Real-time bidding (RTB) is the method of buying and selling individual ad impressions through an automated auction that runs in a fraction of a second while a web page or app is loading. In digital advertising, it is the core transaction mechanism of programmatic advertising: each time an ad slot becomes available, buyers evaluate that specific impression and bid for it.
RTB replaced much of the older practice of buying fixed quantities of impressions in advance at a negotiated CPM. Instead of paying the same price for every impression in a package, advertisers pay according to the value of each impression to them, based on the user context, placement and their own data.
How a real-time bidding auction works
An RTB auction connects the publisher side and the advertiser side through intermediaries that exchange standardized bid requests and responses. The whole cycle typically has to finish within a timeout of roughly 100 to 300 milliseconds, set by the exchange or the publisher’s setup, so that the ad appears without delaying the page.
- Impression opportunity: a user opens a page or app with an ad slot, and the publisher’s ad server or SSP starts a sale.
- Bid request: the SSP or ad exchange sends a description of the impression to connected DSPs.
- Evaluation: each DSP checks the request against its campaigns, targeting rules, budgets and predictions.
- Bid response: interested DSPs return a price and the creative or a reference to it.
- Auction and win notice: the exchange selects the winning bid and notifies the winner.
- Ad delivery: the creative loads, and impressions and clicks are tracked for billing and reporting.
Most exchanges and SSPs use the OpenRTB specification maintained by the IAB Tech Lab, which defines the structure of bid requests and responses. A shared protocol lets one DSP connect to many supply sources without custom integrations for each.
What a bid request contains
The bid request is the only information a buyer has when deciding how much an impression is worth. Its fields describe the placement, the device, the context and, where permitted, the user.
| Field group | Typical contents | Use in bidding |
|---|---|---|
| Impression | Ad size, format, position, floor price | Matching creatives and setting a minimum bid |
| Site or app | Domain or bundle ID, category, page URL | Brand safety and contextual targeting |
| Device | Device type, operating system, connection type | Format choice and performance prediction |
| User | Pseudonymous IDs, consent signals | Audience targeting and frequency capping |
| Geography | Country, region, approximate location | Geotargeting and local pricing |
| Supply chain | Sellers involved in the transaction | Verifying authorized resale and avoiding fraud |
The amount of user data available varies strongly with privacy regulation and browser policies. In the European Union, consent signals under frameworks such as the IAB Transparency and Consent Framework determine what may be processed, and restrictions on third-party cookies have shifted buyers toward contextual signals and first-party data.
Auction types and pricing in RTB
RTB auctions historically used a second-price rule, where the winner paid just above the second-highest bid. Around 2019, most major exchanges moved to first-price auctions, where the winner pays what it bid, largely because header bidding made second-price outcomes difficult to interpret.
| Auction rule | Price paid by the winner | Bidder behavior |
|---|---|---|
| Second-price | Second-highest bid plus a small increment | Bidding true value is a sensible strategy |
| First-price | The winner’s own bid | Bidders reduce bids through bid shading |
| Soft floor | Varies depending on whether the floor is reached | Hard to predict, now less common |
| Hard floor | At least the publisher’s minimum price | Bids below the floor are rejected |
Under first-price rules, DSPs use bid shading algorithms that estimate the lowest price likely to win and bid near it. This makes the DSP’s price prediction quality a major factor in media efficiency.
RTB within the programmatic ecosystem
RTB is one of several programmatic transaction types. They differ in whether the auction is open to all buyers, whether a price is fixed and whether inventory is guaranteed.
| Transaction type | Auction | Price | Inventory guaranteed |
|---|---|---|---|
| Open auction RTB | Open to many buyers | Determined per impression | No |
| Private marketplace | Invited buyers only | Auction with a floor | No |
| Preferred deal | No auction, first look | Fixed | No |
| Programmatic guaranteed | No auction | Fixed | Yes |
On the supply side, header bidding lets several SSPs and exchanges compete at the same time before the publisher’s ad server makes its decision, which raises competition and publisher revenue. On the demand side, DSPs aggregate many exchanges so that advertisers can buy across them through one interface.
Risks and practical challenges
The speed and scale of RTB create transparency and quality problems that buyers must actively manage. Each intermediary takes a fee, and a meaningful share of the advertiser’s budget may never reach the publisher.
- Ad fraud: bots and spoofed domains generate fake impressions; ads.txt, sellers.json and the supply chain object help verify sellers.
- Brand safety: ads may appear next to unsuitable content unless exclusion lists and verification are used.
- Supply path complexity: the same impression can be offered through several resellers, inflating auction volume.
- Data protection: broadcasting user data in bid requests has been challenged by regulators, especially in Europe.
- Measurement gaps: view-through effects and cross-device behavior are hard to attribute reliably.
Experienced buyers respond with supply path optimization, which means buying through fewer, more direct routes, and with independent verification of viewability and invalid traffic.
A plain-language summary
Real-time bidding is an auction for a single ad view that happens while a page is loading. Many advertisers look at a description of the space and the visitor, and the highest bidder shows its ad.
It is like a very fast auction house where every seat at a stadium is sold separately at the moment someone sits down. Each advertiser pays for the views it actually values, instead of buying a whole block in advance.
An example from practice
An online travel agency wants to reach people who recently searched for flights to Lisbon but did not book. It sets up a retargeting campaign in its DSP with its own visitor data, a maximum CPM and a frequency cap of three ads per user per day.
When one of those visitors opens a news site, the site’s SSP sends a bid request. The DSP recognizes the visitor’s pseudonymous ID, predicts a relatively high chance of booking and submits a bid well above its usual average together with a hotel offer creative. Impressions for unknown users on the same site receive low bids or none.
After four weeks, the agency compares the campaign with a broad display campaign. Retargeting through RTB shows a much lower cost per booking, although the team also runs a holdout test to check how many of those travelers would have booked anyway.