On this page
Real-Time Bidding (RTB)
Real-Time Bidding (RTB)
Real-time bidding is an automated, per-impression auction mechanism within programmatic advertising where digital ad inventory is bought and sold in the milliseconds it takes a webpage to load. Each time a user's browser requests a page, eligible ad slots are offered to a pool of competing buyers, a winner is determined algorithmically, and the winning ad is served — all before the page finishes rendering. RTB underpins the majority of open-exchange display spending globally and is increasingly central to Retail Media Network monetisation strategies.
How RTB works
RTB is facilitated through three interconnected platform types, described by Amazon Ads (advertising.amazon.com, undated):
- Supply-Side Platform (SSP): programmatic software for publishers to offer their ad impressions to multiple exchanges and demand sources simultaneously, setting floor prices and inventory rules
- Demand-Side Platform (DSP): programmatic software for advertisers to evaluate bid requests and compete for impressions across multiple exchanges from a single interface
- Ad Exchange: the online marketplace where SSPs and DSPs transact; receives bid requests from SSPs, distributes them to connected DSPs, runs the auction, and returns the winning creative
Step-by-step process (Amazon Ads; multiple secondary sources corroborating):
- A user loads a page — the publisher's SSP packages available signals (device, location, contextual metadata, floor price) into a bid request and sends it to the ad exchange
- The exchange distributes the bid request to connected DSPs; each DSP evaluates it against advertiser goals and campaign parameters
- DSPs that find a match submit a bid price; the exchange selects the winner per auction rules and returns the winning creative
- The ad is rendered in the user's browser
The entire sequence takes 50–300 milliseconds (multiple secondary sources; Amazon Ads references ~100ms; Teqblaze references 50–100ms).
Programmatic deal types
RTB is one of four deal types within programmatic advertising, not a synonym for it (Publift, 2026-02-27):
| Deal Type | How it works | Key characteristic |
|---|---|---|
| Open auction (RTB) | Any eligible buyer can bid; highest bid wins | Maximum reach, lowest CPM floor |
| Private Marketplace (PMP) | Invite-only RTB with publisher-set floor prices | Premium inventory, transparency on placement |
| Preferred deals | Publisher gives preferred advertisers first-refusal at fixed price; no obligation to buy | Premium access without auction |
| Programmatic guaranteed | Buyer and publisher negotiate directly; no bidding | Guaranteed volume, highest CPMs |
Auction pricing models
- First-price auction: winner pays their exact bid; now dominant — accounts for 96% of programmatic bids as of 2026 (as-of 2026, digitalapplied.com — secondary, vendor-compiled)
- Second-price auction: winner pays $0.01 above the second-highest bid; effectively extinct on the open exchange after the industry transition that began in 2019 (Publift; multiple sources)
- Bid shading: algorithmic technique to help buyers avoid overpaying in first-price auctions (Publift, 2026-02-27)
Header bidding
Header bidding is an evolution of RTB where publishers offer inventory simultaneously to multiple demand sources before calling their primary ad server (Publift, 2026-02-27). It emerged to address inefficiencies in the older waterfall model, where demand sources were queried sequentially. See Header Bidding for dedicated coverage.
Market scale
- Global programmatic ad spend: estimated ~$755 billion in 2025, projected ~$821 billion in 2026 — approximately 90% of all digital display investment globally (as-of 2026-04, digitalapplied.com)
- 91.5% of all display ads transacted programmatically in 2026 (as-of 2026-04, digitalapplied.com)
- RTB global market size: USD 21.02 billion (2025) → USD 26.32 billion (2026), CAGR ~25.2% (as-of 2026, secondary sources citing unnamed market research firms — treat as illustrative)
- Global RTB bid request volume: ~14.2 trillion per day across major SSPs/exchanges, Q1 2026, +11% YoY (as-of Q1 2026, digitalapplied.com — single secondary source, unverified)
CPM benchmarks
- Open exchange (standard display): CPM $1–4 (as-of 2026)
- Private marketplace (PMP): CPM $5–15 (as-of 2026)
- Retail media (open exchange equivalent): ROAS 3–8x by category (as-of 2026, vendor-sourced via Skai and secondary sources)
Ad quality metrics
- Average display viewability rate (open exchange): 55%; benchmark for "good" = 70%+ (as-of 2026, secondary)
- Invalid traffic (IVT) rate on open exchange: 8–15%; good DSPs reduce to 2–3% (as-of 2026, secondary)
RTB in retail media
Retail Media Networks are integrating RTB infrastructure as the channel matures. Key signals from 2026 sources:
- Retail media reached ~$62 billion in 2026, the fastest-growing ad channel by percentage (as-of 2026, secondary — multiple sources including Skai, Tinuiti)
- Amazon holds approximately 79.7% of US retail media ad spend (as-of 2025, Skai/secondary); Amazon DSP is a major programmatic RTB buyer and seller
- Walmart Connect (#2 at ~8% US share) runs its off-site programmatic via a partnership with The Trade Desk — a classic DSP-powered RTB architecture (as-of 2026, improvado.io/secondary)
- In 2026, classic RTB plumbing (DSP/SSP pathways) is "creeping into the mix" for retail media — previously bespoke, it is converging with standard programmatic practice (Skai, skai.io, 2026)
- 52% of retail media advertisers anticipate redirecting display investment from open-web DSPs toward retail media DSPs (as-of 2026, survey n=166 — source: multiple secondary roundups; primary survey org not identified)
Amazon's retail media market share is variously cited as ~69% ("approximately 69%", aidigital.com) and ~79.7% (improvado.io Top 15 Retail Media Networks, 2026). Both 2026 sources; denominator definitions likely differ (total US retail media spend vs US display/programmatic retail media spend). Neither cites a primary measurement firm.
Privacy impact on RTB
The third-party cookie erosion narrative significantly affects RTB's user-targeting capability:
- Google Chrome reversed full cookie deprecation in April 2025, opting instead for a user-facing "global consent prompt" rather than blocking by default (multiple sources: groas.com, peer39.com, cometly.com). Industry observers expect a large majority of users to decline, creating an environment comparable to Safari/Firefox (which already block third-party cookies).
- France's CNIL imposed a €325 million fine on Google in 2025 for advertising cookie and consent failures — the largest such sanction recorded (as-of 2025, adrenalead.com/secondary citing CNIL)
- As of January 2026, more than 19 US states had comprehensive privacy laws on the books (as-of 2026-01, digitalapplied.com)
- 71% of publishers (Q1 2025) identified first-party data as a key source of positive advertising results; 85% expect first-party data's role in monetisation to increase further in 2026 (as-of Q1 2025, secondary — primary survey source not identified)
- Contextual advertising matches cookie-based behavioral targeting within 5–8% on click-through rates and conversion quality, per 2025 studies (as-of 2025, aerospike.com citing unnamed studies — confidence: low)
- Clean room solutions (privacy-safe data matching without raw data exchange) expanded significantly in 2025: WPP acquired InfoSum (a clean room platform) to enable privacy-safe data sharing between advertisers and publishers (as-of 2025, digitalapplied.com)
- Protected Audience API (PAAPI / FLEDGE): Google's Privacy Sandbox alternative for interest-based targeting in RTB without third-party cookies. The auction runs in an isolated JavaScript worklet; reporting is aggregated and delayed, not real-time. As of 2026, most DSPs have limited or no integration (as-of 2026, peer39.com/secondary)
OpenRTB standard — 2025–2026 updates
IAB Tech Lab maintains the OpenRTB specification, the industry protocol defining how bid requests and responses are structured between SSPs and DSPs. Key 2025–2026 updates from IAB Tech Lab (iabtechlab.com, published 2026-01-14 — primary source: industry standards body):
- OpenRTB 2.6 is the active specification; releases in 2025 focused heavily on CTV (Connected TV) programmatic, the fastest-growing programmatic channel
- Content Taxonomy 3.1 (January 2025): nearly 3× more content categories than deprecated Taxonomy 1.0; specialised CTV genre support; decouples subject matter from format via vectors. Required for CTV inventory to be correctly valued by buyers. IAB Tech Lab: "If you're still using Content Taxonomy 1.0, you're actively putting revenue at risk."
- Extended Content Identifiers (ECIDs) (May 2025 OpenRTB 2.6x update): enables privacy-safe content metadata sharing — publishers pass content IDs (not show titles) to DSPs, protecting content rights while enabling contextual targeting
- Sidecar APIs (2025): supplementary APIs alongside the core spec to handle asynchronous activity and QPS spikes during live events:
- Concurrent Streams API: real-time viewer count signals from publishers to ad systems, helping scale infrastructure for live events ("tsunami of supply" problem)
- Forecasting API: upcoming events data for advance planning (arriving early 2026)
- Deals API (public comment until January 2026): standardised deal metadata protocol; addresses the problem that two-thirds of deals deliver no or minimal revenue due to inconsistent deal configuration across systems; reduces deal setup from hours to minutes
- Live Content Signals (April 2026 public comment): new attributes to signal whether content is a livestream, real-time, or first broadcast; includes substitution macros for clearer pricing signal communication in the bidstream
- Protocol Buffers standardisation (proposed): can reduce data parsing time by up to 50% compared to JSON for bid request/response handling (as-of 2026, IAB Tech Lab)
Key terms
| Term | Meaning |
|---|---|
| SSP | Supply-Side Platform — publisher-side software to sell ad impressions |
| DSP | Demand-Side Platform — advertiser-side software to buy ad impressions |
| Ad exchange | Marketplace where SSPs and DSPs transact via RTB |
| Bid request | Package of signals (device, context, floor price) sent to buyers when an impression becomes available |
| CPM | Cost per mille — price per 1,000 impressions; RTB's standard pricing unit |
| Floor price | Minimum price below which a publisher will not sell an impression |
| Header bidding | Simultaneous multi-buyer auction before primary ad server call; evolution of RTB waterfall model |
| PMP | Private Marketplace — invite-only RTB auction with publisher-set floor prices |
| First-price auction | Winner pays their bid; dominant model since ~2019 transition |
| Second-price auction | Winner pays $0.01 above second bid; effectively extinct on open exchange |
| Bid shading | Algorithm helping buyers avoid overpaying in first-price auctions |
| OpenRTB | IAB Tech Lab's open protocol specification for RTB bid request/response format |
| IVT | Invalid traffic — bot/fraudulent impressions; 8–15% on open exchange |
| PAAPI/FLEDGE | Google's Privacy Sandbox API for interest-based targeting without cookies |
| Clean room | Privacy-safe environment for multi-party data matching without raw data exchange |
What practitioners report
No Reddit MCP data available this run (tool not loaded in session). YouTube found one directly relevant video — "Developments in RTB (Real-Time Bidding) for Retail Media" featuring Andreas Reiffen, CEO of Pentaleap [1] — transcript not retrieved due to rate limiting. Listed as a gap.
Contradictions
First-price auction prevalence: digitalapplied.com (2026-04) states first-price auctions account for "96% of programmatic bids (up from 88% in 2022)." Publift (2026-02-27) describes second-price auctions as still present: "winner only pays $0.01 more than their closest rival." The Publift piece may describe PMPs or legacy exchange behaviour, whereas the 96% figure likely refers to Google Ad Manager's transition (announced 2019, completed ~2021). The 96% figure is secondary/aggregator and lacks a named primary source.
Amazon retail media share: cited as ~69% (aidigital.com, 2026) and ~79.7% (improvado.io, 2026). Likely reflects different base definitions — total US retail media spend vs display-only. Neither cites a primary measurement organisation.
Gaps
- No Reddit practitioner signal (MCP unavailable in session)
- YouTube transcript for "Developments in RTB for Retail Media" (Pentaleap/Andreas Reiffen, Nov 2025) not retrieved — rate limiting
- No EU/UK-specific RTB regulatory enforcement data beyond CNIL France fine
- No independently audited CPM benchmarks (all secondary/aggregator)
- No ecommerce-specific (fashion/apparel) RTB performance data
- No primary DSP market share data (The Trade Desk, DV360, Amazon DSP relative share)
References
- YouTube, November 2025 — www.youtube.com/watch?v=gqmY1lbiBpg