Scalable SSP Development for Retail Media Network

About the client

The client is a mid-market omnichannel retailer with a strong eCommerce presence and a network of physical stores across the USA. Alongside their core retail operations, the company runs an in-house retail media network that monetizes digital inventory across product listings, search results, and category pages.

The client’s retail media supply-side business relies on direct relationships with advertising partners. The internal team sells placements via direct IOs, configures campaigns in third-party ad servers, and focuses on guaranteed deals with key brands.

Over time, the network onboarded more advertisers, increased campaign volume, and began working with agencies requesting programmatic access through DSPs. As daily impression volumes grew, retail media became a core revenue stream, and the business moved toward a more scalable ad-delivery model.

Client Challenges

As the retailer’s network expanded, the existing ad stack could no longer support increasing scale and performance load. The main challenges included:

• Slow campaign launches due to manual setup in third-party ad servers.
• Difficulty scaling campaign volume without adding operational overhead.
• Low fill rates during peak traffic events due to latency and limited buyer participation.
• Missed impressions and underdelivery on high-demand placements.
• Limited auction competition due to restricted RTB access and a small pool of demand partners.
• High ad response times under load, resulting in unstable delivery.

Technologies Used

Goals

To scale the retail media business, the client planned to deepen programmatic investment and build a platform focused on performance and growth. The project focused on developing a retail media SSP built as a dedicated supply-side platform with native support for real-time bidding and auction management. It was designed to handle growing traffic, improve auction outcomes, and streamline operations:

• Move from manual IOs to a unified SSP supporting direct and programmatic campaigns.
• Integrate with multiple demand sources without performance issues.
• Maintain low-latency ad responses under peak load.
• Build scalable infrastructure to support growth in impressions.
• Increase programmatic fill rates and competition for ad space.
• Centralize auction logic for complete control and better efficiency.

Work Description

The team delivered the retail media supply-side platform end-to-end, developing every core component from inventory management to real-time auctions. The project demanded careful architectural design and high-quality implementation to address high-load performance and scalability requirements.

Stage 1. Planning and analysis

• Duration: 4 weeks.
• Team: Business analysts, technical architects, project manager.

The team reviewed the retailer’s existing ad operations, including campaign workflows, inventory management, and relationships with core demand patterns. They identified bottlenecks, performance limits, and growth constraints.

Using these insights, the specialists outlined the initial SSP approach, defining high-level architecture, auction logic, and integration priorities.

Stage 2. Architecture and MVP design

• Duration: 6,5 weeks.
• Team: Solution architect, back-end and front-end developers.

Geomotiv’s developers designed a modular retail media SSP architecture to handle high traffic, automated auctions, and scalable programmatic operations. They:

• Defined core modules, including inventory management, auction engine, and DSP gateway.
• Scoped an MVP to test programmatic campaigns and real-time auctions.
• Mapped system workflows to process large impression volumes and maintain low-latency ad responses.

The team reviewed the proposed architecture through internal simulations and prototype tests, validating that it could support peak traffic loads and a growing pool of programmatic partners.

Stage 3. Development and integration

• Duration: 5,5 months.
• Team: Back-end developers, front-end developer, QA engineer.

The development team delivered the prioritized modules for the retail media SSP in 2-week sprints. The platform evolved into a unified system for automated programmatic selling, with built-in inventory management, live auction visibility, and DSP endpoints. Our developers integrated the platform into the client’s retail media network, unifying digital inventory across product listings, search results, and category pages into a single interface.

Geomotiv’s team selected the tech stack to support high traffic and stable performance under load:

• To handle high traffic and maintain low-latency ad responses, the team built the back-end using Go.
• PostgreSQL was used to structure campaign and inventory data in an automated way, even as traffic grew.
• The team leveraged Redis for real-time caching and state management, which was critical for processing large volumes of concurrent bid requests and auctions.
• Kubernetes orchestrated the platform’s containers, so the software could scale and perform under peak loads.

Stage 4. Testing, optimization, and release

• Duration: 8 weeks.
• Team: QA engineers, back-end developers, DevOps.

When the core SSP features were ready, the team tested the platform under load and prepared the system for full production rollout. Our developers ran load and stress tests using patterns that matched real-life conditions. They optimized caching logic and reduced request latency to keep bid processing fast under load.

The team rolled out the retail media supply-side platform alongside the existing ad setup and gradually increased live traffic. They monitored delivery metrics, response times, and auction outcomes in real time. After the platform met production benchmarks, live traffic was routed through the SSP.

Results

Our client replaced fragmented workflows with a unified retail media SSP. The platform now runs high-volume auctions, connects multiple DSPs, and centralizes inventory management. Measurable performance gains followed within the first three months after launch:

• Fill rates on high-demand placements rose by 20%.
• Average server-side ad response times dropped from 120ms to 45ms.
• eCPMs increased by 18%.
• Daily ad impressions reached 5 million without slowing delivery.