Building a Deal-Based Supply-Side Platform: SSP, Ad Exchange, and Prebid Adapter Development

The Scenario

A U.S.-based programmatic advertising technology company specializing in direct and deal-based ad transactions set out to build what would be the first supply-side platform designed specifically for deal-based programmatic transactions — a premium supply platform (PSP). The organization already operated a mature direct-sales platform and wanted to extend it into the programmatic layer by adding a fully integrated SSP, ad exchange, and Prebid adapter.

The timing reflected a real market shift: increasing privacy restrictions across browsers and operating systems continue to erode identity resolution, limiting buyers' ability to execute audience-driven programmatic strategies. That erosion creates measurable demand for deal-based direct transactions with premium publishers — an environment where this kind of platform fits naturally.

The Approach

The central design goal was to support multiple simultaneous deal types — programmatic guaranteed, preferred deal, and private auction — all running concurrently alongside open-market programmatic and direct deals within a publisher's ad server. That multi-deal structure introduced several technical complexities that a standard SSP architecture would not have needed to solve.

Key-Value Signaling Instead of Price-Only Clearing

By default, Prebid only collects the highest bid from connected SSPs. Because the platform supports multiple simultaneous deals of different types and priority levels, auction clearing cannot be performed on price alone. After bid responses are collected, the platform must pass signals to Google Ad Manager (GAM) reflecting the full scope of demand received from DSPs — not just the top price — for final decisioning.

The solution was to craft a key-value signaling strategy and apply Prebid's sendAllBids method, which sends all relevant bids from the various active deals to GAM rather than limiting the payload to the default of three. This required modifying the character string in the Prebid adapter to include all bids from connected demand sources, and adapting the key-value pairs so they could be correctly matched and interpreted by GAM's line item structure.

Private Auction vs. Programmatic Guarantee

The private auction workflow functions similarly to the programmatic guarantee setup, with one deliberate difference: instead of forwarding all bids, only the highest bid across all deals assigned to that workflow is sent. This distinction reflects the clearing logic appropriate to each deal type and had to be preserved in the adapter logic explicitly.

Line items for the private auction workflow had to be created manually — a labour-intensive step inherent to how Prebid integrates with GAM to pass programmatic demand. Automating this process for publishers on live deals is a planned capability within the direct-sales platform layer.

Implementation Components

The Ad Exchange

The exchange is the core matching layer. Impressions from the Prebid adapter are matched with deals from the direct-sales platform and passed as bid requests to integrated DSPs. The exchange processes bid responses and sends them back via Prebid to GAM.

Unlike a traditional SSP — where a single auction is held and the winning bid is selected — this exchange handles multiple concurrent deal types. Processing bid responses and assembling a coherent batch of signals for GAM required real-time post-RTB processing: taking deal instructions from the direct-sales platform, running real-time processing against DSP integrations, performing decisioning, and then facilitating post-auction handling.

Prebid Adapter

The official Prebid adapter for this platform serves as the header bidding layer connecting publisher pages to the exchange. The primary engineering challenge was adjusting the adapter to send all bids from demand sources to GAM, rather than the Prebid default of three. Integrating correctly with GAM also required learning how to configure and match key-value pairs — including line items — in a way that supports the platform's multi-deal logic.

Programmatic Lab Environment

A self-contained programmatic lab environment was built for business and performance testing. The environment mirrors the production setup but uses mock DSPs and websites. Operators can create deal files, define scenarios, and upload creatives — replicating everything present in a live environment — without affecting production traffic.

Reporting Service

The reporting service was implemented via APIs without a dedicated UI. An endpoint was exposed for data access, with the intent to integrate reporting directly into the direct-sales platform in a future phase.

Technologies Used

The platform was built using the following technologies:

• Backend: Node.js, Nest.js, GO
• Frontend: React.js, Material UI
• Cloud & Infrastructure: Azure (DataBricks and Blob), Kubernetes, Terraform
• Observability: Grafana

Outcomes and Tradeoffs

The result is a functioning SSP specialized for deal-based programmatic transactions — a configuration that did not previously exist as a packaged platform. The multi-deal auction model addresses a genuine gap: standard SSP and header bidding tooling is designed for price-only auction clearing, not for the priority and deal-type logic that premium publisher relationships require.

The key tradeoffs in this approach are worth noting:

• Manual line item creation is required for private auction workflows as long as GAM's integration with Prebid relies on pre-configured line items. Automation at the direct-sales platform layer can eventually absorb this burden for publishers.
• Signaling complexity increases as more deal types are added simultaneously. The sendAllBids approach resolves the immediate problem but requires disciplined key-value management as deal volume grows.
• Reporting without a UI keeps the initial build lean but shifts integration work to the next phase — a reasonable tradeoff when the reporting destination (the direct-sales platform) is still evolving.

For publishers and platforms operating in a privacy-constrained environment where audience-based targeting is becoming less reliable, deal-based programmatic infrastructure of this kind represents a durable strategic direction. The technical pattern described here — exchange matching layer, adapted Prebid signaling, multi-deal type support, and lab-based demand simulation — provides a replicable architecture for organizations building in this space.