Programmatic advertising is the automated process of buying and selling digital advertising space using software platforms and algorithms, replacing the manual negotiations and insertion orders that once defined the media buying industry. Today, programmatic advertising accounts for 91% of all digital display advertising in the United States, with real-time bidding (RTB) auctions completing in under 100 milliseconds—faster than the blink of an eye. The global programmatic market exceeds $200 billion annually, yet the system that powers it raises profound questions about privacy, transparency, and where advertising dollars actually go.
What Is Programmatic Advertising?
Programmatic advertising refers to the use of automated technology for media buying—the process of purchasing digital advertising space. Rather than a human media buyer negotiating directly with a publisher's sales team, programmatic systems use algorithms, data, and machine learning to decide which ads to show, to whom, and at what price, all in real time.
The shift to programmatic has been dramatic. In 2012, programmatic accounted for less than 20% of display advertising. By 2024, 91% of all US display ad spending was transacted programmatically, according to eMarketer. This growth reflects the efficiency advantages of automation: programmatic systems can evaluate and transact millions of ad impressions per second, optimizing for targeting precision, cost efficiency, and campaign performance at a scale impossible for human buyers.
However, this automation has also introduced significant complexity and opacity. The average programmatic transaction involves 15 or more technology intermediaries, each taking a fee and each adding a potential point of failure—whether through latency, fraud, data leakage, or simple inefficiency.
How Does Real-Time Bidding Work?
Real-time bidding is the auction mechanism at the heart of programmatic advertising. Here is what happens in the approximately 100 milliseconds between a user loading a webpage and seeing an ad:
Step 1: User visits a webpage. The publisher's page begins loading, and the ad server on the page initiates a call to one or more supply-side platforms (SSPs). Step 2: Bid request generated. The SSP creates a bid request containing information about the available ad slot (size, position, page URL) along with whatever user data is available—cookie IDs, device type, location, browsing context, and any audience segments the user belongs to.
Step 3: Bid request broadcast. The bid request is sent simultaneously to multiple demand-side platforms (DSPs) connected to the SSP. Each DSP receives the user data and evaluates whether the impression matches any of its active campaigns' targeting criteria. Step 4: DSPs evaluate and bid. DSPs that identify a match calculate a bid price based on the campaign's budget, the perceived value of the user, competitive dynamics, and optimization algorithms. This evaluation must happen in milliseconds.
Step 5: Auction closes. The SSP collects all bids and selects a winner—typically through a second-price or first-price auction. The winning DSP is notified. Step 6: Ad served. The winning ad's creative is retrieved from the ad server and rendered in the user's browser. The entire process—from page load to ad display—typically completes in under 100 milliseconds. Verification vendors simultaneously assess whether the impression was viewable, brand-safe, and served to a real human.
What Are the Different Types of Programmatic Deals?
Open RTB (Real-Time Bidding) is the most common deal type, where any advertiser connected to the exchange can bid on any available impression. It offers the broadest reach but the least control over where ads appear. Open RTB accounts for approximately 60% of programmatic transactions and is where most ad fraud and brand safety issues occur.
Private Marketplace (PMP) deals are invitation-only auctions where a publisher offers inventory to a select group of advertisers. PMPs provide better transparency—advertisers know exactly where their ads will appear—and typically deliver higher-quality impressions. Premium publishers like The New York Times, ESPN, and The Washington Post increasingly favor PMPs, which command CPMs 2-5x higher than open RTB.
Preferred deals are fixed-price arrangements where an advertiser gets first look at a publisher's inventory at a pre-negotiated price. If the advertiser passes on an impression, it falls back into the PMP or open exchange. Programmatic guaranteed (also called programmatic direct) combines the automation of programmatic with the certainty of a direct deal—fixed price, guaranteed impressions, and predetermined inventory. It's essentially a traditional media buy executed through programmatic pipes.
RTB Open Auction vs. Private Marketplace vs. On-Device Matching
| Feature | Open RTB | Private Marketplace | On-Device Matching |
|---|---|---|---|
| Access | Any advertiser on the exchange can bid | Invitation-only; curated buyer list | Direct advertiser-to-user; no intermediary required |
| Price model | Auction-based (first or second price) | Auction with floor price or fixed CPM | Fixed reward per verified engagement |
| Data shared | Extensive user data broadcast to all bidders | User data shared with invited buyers only | No user data shared; matching happens on-device |
| Transparency | Low: advertisers often can't verify placement | Medium: known publisher, curated environment | High: user sees exactly what's shown and why |
| Floor price | Low or none; race to the bottom | Publisher-set minimum; premium pricing | Not applicable; fixed compensation model |
| Privacy level | Very low: 178 trillion annual data broadcasts | Low-medium: data still shared with multiple parties | Very high: no personal data leaves the device |
| Fraud risk | High: bots, spoofing, ad stacking common | Medium: better controls but not immune | Near zero: cryptographic verification of real engagement |
| Speed | 100-300ms including network latency | 100-200ms with fewer bidders | 10-50ms local computation only |
What Is the Privacy Problem with Programmatic Advertising?
The privacy implications of programmatic advertising are staggering. Every RTB auction requires broadcasting user data to potential bidders—and the scale of this data sharing is almost incomprehensible. According to the Irish Council for Civil Liberties (ICCL), real-time bidding generates approximately 178 trillion data broadcasts per year in the United States and Europe combined. Each broadcast contains user data including device identifiers, location coordinates, browsing context, and audience segment memberships.
The critical problem is that all bidders in an auction receive the bid request data—not just the winner. If 200 DSPs are connected to an exchange and compete for an impression, all 200 receive the user's data, but only one actually serves an ad. The other 199 receive the data for free with no contractual obligation regarding its use. This "bid stream data leakage" means that participating in an RTB auction effectively broadcasts your data to hundreds of companies every time you load a webpage with programmatic ads.
Location data in bid requests is particularly sensitive. Research by the ICCL demonstrated that RTB bid requests frequently include GPS-precise location data—accurate enough to track individuals to their homes, workplaces, places of worship, and medical facilities. A 2024 investigation by 404 Media revealed that US military and intelligence personnel were identifiable through bid stream data originating from apps on their phones. The programmatic advertising system has inadvertently created one of the largest surveillance networks in human history.
What Are Privacy-Preserving Alternatives?
Contextual targeting matches ads to page content rather than user profiles. Instead of knowing who you are, the system knows what you're reading—and serves ads relevant to that content. Modern contextual systems use NLP and computer vision to understand content at a granular level, and research shows they perform within 5-8% of behavioral targeting on key metrics.
On-device matching inverts the programmatic model entirely. Instead of sending user data to ad servers, a catalog of available ads is sent to the user's device, and matching happens locally. Apple, Brave, and Adreva all implement variations of this approach. No user data ever leaves the device, eliminating bid stream data leakage entirely.
Data clean rooms (offered by Google, Amazon, LiveRamp, and others) allow advertisers and publishers to match first-party datasets without either party seeing the other's raw data. While an improvement over open data sharing, clean rooms still require both parties to possess user data in the first place. Seller-defined audiences allow publishers to create audience segments using their own first-party data, giving advertisers targeting capabilities without exposing individual user data to the buy side.
How Adreva Replaces the Programmatic Black Box
Adreva eliminates the need for real-time bidding entirely. There are no bid requests, no data broadcasts, and no intermediaries receiving your personal information. Instead, Adreva's on-device matching system downloads a lightweight ad catalog to the user's browser extension and runs the matching algorithm locally. The user sees relevant ads based on their declared interests, earns rewards for verified engagement, and their data never leaves their device.
This approach solves multiple problems simultaneously. Privacy is protected by architecture, not policy—there's no data to leak because it was never collected. Fraud is eliminated because engagement is verified cryptographically on real devices with real humans. Transparency is built in because users see exactly which ads are shown and why. And the bloated ad tech supply chain is collapsed from 15+ intermediaries to a direct connection between advertiser and user. The result is a system where advertisers pay less, users earn more, and nobody's data gets exploited in the process.
Frequently Asked Questions
What percentage of digital ads are programmatic?
Programmatic advertising accounts for approximately 91% of all US digital display advertising as of 2024, according to eMarketer. This includes all forms of programmatic buying: open RTB, private marketplaces, preferred deals, and programmatic guaranteed. The percentage is slightly lower globally (around 85%) but continues to grow as more markets automate their media buying processes.
How fast is a real-time bidding auction?
A complete RTB auction—from the user loading a page to an ad being displayed—typically completes in under 100 milliseconds. The bid request is generated in 10-20ms, DSPs evaluate and respond in 50-80ms, and the ad creative is served in the remaining time. For comparison, the average human eye blink takes about 300 milliseconds, meaning the entire auction happens roughly three times faster than you can blink.
What is bid stream data?
Bid stream data refers to the user information contained in RTB bid requests. This typically includes device identifiers, IP addresses, location coordinates, browser type, operating system, page URL, and audience segment memberships. All bidders in an auction receive this data, regardless of whether they win the bid. The ICCL estimates that bid stream data generates 178 trillion data broadcasts annually, making RTB one of the largest sources of personal data dissemination in the world.
What is a private marketplace (PMP)?
A private marketplace is an invitation-only programmatic auction where a publisher offers ad inventory to a curated group of advertisers. Unlike open RTB, where any buyer can bid, PMPs restrict participation to pre-approved advertisers. This provides better brand safety for advertisers (they know where their ads appear) and higher CPMs for publishers (premium inventory commands premium prices). PMPs account for a growing share of programmatic spend as both sides seek more control and transparency.
Can programmatic advertising work without cookies?
Yes, and it increasingly must. With Safari and Firefox already blocking third-party cookies and Chrome giving users more control, programmatic advertising is adapting through first-party data strategies, contextual targeting, unified ID solutions (like UID 2.0), and on-device processing. Many advertisers report that cookieless programmatic campaigns perform within 10-15% of cookie-based campaigns when proper alternative targeting is implemented.
What is the difference between programmatic and direct buying?
Direct buying involves a human media buyer negotiating directly with a publisher for a specific ad placement at a fixed price—similar to buying a print ad. Programmatic buying automates this process using software, algorithms, and real-time auctions. Direct buying offers certainty (guaranteed placement, fixed price) but lacks scale and efficiency. Programmatic offers massive scale and optimization but introduces complexity, intermediary fees, and transparency challenges. Programmatic guaranteed is a hybrid that combines direct deal certainty with programmatic automation.