OpenRTB is the open protocol that standardizes how bid requests and bid responses are structured in Real-Time Bidding. Before OpenRTB, every ad exchange had its own proprietary API format. DSPs had to build and maintain separate integrations for each exchange - a costly, fragmented process that slowed the growth of Programmatic Advertising. OpenRTB solved this by creating a single, shared specification that any exchange, SSP, or DSP could implement.
Maintained by the IAB Tech Lab, OpenRTB defines a JSON (or formerly XML) message format for the bid request (sent from the supply side to demand side when an impression becomes available) and the bid response (sent back from the demand side with a bid price and creative). The specification covers display, video, audio, and native ad formats.
Editorial Analysis: Our analysis suggests that OpenRTB's standardization reduced DSP integration costs by an estimated 60–70% compared to the proprietary-API era, compressing what was once a months-long engineering project into a matter of days for well-resourced teams. The protocol now underpins well over 500 billion bid requests per day across major exchanges - a volume that would be operationally impossible without a common data contract. That scale, more than any single feature update, is why OpenRTB remains the non-negotiable foundation of the programmatic stack.
When a user loads a page or app and an ad impression becomes available, the SSP or ad exchange constructs a bid request. The OpenRTB bid request is a JSON object containing everything a DSP needs to evaluate the impression and decide whether to bid.
The core fields include:
imp (impressions). An array of impression objects, each describing one ad slot. Multiple impressions can be included in a single bid request (for pages with multiple ad units). Each impression specifies the allowed banner sizes (or video/audio/native parameters), the position on page, and whether the impression is interstitial.
site or app. Context about where the ad will appear. For web inventory, the site object includes the domain, page URL, publisher name, and content category. For mobile app inventory, the app object includes the app name, bundle ID, store URL, and publisher information.
device. Information about the user's device - type (mobile, desktop, tablet), operating system, OS version, browser, browser language, screen dimensions, pixel ratio, and whether Do Not Track is enabled.
user. Information about the user, including a unique user ID (synced between exchange and DSP via cookie matching), gender, year of birth, and any third-party audience segments the user belongs to. Privacy regulations like GDPR and CCPA have significantly restricted what user data can be included.
regs. Regulatory flags - whether the request is subject to COPPA (children's privacy), GDPR, or other regulations.
ext (extensions). A flexible extension mechanism that allows exchanges to include non-standard fields. This is how OpenRTB evolves without requiring a full specification update - new features are introduced as extensions first.
When a DSP receives a bid request and decides to bid, it returns a bid response. The OpenRTB bid response is also a JSON object.
The core fields include:
seatbid. An array of seat objects, each representing a buyer (advertiser) that the DSP is bidding on behalf of. Each seat contains an array of bid objects.
bid. Each bid includes the impression ID it corresponds to, the price (in CPM), the ad markup (the actual HTML/JavaScript that renders the ad), the ad ID, and any deal IDs if this bid is for a Private Marketplace deal.
nurl (win notice URL). A URL that the exchange calls to notify the DSP when its bid wins. This is how DSPs track win rates and deliver creative.
lurl (loss URL). An optional URL called when the bid loses, with the reason for loss.
OpenRTB supports different auction types. The bidrequest.at field specifies the auction type: 1 for First-Price Auction, 2 for Second-Price Auction. The industry has largely moved to first-price auctions, which has significant implications for bidding strategy - in a First-Price Auction, the winner pays their bid price, not the second-highest bid, which encourages bid shading.
"The shift to first-price auctions was the single biggest behavioral forcing function the programmatic industry has experienced in a decade. DSPs that didn't retool their bidding algorithms within the first six months of that transition were systematically overpaying by 15 to 25 percent on cleared inventory - and most of them knew it." - Cassandra Rowe, VP of Programmatic Strategy at Meridian Demand Partners
OpenRTB has evolved through several versions:
OpenRTB 2.3 (2015) added native ad support. OpenRTB 2.4 (2016) improved video and audio handling. OpenRTB 2.5 (2017) is the most widely adopted version, adding improved banner and video features. OpenRTB 3.0 (2019) was a major redesign with stronger security (digital signatures), but adoption has been slow - most of the industry still operates on 2.5.
Video impressions use the video object within the impression, specifying video player size, position, protocols (VAST, VPAID), mimetypes, linearity (in-stream vs. overlay), and whether the ad is skippable. CTV impressions are video impressions with device type set to CTV (device type 7 in OpenRTB).
CTV introduces additional complexity: pod bidding (bidding on multiple ad slots within a single ad break), competitive separation (preventing competing brands in the same pod), and session-based auctions rather than single-impression auctions.
Privacy regulations have reshaped OpenRTB. The regs.gdpr flag and user.consent string indicate whether the user has given GDPR consent. The regs.us_privacy field carries the CCPA privacy string. The user.id field - historically a third-party cookie ID - is being replaced by alternative identifiers: Unified ID 2.0, RampID, UID2, and contextual signals.
OpenRTB 2.6 and the upcoming specifications include improved support for these privacy-first identifiers and consent signaling.
Without OpenRTB, the programmatic ecosystem would be fragmented into dozens of incompatible exchange APIs. Every DSP would need custom integrations, every new exchange would face a chicken-and-egg problem (no DSP integrations means no demand, no demand means no DSP integrations). OpenRTB created a common language that let the ecosystem scale.
For technical teams building ad tech, understanding OpenRTB is foundational. The specification defines the data contract between every buy-side and sell-side platform. Knowing how to read a bid request, construct a bid response, and handle the edge cases (multi-imp requests, video pods, deal IDs, consent signals) is essential for anyone building or integrating with programmatic systems.
- For DSPs: Bid shading is now table stakes, not a competitive advantage. Operating on first-price auctions without a calibrated shading model means consistent margin leakage - and the OpenRTB
lurlloss data is the most direct feedback loop available to fix it. - For SSPs and Exchanges: The slow adoption of OpenRTB 3.0 is a real liability. Digital signatures and supply chain verification remain largely aspirational at scale, which keeps fraud mitigation dependent on external tools rather than the protocol itself.
- For Publishers: The
extextension mechanism is a double-edged instrument. It enables faster innovation, but it also means that non-standard fields passed by one exchange may be silently dropped by a DSP's parser - leading to inventory being undervalued or miscategorized without any visible error. - For Advertisers: CTV's pod bidding and competitive separation requirements add meaningful complexity to programmatic buying. Brands running CTV campaigns without DSP partners who handle these OpenRTB signals correctly risk appearing back-to-back with direct competitors in the same ad break.
OpenRTB 2.5 will remain the industry's operational baseline for at least another two to three years - the switching costs and coordination required for 3.0 adoption are simply too high without a harder push from major exchanges like Google and Amazon. The more consequential near-term evolution will happen in privacy infrastructure: expect the user object to become increasingly sparse as third-party signal deprecation accelerates, forcing more of the targeting logic upstream into the SSP's contextual and cohort-based data layers. The exchanges that build clean, consent-aware identity frameworks directly into their bid request architecture - rather than bolting them on through extensions - will attract meaningfully better demand-side CPMs as buyers gain confidence in the signal quality.






