Ad Tech Infrastructure: Building the Engine Behind Retail Media (2026 Guide)

Ad tech infrastructure is the set of systems that power ad delivery, targeting, bidding, and measurement across a retail media network -- and it is the single biggest factor determining whether your retail media program scales or stalls. With US advertisers projected to spend $69.33 billion on retail media in 2026, up from $58.79 billion in 2025 (eMarketer, 2025), the infrastructure beneath every ad impression has never mattered more. This guide maps the five core components of a retail media ad tech stack, explains auction mechanics, compares the vendor landscape, and outlines how API-first, vendor-agnostic infrastructure like the Osmos API Hub lets retailers launch commerce media in as little as two weeks without locking into a single vendor's roadmap.

Retail media has entered what industry analysts call the "infrastructure era," where scale, consistency, and operational efficiency matter more than experimentation (AdButler, 2026). Globally, retail media spend is growing from $184 billion in 2025 toward a projected $312 billion by 2030 (Adtelligent, 2026). Yet many retailers are building on infrastructure that was never designed for retail media -- publisher ad servers, fragmented point solutions, or closed ecosystems that restrict data portability and ad format flexibility. This article is the central reference for Osmos's entire ad tech infrastructure content series: each section introduces a core infrastructure topic and links to deeper-dive articles on build vs buy decisions, ad serving architecture, targeting and attribution, auction mechanics, and more.

What Is Ad Tech Infrastructure for Retail Media?

Ad tech infrastructure is the complete technology stack that enables a retailer to sell, serve, target, price, and measure advertising across its owned digital (and physical) properties. It is distinct from publisher ad tech -- systems built for media owners selling display and video inventory on websites -- because retail media operates on fundamentally different data, ad formats, and monetization models.

A retail media network (RMN) is an advertising business operated by a retailer using its own first-party shopper data and owned media inventory (product listing pages, search results, category pages, checkout flows, in-store displays, and offsite extensions). The infrastructure behind an RMN must handle sponsored product auctions, catalog-based targeting, real-time purchase attribution, and advertiser self-serve workflows -- none of which exist in traditional publisher ad tech stacks.

The five core systems that compose retail media ad tech infrastructure are:

1. Ad server -- the decision engine that selects, delivers, and paces ads in real time
2. Targeting engine -- the data layer that matches ads to shoppers using first-party purchase, search, and behavioral data
3. Auction engine -- the bidding system that determines which ad wins a given placement and at what price
4. Reporting and analytics -- the dashboards and APIs that give advertisers performance visibility and closed-loop attribution
5. API and integration layer -- the connective tissue that lets components talk to each other and to external demand sources

These components interact in a live ad request lifecycle: when a shopper loads a product listing page, the ad server receives a request, the targeting engine identifies relevant audience segments, eligible campaigns enter the auction engine, the winner is selected within milliseconds, the creative is delivered, and the impression (and any subsequent conversion) flows into the reporting layer. For a deeper look at why publisher-native ad tech cannot handle this lifecycle, see our article on why publisher ad tech falls short for retail media.

The Five Core Components of a Retail Media Ad Tech Stack

Every retail media network, regardless of size, needs these five infrastructure components working in concert. The difference between a $10 million and a $100 million ad revenue program often comes down to how well these components are built, integrated, and scaled.

Component 1: Ad Server

An ad server in retail media is the decision engine that evaluates available campaigns, applies targeting rules, runs auction logic, selects the winning ad, and delivers the creative to the shopper -- all within a sub-100-millisecond response window. Unlike publisher ad servers (which primarily traffic display banners and pre-rolls), a retail media ad server must handle product ads with dynamic catalog data, keyword-triggered sponsored listings, and real-time inventory availability checks.

The ad server is the single most latency-sensitive component in the stack. In one documented case study, deploying a purpose-built scalable SSP architecture reduced average ad response time from 120ms to 45ms, improved fill rates by 20% on high-demand placements, and increased eCPM by 18% (Geomotiv, 2025). That platform was built on Go with Redis for real-time bid processing and Kubernetes orchestration on AWS, handling 5 million daily ad impressions via the OpenRTB protocol.

The role of the ad server in a retail media network extends beyond simple ad delivery: it also handles creative trafficking, frequency capping, pacing (spreading budget evenly across a campaign flight), and ad quality enforcement (ensuring brand safety and policy compliance). For a deeper look at smart ad serving capabilities in 2026, see Smart Ad Servers 2025: Retail Media Ad Technology.

Component 2: Targeting Engine

The targeting engine is where retail media's structural advantage becomes concrete. As Topsort's market analysis noted, "Retail media has a structural advantage: it doesn't need third-party cookies" (Topsort, 2025). Retailers sit on first-party purchase data, search query history, browsing behavior, and loyalty program information -- data that is both more accurate and more privacy-compliant than the third-party cookie signals that open web advertising relies on.

A customer data platform (CDP) aggregates and unifies shopper data from across touchpoints (online transactions, in-store purchases, loyalty program interactions, app engagement) into individual profiles that the targeting engine can query at ad request time. A data management platform (DMP) serves a complementary role, ingesting and segmenting broader audience data -- including anonymized third-party data -- for campaign planning and reach estimation.

In a well-built retail media stack, the targeting engine processes shopper segments at query time: when a shopper searches for "wireless headphones," the engine identifies which brand campaigns are targeting that keyword, which audience segments the shopper belongs to (e.g., "electronics repeat buyer," "Prime member," "price-sensitive"), and which campaign has the highest combined bid and relevance score. The speed and accuracy of this matching directly determines both fill rate and advertiser satisfaction.

However, first-party data has limits. According to eMarketer, approximately 90% of US retail media investment flows to just Amazon and Walmart (eMarketer, 2026) -- in part because these platforms have the deepest first-party data pools. Smaller and mid-market retailers must supplement first-party sources with privacy-compliant external data enrichment to fill gaps for irregular buyers and high-value prospects who have not yet logged in or transacted. This is where CDP and DMP integration becomes critical infrastructure, not an optional add-on.

For a deeper exploration of how ad serving, targeting, and attribution work together as a system, see Ad Serving, Targeting & Attribution: The Core of Retail Media Success.

Component 3: Auction Engine

The auction engine determines which ad wins a given placement and at what price. Retail media predominantly uses two auction types:

• First-price auction: the winning bidder pays exactly what they bid. This is the most common model in retail media sponsored product listings, where advertisers set cost-per-click (CPC) bids.
• Second-price auction: the winning bidder pays one cent above the second-highest bid. Target's Roundel moved to this model on February 4, 2025 (Tinuiti, 2025), signaling a broader industry shift toward auction models that reduce bid inflation and increase advertiser trust.

The IAB Tech Lab released its Programmatic Auction Definitions document in January 2026, establishing standardized vocabulary for auction mechanics: 12 roles and a 12-step workflow from inventory setup through bid solicitation, execution, rendering, transaction recording, and monthly reconciliation (PPC Land, 2026). This standardization effort reflects the industry's recognition that auction mechanics need transparency and consistency -- a point we cover in more depth in Transparent Auctions: How Retailers Can Stay Fair and Profitable.

Beyond simple price-based bidding, modern retail media auction engines incorporate relevance scoring (is the ad relevant to the shopper's query and context?), floor pricing (minimum CPC thresholds to protect yield), and automated bidding (AI-driven bid optimization that adjusts CPCs dynamically based on conversion probability). According to Topsort, Poshmark achieved 3.8x ROAS across 1,000+ sellers after deploying automated bidding (Topsort, 2025).

For more on auction automation and scalability, see Automation & Auctions: The Science of Scalable Retail Media.

Component 4: Reporting and Analytics

Reporting infrastructure is where many retail media networks fall short. According to Bain & Company data cited by Treasure Data, only one-third of retail media networks today can report sales at the campaign level (Treasure Data, 2025). This measurement gap is the single largest barrier to scaling advertiser spend: brands will not increase investment in channels where they cannot prove return.

A modern reporting layer must include:

• Real-time impression and click tracking with sub-second latency
• Closed-loop attribution connecting ad impressions to purchases (both online and in-store)
• Advertiser dashboards with campaign-level performance visibility (ROAS, CPA, CTR, conversion rate)
• API-accessible reporting so advertisers can pull data into their own BI tools and agency platforms

According to Kevel, retailers requesting measurement capabilities typically receive only 10-50% of the variables needed for sophisticated analysis (Kevel, 2026). Similarly, only 25% of retail media networks are proficient at incremental measurement (Topsort, 2025). The gap between what advertisers need and what most RMN reporting layers deliver is a core infrastructure deficit, not merely a product feature request.

Component 5: API and Integration Layer

The API and integration layer is the connective tissue of the entire stack. It determines how easily components can communicate with each other, how quickly new features can be deployed, and how flexibly the stack can integrate with external demand sources, attribution partners, and agency platforms.

Melissa Burdick, President at Pacvue, put it directly: "My big phrase that I say to every retailer is API first. It's really what unlocks the ability to be fast, flexible, and be able to scale" (Kevel, 2026). API exposure directly correlates with advertiser spend -- capabilities that are not available programmatically effectively do not exist for large brand platforms running campaigns across dozens of retail media networks.

The Osmos API Hub embodies this principle: it provides open APIs for ad serving, campaigns, events, and reporting -- enabling commerce media setup in as little as two weeks. Because the API Hub connects directly into the Osmosphere retail media operating system (which includes Adscape for ad formats, ControlHub for ad operations, and StratEdge for yield management), retailers get API-first flexibility with a full operating environment behind it. This is a distinct architecture from platforms that offer APIs without a complete operating layer, or platforms that offer a complete UI without API access.

Build vs Buy vs Open API: Choosing Your Infrastructure Path

Every retailer launching or scaling a retail media network faces a foundational infrastructure decision: build the ad tech stack in-house, buy a turnkey platform from a vendor, or adopt an open API-first approach that sits between the two. Each path involves different tradeoffs on cost, time to market, control, and vendor dependency.

As Jaclyn Nix, COO at Kevel (with 18+ years of retail media and ad tech experience, previously leading scaling for 40+ retail media businesses at CitrusAd/Epsilon) wrote in MarTech Series: "In 2026, more retailers will prioritize control over their media stack -- from pricing and formats to data governance and roadmap decisions. This shift is not about rejecting partners, but about redefining relationships" (MarTech Series, 2026).

The Three Paths

Building in-house gives maximum control: you own the data, the roadmap, and the ad formats. The tradeoff is cost and time. A custom build requires ad tech engineers, data scientists, infrastructure ops, and ongoing maintenance investment. Technical debt accumulates rapidly in ad tech systems under production load, and the engineering talent required to maintain a custom ad server is scarce and expensive. For retailers with the resources and scale to justify it, in-house builds can deliver competitive advantage -- but most mid-market and regional retailers cannot staff or fund this path.

Buying a turnkey platform gets you to market fast. Osmos offers a Turnkey Solution that enables go-live in 4 weeks without building from scratch. Closed-ecosystem platforms like Criteo provide immediate access to advertiser demand and campaign management tools. The risk is vendor lock-in: your ad formats, pricing models, and data access are limited to what the vendor supports, and switching costs can be significant.

The open API-first approach represents a third path. Open APIs let you compose your stack from best-of-breed components, connect to multiple demand sources, and maintain control over your data and roadmap without building everything from scratch. Topsort reported that Garmentory integrated retail media infrastructure in 48 hours using API-first tooling (Topsort, 2025), demonstrating the speed advantage of this approach. The Osmos API Hub enables commerce media setup in two weeks, with the full Osmosphere operating system accessible via the same APIs.

For a deeper analysis of the build vs buy decision framework -- including cost modeling, team requirements, and when each path makes sense -- see our dedicated article: Build or Buy? Retail Media Ad Technology Considerations. For guidance on building a scalable stack without third-party lock-in, see How to Build a Scalable Ad Tech Stack Without Third-Party Lock-In. And for a deeper look at the stack ownership thesis, see Owning Your Retail Media Ad Stack.

Auction Mechanics: How Bids Become Placed Ads

Auction mechanics sit at the heart of retail media monetization. Every sponsored product listing, display placement, and offsite extension runs through an auction process that balances advertiser bids, shopper relevance, and retailer yield.

The Ad Request Lifecycle

When a shopper triggers an ad-eligible event (loading a search results page, browsing a category, viewing a product detail page), the following occurs in under 100 milliseconds:

1. Ad request -- the page fires a request to the ad server with context: search query, category, shopper ID, device type, geo
2. Targeting match -- the targeting engine identifies eligible campaigns based on keyword match, audience segments, and advertiser targeting rules
3. Bid collection -- eligible campaigns submit bids (either pre-set CPCs or dynamically computed via automated bidding algorithms)
4. Auction execution -- the auction engine ranks bids by a combination of bid amount, relevance score, and ad quality score; applies floor pricing; determines the winner
5. Ad delivery -- the winning creative is rendered on the page
6. Event tracking -- impressions, clicks, and downstream conversions are recorded for reporting

This lifecycle must complete within a strict latency budget. Industry benchmarks indicate that DSPs typically must submit bids within 50 milliseconds (Xapads, 2025), and total ad response times should stay under 100ms to avoid impacting page load speed and fill rate. The Geomotiv SSP case study demonstrated a concrete improvement: reducing ad response time from 120ms to 45ms directly improved fill rates by 20% (Geomotiv, 2025).

Auction Transparency and Fairness

Auction fairness is becoming a critical infrastructure concern. According to a 2024 survey cited by Marketing Agent blog, 61% of media buyers said they do not fully trust reported auction fairness (Marketing Agent, 2025). According to the same source, an internal audit at Unilever (February to May 2025) found that 19% of bids were suppressed by undisclosed floor prices; after standardizing auction practices, cost efficiency improved 14%.

The IAB Tech Lab's Programmatic Auction Definitions document (January 2026) is a direct response to these trust gaps. By standardizing the 12 roles and 12-step auction workflow, the industry is moving toward a model where advertisers can verify that auctions are conducted fairly and transparently.

For retailers building or selecting ad tech infrastructure, auction transparency is not a nice-to-have -- it is an advertiser retention requirement. Advertisers who cannot trust the auction will shift budget to platforms where they can. For more on fair auction design, see Transparent Auctions: How Retailers Can Stay Fair and Profitable.

API-First Design: Why Vendor-Agnostic Infrastructure Wins

API-first design means that every capability of the ad tech stack -- ad serving, campaign management, event tracking, reporting -- is accessible and controllable via programmatic APIs, not just through a proprietary user interface. This architectural principle determines how flexibly a retailer can integrate with demand partners, build custom workflows, and iterate on ad products without waiting for a vendor's roadmap.

Why API-First Matters for Retail Media

API-first architecture unlocks three structural advantages:

1. Composability -- plug in best-of-breed tools (attribution provider, creative management platform, demand aggregator) without ripping out your core stack
2. Speed of iteration -- launch new ad formats, targeting rules, or pricing models via API calls, not platform release cycles
3. Vendor independence -- avoid single-vendor dependency by maintaining the ability to swap components without rebuilding the entire system

As Jaclyn Nix noted, "The trajectory of adtech and retail media is clear. Retail media will sit at the intersection of marketing, commerce, and technology -- demanding new skills, new structures, and new ways of thinking" (MarTech Series, 2026). API-first infrastructure is the architectural pattern that makes this intersection workable at enterprise scale.

Osmos API Hub: Open APIs for Commerce Media

The Osmos API Hub provides four core APIs:

• Ad Server API -- serve ads with real-time auction decisions, targeting, and pacing
• Campaigns API -- create, manage, and optimize campaigns programmatically
• Events API -- track impressions, clicks, conversions, and custom events
• Reporting API -- pull performance data into any analytics or BI platform

Because the API Hub connects into the full Osmosphere retail media operating system -- Adscape for ad formats (product ads, video ads, in-store, display, offsite, gamified, carousel ads), ControlHub for ad operations (campaign review, wallet management, advertiser onboarding), and StratEdge for yield management and demand generation -- retailers get the composability of an API-first platform with the operational depth of a full-stack OS. This architecture enables commerce media setup in as little as two weeks.

For retailers who want a hybrid approach, the Osmos Custom Solution supports multiple demand sources and flexible integration models -- combining API-first flexibility with the ability to customize workflows based on specific business requirements.

Scalability and Performance Requirements

Retail media infrastructure must handle traffic volumes that scale unpredictably -- sale events, seasonal spikes, new advertiser onboarding -- without degrading performance. Scalability is not a future concern; it is a day-one infrastructure requirement.

Latency Standards

The industry standard for retail media ad response time is sub-100 milliseconds. Widely cited industry research attributes significant revenue impact to latency: every 100ms increase in page load time has a measurable negative effect on conversion rates, which directly impacts both retail sales and ad fill rates. Moloco's documentation cites a p95 latency of approximately 100ms for retail media ad serving decisions including real-time ML personalization (Moloco, 2025).

The Geomotiv SSP case study provides a concrete benchmark: their platform architecture (Go backend, Redis for real-time bid processing, Kubernetes orchestration on AWS with OpenRTB protocol) handles 5 million daily ad impressions with 45ms average response time after optimization (Geomotiv, 2025).

Infrastructure Cost at Scale

Scaling ad tech infrastructure means scaling cloud compute, database I/O, real-time event streaming, and ML inference simultaneously. The cloud advertising market is growing from $5.06 billion in 2025 to a projected $12.64 billion by 2031 at 16.52% CAGR -- driven by retail media, clean rooms, and edge AI bidding (Cloud Advertising Market Report, 2026). For individual retailers, cloud infrastructure costs scale with impression volume: a retailer serving 5 million daily impressions faces very different infrastructure economics than one serving 500 million.

The build vs buy decision is heavily influenced by these cost dynamics. Managed platform providers (including Osmos) amortize infrastructure costs across their client base, making enterprise-grade performance accessible to mid-market retailers who could not afford to build and operate equivalent infrastructure in-house.

Vendor Landscape: Criteo vs Kevel vs Google Ad Manager vs Osmos API Hub

The retail media ad tech vendor landscape includes closed ecosystems, API-first platforms, publisher-native ad servers being repurposed for retail, and full-stack operating systems. Choosing the right vendor (or combination of vendors) depends on your scale, market, technical maturity, and strategic priorities.

Comparison Table

Vendor Analysis

Criteo remains the largest independent retail media technology provider. In September 2025, Google designated Criteo as its first onsite retail media partner, enabling advertisers to manage campaigns across Criteo's retailer inventory via Google Search Ads 360 (Criteo, 2025). As Adtelligent's analysis summarized: "Criteo supports an open ecosystem across multiple retailers with strong analytics and cross-retailer campaign management, but has limited in-store and connected TV integrations" (Adtelligent, 2026). Criteo's strength is scale and brand demand; its weakness is ecosystem lock-in and limited omnichannel coverage.

Kevel is the closest direct competitor to Osmos in the API-first positioning. "Kevel's API-first architecture gives teams full control over ad formats and messaging; rapid deployment in weeks vs. months" (Adtelligent, 2026). Kevel powers retailers like Kleinanzeigen, Dollar General, and Haypp Group. However, Kevel is positioned as an infrastructure layer -- it does not include a built-in demand source, advertiser onboarding workflows, or a unified operating system above the API layer. Retailers who choose Kevel still need to build or source their own campaign management UI, advertiser portals, and demand generation programs.

Google Ad Manager is familiar to publisher ad operations teams and has strong display and video infrastructure. However, it was built for media publishers, not retailers. GAM lacks native support for sponsored product auctions, catalog-based targeting, and closed-loop purchase attribution -- all of which are foundational to retail media. For a detailed side-by-side comparison, see the GAM vs Osmos comparison.

Topsort is an emerging challenger with an AI-native, auction-first architecture. In November 2025, W23 Global's investment connected Topsort to a retailer network including Tesco, Ahold Delhaize, Woolworths Group, and Shoprite Group (PPC Land, 2025). Topsort's strength is its clean auction-first design and rapid deployment speed. Its tradeoff is a narrower enterprise track record compared to established players.

Osmos API Hub occupies a unique position: it provides API-first infrastructure flexibility (comparable to Kevel and Topsort) backed by a full retail media operating system (Osmosphere) that includes ad formats (Adscape), operations (ControlHub), and strategy (StratEdge). This means retailers get API composability without needing to build their own operating layer from scratch. Proven results include 195% ad revenue boost for an India baby and kids retailer, 112% ad revenue increase in 2 months for Konvy Thailand, and scaling in-store retail media across 1,300+ stores for SE Asia's largest multi-brand retail group.

For marketplace-specific infrastructure guidance, see How Marketplaces Can Build Their Own Ad Platforms Without Big Tech.

Data Privacy, Compliance, and Governance

Data privacy is not a feature layer -- it is an infrastructure obligation. Every component of the ad tech stack must be designed with privacy compliance built in, from consent management in the targeting engine to data residency in the reporting layer.

Regulatory Landscape

Retail media infrastructure must comply with a growing patchwork of privacy regulations:

• GDPR (EU/EEA) -- requires explicit consent for data processing, data portability rights, right to erasure, and data protection by design. Enforcement has been aggressive and escalating.
• CCPA/CPRA (California) -- gives consumers opt-out rights for data selling and sharing, with specific requirements for financial incentive programs (relevant to loyalty-linked retail media targeting).
• US state privacy laws -- multiple US states have enacted comprehensive privacy laws, creating a compliance mosaic for retailers operating nationally.
• India DPDP Act (2023) -- India's data protection framework imposes consent requirements and data localization expectations for processing Indian consumer data.
• Australia Privacy Act reform -- tightening rules around targeted advertising and data consent.

For retailers operating across markets (India, SE Asia, Australia, South Africa), data residency requirements add another infrastructure layer: where shopper data is stored and processed may be regulated by jurisdiction. This means the ad tech infrastructure must support configurable data residency -- not a one-size-fits-all architecture.

Build vs Buy Compliance Implications

When you build ad tech in-house, you own the compliance liability entirely: your engineering team must implement consent flows, data deletion pipelines, audit logging, and regulatory update processes. When you buy from a vendor, compliance obligations are shared -- but the retailer remains the data controller under most frameworks. The compliance question is not "who builds it" but "who is accountable when something goes wrong."

For a comprehensive guide to compliance and regulatory considerations, including GDPR and CCPA implementation specifics, see our upcoming article on compliance and regulatory considerations for ad tech infrastructure.

Global Market Context: India, SE Asia, and Australia

Retail media infrastructure is not just a US story. Emerging markets present both the highest growth rates and the most complex infrastructure challenges.

India

India's retail media market is projected to reach over INR 30,000 crore (~$3.6 billion USD) in 2026, representing approximately 15% of total Indian ad revenue (AdTech Today, 2025). The market is driven by major e-commerce players (Amazon India, Flipkart), quick commerce platforms (Blinkit, Zepto, Instamart), and vertical-specific marketplaces. India's quick commerce boom is creating a new category of retail media infrastructure demand: platforms that can serve ads within 10-minute delivery promise windows, requiring ultra-low-latency ad serving and hyper-local targeting.

Osmos has deep India market presence, with proven client results including 200% revenue growth in 2 months for Apollo 24x7, 100% QoQ ad revenue growth for India's leading online pharmacy, 8% revenue growth for BigBasket, and 195% ad revenue boost for a leading baby and kids retailer.

Australia

Australian retail media is the fastest-growing advertising channel in the market. According to IAB Australia, 70% of Australian advertisers and agencies increased their retail media investment over the past 12 months, and 77% of respondents now work with three or more retail media networks -- up from 58% the previous year (IAB Australia, 2025). According to B&T and IAB Australia, retail media is forecast to grow 28.1% in 2025 and 24.4% in 2026, on track to surpass total TV ad revenue for the first time in 2027 (B&T, 2025). The multi-network environment in Australia creates particular infrastructure challenges around cross-network measurement and standardization.

Southeast Asia

SE Asia is experiencing rapid retail media growth, particularly in Indonesia, Thailand, Vietnam, and the Philippines. The region's e-commerce-first retail landscape and mobile-heavy consumer base create unique infrastructure requirements: mobile-first ad formats, regional payment integration, and multi-language support. Osmos has demonstrated scalability in the region, with Konvy Thailand achieving 112% ad revenue increase in 2 months and SE Asia's largest multi-brand retail group scaling in-store retail media to 1,300+ stores.

As Terry Guyton-Bradley, Head of Retail Media at Tata Consultancy Services, observed: "Retail media is no longer experimental. It has stepped into its role as one of the most important, fastest-growing channels in digital advertising" (AdMonsters, 2025). For retailers in emerging markets, the infrastructure decisions made today will determine their competitive position in a channel growing at 20-30%+ annually.

Frequently Asked Questions

What is an ad server in retail media?

An ad server in retail media is the decision engine that evaluates available advertising campaigns, applies targeting rules and auction logic, selects the winning ad, and delivers the creative to a shopper's screen -- all within a sub-100-millisecond response window. Unlike publisher ad servers designed for display banners and video pre-rolls, retail media ad servers must handle product-level catalog data, keyword-triggered sponsored listings, and real-time inventory availability. The ad server also manages pacing, frequency capping, and ad quality enforcement.

What is the difference between a DSP and SSP in retail media?

A demand-side platform (DSP) is the technology used by advertisers (brands) to buy ad inventory across multiple retail media networks from a single interface. DSPs allow brands to manage bids, targeting, and budgets across platforms. A supply-side platform (SSP) is the technology used by the retailer (the publisher/inventory owner) to manage, price, and sell their available ad placements. In practice, most retail media networks operate their own SSP functionality as part of their ad tech stack, while brands use DSPs (or direct self-serve portals) to access that inventory.

How does retail media ad tech differ from publisher ad tech?

Publisher ad tech was built for media companies selling display and video inventory on content websites. Retail media requires fundamentally different infrastructure: sponsored product auction mechanics (not just display bidding), catalog-based targeting (matching ads to specific products and categories), first-party purchase data integration (not third-party cookies), and closed-loop attribution (connecting ad impressions to actual purchases). For a full comparison, see Ad Serving for Retail Media: Why Publisher Ad Tech Falls Short.

What does API-first mean for retail media ad tech?

API-first means every capability of the ad tech platform -- ad serving, campaign management, event tracking, reporting, billing -- is accessible via programmatic APIs, not just through a proprietary UI. This allows retailers to integrate the ad tech stack into their existing commerce platform, connect to multiple demand sources, build custom workflows, and avoid vendor lock-in. API-first architecture is the recommended approach for retailers who want both speed of deployment and long-term flexibility.

How long does it take to build retail media ad tech in-house?

A full in-house retail media ad tech stack (ad server, targeting engine, auction engine, reporting, advertiser portal) typically requires 12-18 months and a team of 15-30+ engineers for an initial build. Ongoing maintenance, feature development, and infrastructure scaling require permanent staffing. By comparison, API-first platforms like the Osmos API Hub enable commerce media setup in 2 weeks, and the Osmos Turnkey Solution enables go-live in 4 weeks -- dramatically reducing both time and engineering investment.

What is vendor lock-in risk in retail media ad tech?

Vendor lock-in occurs when a retailer's ad tech stack is deeply integrated with a single vendor's platform, making it costly and complex to switch. In retail media, lock-in manifests as restricted data portability (you cannot export shopper audience data), limited ad format flexibility (you can only run the ad types the vendor supports), and dependency on the vendor's pricing and product roadmap. Closed ecosystems like Amazon's operate as what Adtelligent described as a "walled garden with restricted cross-platform visibility and rising cost-per-click inflation" (Adtelligent, 2026). Open, API-first platforms mitigate lock-in by ensuring that components can be swapped, data can be exported, and integrations are standards-based.

Is there an open source retail media ad server?

There is no widely adopted, production-ready open source ad server purpose-built for retail media. General-purpose open source ad servers (such as Revive Adserver) exist for publisher display advertising, but they lack the catalog-based targeting, sponsored product auction mechanics, and first-party data integration that retail media requires. Most retailers in 2026 choose between API-first platforms (Osmos, Kevel, Topsort) and closed-ecosystem vendors (Criteo, Amazon) rather than attempting an open source build.

Who owns compliance when you build vs buy retail media ad tech?

Under most privacy frameworks (GDPR, CCPA, India DPDP Act), the retailer remains the data controller regardless of whether ad tech is built in-house or purchased from a vendor. The vendor is typically the data processor. This means the retailer retains ultimate compliance liability -- including consent management, data subject access requests, breach notification, and data deletion. When buying ad tech, compliance obligations should be specified in a Data Processing Agreement (DPA) with the vendor. When building in-house, the retailer must implement all compliance infrastructure directly. For more detail, see our upcoming article on compliance and regulatory considerations for ad tech.

How do retailers manage retail media ad tech after launch?

Post-launch management involves ongoing optimization across multiple dimensions: auction tuning (adjusting floor prices, bid algorithms, and relevance scoring), advertiser onboarding (streamlining signup, campaign review, and payment workflows), ad operations (monitoring fill rates, latency, and ad quality), and reporting calibration (ensuring attribution accuracy and expanding measurement variables). Tools like Osmos ControlHub automate many of these workflows -- including campaign review, wallet management, and advertiser onboarding -- reducing the operational team size required to run a retail media network at scale.

What are the key ad tech stack components for scalability?

The critical scalability components include: (1) a horizontally scalable ad server that can handle traffic spikes during sale events without latency degradation, (2) a real-time event streaming pipeline (typically using technologies like Redis or Kafka) for bid processing and event tracking, (3) cloud-native orchestration (Kubernetes or equivalent) for auto-scaling compute resources, (4) edge-deployed ML models for real-time targeting and bid optimization, and (5) an API layer that can handle thousands of concurrent advertiser integrations without bottlenecking. For more on building a scalable stack, see How to Build a Scalable Ad Tech Stack Without Third-Party Lock-In.