Programmatic advertising has been a mainstay of digital advertising and marketing for much of the internet era. But the advertising approach, which comprises over 90% of US digital display advertising, continues to evolve. Once dominated by desktop computer-based impressions, programmatic today is moving to mobile formats and has branched into connected TV (CTV).
In this guide, we discuss key definitions related to programmatic advertising and explore the ins and outs of the ad buying process, first- and third-party identifiers, ad targeting, and more.
- Want to learn more about programmatic advertising and other marketing trends? Sign up for the eMarketer Daily newsletter.
What is programmatic advertising?
Programmatic advertising refers to any ad that is transacted or fulfilled via automation, which means technology takes on decision-making in the ad serving process and that there is no need for a manual insertion order, according to eMarketer’s Programmatic Advertising Explainer. Programmatic advertising will make up the lion’s share (91.3%) of US digital display advertising in 2024, per a December 2023 eMarketer forecast.
What isn’t programmatic advertising?
Any direct-sold inventory with fixed pricing and a predetermined, limited campaign window is not programmatic advertising. These more manual ad types are more time-consuming on both the buy and sell sides.
How big is the programmatic ad market?
US programmatic digital display ad spend will total $157.35 billion in 2024, according to a December 2023 eMarketer forecast. After several years of rapid growth, programmatic online advertising spend growth has slowed but will still reach 15.9% in 2024.
How does programmatic advertising differ from channel to channel?
Though we forecast 91.3% of US digital display ad spend will be programmatic in 2024, not every digital channel has the same degree of programmatic penetration.
Programmatic has become mobile-dominated
From 2013 to 2023, programmatic ad spend exploded on mobile and expanded into CTV. Over that same time period, US programmatic digital display ad spend increased from $4.99 billion to $135.72 billion, according to an eMarketer December 2023 forecast.
Most growth is coming from video
US programmatic video ad spend passed non-video digital display ad spend for the first time in 2022, according to eMarketer’s April 2023 forecast. Programmatic video ad spend is increasing faster than non-video. The uptick in video ad spend is driven by CTV, mobile, and social media advertising on TikTok and other platforms.
Out-of-home is moving toward programmatic
US programmatic digital out-of-home (DOOH) ad spend will reach $875.1 million in 2024, per an eMarketer December 2023 forecast. That’s a growth of 36.1% YoY. In 2024, programmatic will account for 26.7% of US DOOH ad spend and 9.2% of US OOH ad spend.
Digital audio is also increasingly programmatic
US programmatic digital audio ad spend will total $1.83 billion in 2024, representing 24.6% of the market, according to an eMarketer December 2023 forecast. Spend will increase by 17.2% YoY.
How are programmatic ads bought and sold?
Programmatic ads follow an extensive and fragmented process from when they are bought to when they are served. The ads are purchased either through a programmatic direct transaction where one seller and one buyer interact, or through a bidding process in a marketplace of an open exchange.
On the sell side, publishers, publisher ad servers, and supply-side platforms (SSPs) work together to create, manage, and distribute inventory. On the buy side, advertisers, advertiser ad servers, and demand-side platforms (DSPs) create ads and deliver them. Both sides come together in ad networks and ad exchanges which aggregate supply and demand to facilitate these transactions. These terms are described in more detail within the “Programmatic advertising ecosystem” section further into this guide.
Programmatic advertising can be divided into two main categories—programmatic direct and real-time bidding (RTB). According to eMarketers’s Programmatic Advertising Explainer:
Programmatic direct involves one seller and one buyer. Some level of negotiation is required, and pricing is predetermined. When a certain amount of inventory is guaranteed, the transaction is called programmatic guaranteed. When inventory is not guaranteed, the transaction is called a preferred deal.
RTB involves auctions that take place in … real time, as a webpage or app is loading on a user’s device. Within RTB, there are two main transaction types. Private marketplaces (PMPs) are typically owned by a single publisher or a small group of publishers and are open to only a select number of invited buyers. The open exchange, as the name suggests, is open to all buyers and sellers in the digital ecosystem.
Header auction vs. waterfall auction
Per eMarketer’s Programmatic Explainer: Header auction bidding is a practice that runs ad auctions across multiple demand sources on a user’s device before calling an ad server.
In the waterfall auction structure, one demand source is contacted at a time, and the first bid that surpasses the price floor set by the publisher is accepted.
By all indications, header bidding will continue gaining ground in years to come, as it helps publishers bring in more revenues versus waterfalling.
How are programmatic ads served?
Here are the steps to the programmatic ad process:
- An ad loads on a user’s device.
- Bid requests are sent to header bidding partners.
- The winning header auction bid competes with other demand services on the ad server.
- The publisher’s ad server determines the winning bid based on price and priority.
- The winning bid is selected.
- The advertiser’s ad server delivers the ad.
- The winning ad is served on a user’s device.
Managed service vs. self-service
Two main kinds of service models can be found among the major ad tech vendors, as explained in eMarketer’s Programmatic Advertising Explainer:
Managed service. The vendor provides hands-on assistance, from campaign setup and maintenance to mid- and post-campaign reporting. Managed service often comes with a minimum spend requirement and is more expensive on a per-impression basis. If you’re an advertiser or publisher who has high standards for transparency, this may not be the option for you.
Self-service. The vendor sets up a dashboard, provides training, and leaves the client to it. Self-service is a great option for advertisers and publishers who can’t meet a minimum spend requirement or want more control. Self-service allows users to customize reporting, and those customers often have free rein over optimization levers. But getting the most out of a self-serve campaign requires a lot more time and effort.
How does programmatic ad targeting work?
Programmatic ads are targeted either contextually or behaviorally, leveraging first- and third-party data to direct ads toward people who have high potential to interact.
Contextual vs. behavioral targeting
Contextual targeting relies on content surrounding an ad and serves something that makes sense within that context. For example, an ad for a running shoe might run alongside an article about running. Contextual targeting does not leverage user data.
Behavioral targeting relies on data related to the consumer, such as browsing activity or purchase history. For example, an ad for a running shoe may appear for a user who has previously looked at listings for running shoes, regardless of what content they are looking at now. Behavioral targeting does leverage user data, though it is currently pivoting away from third-party cookies.
What are third-party identifiers, and how do they compare with first- and zero-party identifiers?
Third-party identifiers are means of tracking behavior and purchase history across websites and apps over time. Third-party identifiers like cookies and mobile advertising IDs (MAIDs) follow users around websites, collecting data for behavior targeting. These identifiers came at the cost of consumer privacy, and for that reason, regulators have been cracking down. Here are some examples of third-party identifiers, as defined by eMarketer’s Programmatic Advertising Explainer:
- Third-party cookies are pieces of code that live on browsers and are used to track consumer behavior and purchase history over time. Unlike first-party cookies, which are implemented exclusively by the website that places them, third-party cookies come from another website. Google is currently in the process of phasing out Chrome’s third-party cookies.
- MAIDs are identifiers that live on mobile devices and are used to track consumer behavior and purchase history over time, specifically for use in mobile in-app advertising.
- Advertising ID is Google’s MAID for Android devices.
- Identifier for Advertisers is Apple’s MAID for iOS devices.
AppTrackingTransparency (ATT) is a privacy framework Apple released in April 2021 that requires apps to request permission to track a user across other apps on iOS 14.5+ devices. ATT posed yet another hit to third-party identifiers, specifically Meta’s Pixel.
First-party data is collected by an advertiser about its own users or consumers, like retail media data or data that exists within a walled garden (more on that below).
First-party is filling the gaps left by waning third-party identifiers. It can be shared through data clean rooms, which allow multiple parties to share anonymized first-party data to produce audience and campaign insights.
Zero-party data is like first-party data, but it’s given willingly and enthusiastically by a consumer, especially in exchange for a reward. For example, a consumer might give up their email address for a coupon.
What is a walled garden?
A walled garden is an ad platform that has all of the following features:
- A closed ad ecosystem. All advertising is owned by the platform, as are processes including ad buying, serving, tracking, and reporting.
- Exclusive access to its own ad space. This is when purchased ad inventory is direct from the platform, not via third parties or resellers.
- First-party data on users. The data users share with the platform stay within the garden walls due to commercial interest and consumer privacy.
Some examples include TikTok, Twitter, Snap, Amazon, YouTube, and Meta’s Facebook and Instagram.
US walled garden programmatic digital display advertising will cross the $100 billion mark in 2024, according to an April 2023 eMarketer forecast. That said, in 2023, these ecosystems lost share of the programmatic digital display ad market for the first time since eMarketer began tracking the segment in 2017. Walled garden spending share is being dragged down as the duopoly of Meta and Google becomes less dominant.
Programmatic advertising KPIs
- Programmatic ad spend growth (Subscribers only): Percent by which money spent on advertising has increased over time
- Programmatic average cost per thousand (CPM): The average CPM impressions of a given campaign
- Programmatic percent spend: The percent of ads that are transacted on programmatically
- Programmatic spend distribution: The distribution of where programmatic ads appear
Programmatic advertising ecosystem: Platforms, exchanges, and partners
The following glossary also comes from eMarketer’s Programmatic Advertising Explainer report:
Ad exchanges: Think of these as marketplaces. They bring together people who want to sell ad space (sell side) and those who want to buy it (buy side). This all happens in real time, making programmatic ad buying and selling faster and easier.
Examples: AppLovin, TripleLift, InMobi, PubMatic, Google’s AdX
Ad networks: These are like wholesalers in advertising. They buy ad space from publishers and then sell it to advertisers at a higher price.
Examples: Google Ads, Criteo
Ad verification firms: These are the quality control experts of advertising. They make sure that ads appear where they’re supposed to, reach the right target audience, and stay away from inappropriate content. They also keep an eye out for fake traffic and measure how well ads are doing.
Examples: DoubleVerify, Integral Ad Science, Zefr
Advertiser ad servers: Also known as third-party ad servers, these take ad creative and deliver it directly to a web browser. They also track how ads are doing.
Examples: Equativ (formerly known as Smart AdServer), Google Campaign Manager 360
Customer data platforms: These gather customer information and organize it neatly. This helps companies understand their customers better over time.
Examples: BlueConic, Treasure Data, Segment, Tealium, mParticle
Data clean rooms: Imagine a highly secure meeting room where companies can safely share their data with each other. This helps them understand their audiences better and make smarter campaign decisions.
Examples: Google Ads Data Hub, Amazon Marketing Cloud, LiveRamp’s Safe Haven, InfoSum’s Secure Data Clean Room, Snowflake Data Cloud
Data management platforms (DMPs): These are data collectors. They gather all sorts of data—whether it’s directly from consumers, bought from someone else, or freely available—and crunch it to help advertisers make quick, smart decisions about where to show their ads.
Examples: Lotame, Audience Manager DIL extension in Adobe Experience Cloud, Oracle BluKai DMP
DSPs: These are like one-stop shops for advertisers. They can buy ad space from a bunch of different places (like ad exchanges and networks) all in one go. These platforms also help advertisers manage their ads and figure out new places to advertise (like on streaming TV or digital billboards).
Examples: The Trade Desk, MediaMath, Yahoo DSP, Adobe Advertising Cloud, Amazon DSP, Google Display & Video 360
Measurement and attribution partners: These track ads, gather data, and figure out the impact of an ad, including how many people it reached and what actions they took after seeing it.
Examples: Nielsen, Comscore, Kantar, and AppsFlyer
Publisher ad servers: Also known as first-party ad servers, these manage ad inventory, facilitate direct deals, host RTB auctions, and track campaign results.
Examples: Kevel, OpenX, Equativ, Google Ad Manager
SSPs: These are similar to DSPs but for publishers. Publishers use these platforms to connect with lots of different ad buyers (like ad exchanges and networks) all in one place. They help publishers manage their ad space and make the most money.
Examples: Magnite, PubMatic, OpenX, Amazon Publisher Services, Google Ad Manager
Click fraud potential in programmatic advertising
Because programmatic ads are served automatically, there is potential for ads to end up on “made-for-advertising” websites where engaged users are unlikely to see them, or in contextually inappropriate places, like pornographic or politically controversial websites.
An Adalytics report from 2023 alleged that 80% of Google video ads were misplaced on contextually undesirable websites. While the Media Rating Council claimed that Adalytics overstated its claim, Google took action to avoid serving ads on low-quality websites by turning off ad serving to these sites.