realized that if they could pull product data directly from POS APIs—instead of waiting for CSVs—the entire process could be automated, faster, and more reliable.

Building the solution: direct API integrations

The first POS provider to offer product APIs was SAP. The engineering team integrated Commerce Connect with SAP’s API, enabling direct product retrieval and automatic synchronization.

The results were immediate:

• No more waiting for CSV files.
• No manual mapping or uploads.
• No delays caused by tenant responses.

After SAP’s success, the team replicated the approach for other POS providers, expanding automation across the ecosystem.

Designing for flexibility and scalability

Rather than rewriting the synchronization pipeline for every new POS provider, the team introduced a Factory Pattern to abstract provider-specific logic.

Factory.build("sap_iplanet")

This design allowed dynamic instantiation of provider implementations. Each new provider only needed to define:

• How to authenticate with its API.
• How to fetch products.

This modular approach made it easy to onboard new tenants and support additional POS systems without increasing operational overhead.

A legacy of efficiency and scalability

Years later, this automation project remains a cornerstone of Casa Retail AI’s infrastructure. It didn’t involve writing complex algorithms or scaling distributed systems—just solving a real operational pain point with clean engineering.

By shifting from manual CSV uploads to direct API integrations, the team reduced onboarding time by 40%, eliminated human error, and created a scalable foundation for future growth. The project is a reminder that sometimes the most impactful tech solutions aren’t the most glamorous—but they deliver measurable value where it matters most.