By Luca Passani, @Scientia_CTO, May 2026
“My name is Luca, and I’m a programmatic-holic.”
I’ve spent two-plus years attending Programmatic and Advertising events. Prebid summit, IAB Tech Lab, Programmatic IO, Marketecture, and industry conferences of all kinds. It was a weird feeling initially. I bumped into people who knew what WURFL was, and this made me feel at home, but at the same time, 80% of what was said on stage flew right over my head.
Turns out I was not alone. The advertising industry is very complicated, and its intersection with Programmatic — the technology that lets companies decide which banners and ads get nested in web pages and videos in real time — takes quite a lot of experience to comprehend.
I have decided to distill what I’ve learned over time by osmosis (and a lot of research) into a series of blog posts. Hopefully, it will be useful to others who, like me two years ago, are looking to jumpstart a career in ad tech. By the end of this post, you should have a high-level understanding of the landscape of programmatic advertising.
The languages of Ad Tech — Yieldic and Audient
The self-taught linguist in me came up with this analogy. There are two communities under the Advertising sky, but a mountain separates them. Unsurprisingly, those communities have developed separate languages. There are probably people who are fluent in one and dabbling in the other, but in general, we are looking at two mutually unintelligible idioms in spite of their similarity to English. Those two languages are Yieldic and Audient (no need to look them up; both names are fictional).
Yieldic is the language of publishers: protective, defensive, and mathematically dense. It is spoken by those who own the “land” (the websites) and are constantly worried about it being undervalued or “polluted” by bad ads. The core philosophy is: “My pixels are a finite resource; every millisecond must be monetized.” Yieldic focuses on the here and now.
Audient is an aggressive, hunter-gatherer language. It is spoken by those who have the “gold” (the budget) and are searching the vast digital wilderness for specific types of people. Their core philosophy is “I don’t care where you are. I only care who you are”. Audient is strategic, data-hungry, and focused on “The Result”.
When it comes to doing business, these two communities share a protocol that connects them. OpenRTB — the protocol specification that makes Real Time Bidding possible at scale — isn’t a language per se. It’s a pidgin. A trade tongue that both sides are forced to use to conduct business. It’s a simplified, broken version of both cultures that allows them to transact, even though they go home and grumble in Yieldic or Audient at the end of the day. Understanding this tension explains almost every architectural decision in programmatic, including the ones that seem completely irrational.
Who buys what from whom
In the internet era, advertising has evolved into the automated buying and selling of digital ad space in real-time. Instead of manual negotiations and insertion orders, it uses software and data to decide which ads to show to which users, and all of this, in the milliseconds it takes for a webpage to load. Five to ten companies have worked together to decide which banner or video you should see in this or that area of the webpage. In fact, if you count competing bidders behind the scenes, the number of entities involved in any given impression is much larger.
Figure: The programmatic advertising ecosystem.
If you are approaching ad tech, you might already have heard of the following terms. If not, make sure you are positively familiar with them.
-
- Publishers (e.g., some guy with a popular website or app, all the way up to CNN and The New York Times) sell advertising space in the form of ad units that the ecosystem can bid on. Publishers are on the sell-side (AKA supply-side) of the ad tech ecosystem.
-
- Advertisers/Brands (e.g., a local car dealership promoting this week’s deals, or Coca-Cola running a global brand campaign) are BUYING advertising space (across all media, including programmatic) to promote their products, brands, and services. Advertisers are on the demand side of the ad tech ecosystem.
-
- Programmatic Advertising is the automated buying and selling of digital ads using algorithms to perform full-fledged real-time auctions, known as real-time bidding (RTB). It’s a key part of ad tech that uses machines to optimize ad placement, pricing, and delivery.
-
- Inventory: available ad space or slots on a publisher’s website, app, or other digital platform where ads can be displayed. This can include display, video, mobile, and other ad formats. In programmatic advertising, inventory is typically bought and sold in real-time through ad exchanges, supply-side platforms (SSPs), or demand-side platforms (DSPs).
I’ll take a little break here before I send more names, concepts, and definitions your way. As an incentive to keep on reading, consider the following question:
When a display ad loads on a webpage, who are the 5-10 companies involved in that transaction, and what does each one do?
By the end of this, if you’re ever asked this question, you will be conversant. So, bear with the glossary for a brief moment.
-
- SSPs – supply-side platforms: publishers set up their Prebid auctions (using JavaScript) to offer each ad slot to multiple SSPs in parallel. Each SSP will review the bid request, activate all its secret sauce, and decide whether to run a new auction with DSPs.
-
- DSPs – demand-side platforms: advertisers need to run campaigns. DSPs provide tools for targeting, budgeting, and optimizing campaigns across multiple exchanges and inventory sources. Typically, a DSP will have contracts with many SSPs and accept bids from them. It will apply its secret sauce to each bid to determine how much to bid and whether to bid at all.
-
- Ad Exchanges are marketplace mechanisms that connect publishers to SSPs or directly to DSPs — positioned on the supply side, as you can see in the diagram above. In modern programmatic, the exchange is not a distinct layer sitting between SSP and DSP — most major SSPs have absorbed exchange functionality, and the terms are often used interchangeably. Think of it as the mechanism that enables competitive, real-time price discovery on the supply side.
Now I’m tempted to tell you about DMPs, Identity Graphs, PMPs, and a lot more. But that would kill the momentum completely, and I don’t want that. A practitioner already knows these names. A newcomer would get lost in them. If you are genuinely interested, stay tuned for the future installment (follow me at @Scientia_CTO), and I’ll get there. That’s enough players for now. What matters more than memorizing names is understanding what happened when they all started talking to each other.
Not by design — the Ad Tech Rube Goldberg machine
If you think that the whole programmatic architecture is a bunch of moving parts kept together with scotch tape, that makes two of us. Reality is that programmatic advertising was never designed. It evolved. Each layer was added to solve a problem created by the previous layer, often by a different set of companies with different incentives, and the result is a Rube Goldberg machine of extraordinary complexity that somehow processes trillions of transactions a day.
The original waterfall model (query SSPs sequentially rather than through an auction) was very wasteful for publishers, often leading to lower yield. Header bidding fixed that, but created a ton of traffic (measured in QPS, Queries Per Second). Traffic shaping was invented to fix the QPS explosion. Curation — the selection of high-quality inventory — was invented to add value back to the supply that traffic shaping commoditized. SPO (Supply-Path Optimization) was invented because header bidding created too many redundant paths. Identity graphs were invented because cookies — themselves a hack that was never designed for tracking — face a gradual and contested phase-out, with browsers restricting them and the industry scrambling to build alternatives.
You get the gist. Every solution creates a new problem that requires a new solution. And at every layer, someone is taking a fee. It is generally taken for granted in the industry — but don’t quote me on it — that for every dollar an advertiser spends, somewhere between 40 and 60 cents actually reaches the publisher. The rest disappears into the stack. The industry calls this the “ad tech tax”, and everyone complains about it while continuing to depend on the infrastructure that causes it.
One piece of this architecture deserves special mention. When header bidding emerged as the fix for the waterfall model’s wastefulness, publishers faced a new problem: how to simultaneously ask dozens of SSPs to bid on your inventory without writing custom integration code for each one? The answer was Prebid.js — an open-source JavaScript library, now maintained by the non-profit Prebid.org, that manages the entire header bidding auction from inside the publisher’s webpage. It sends bid requests to multiple SSPs in parallel, collects their responses, and hands the winning bid to the ad server — all before the page finishes loading. Today, it runs on a significant share of the world’s programmatically monetized web pages.
Note: It is also not coincidentally where WurflRTD lives. But that’s a story for later in this series.
The result is a system of such baroque complexity that it processes trillions of transactions a day, and nobody fully understands it. A Rube Goldberg machine of extraordinary scale that somehow keeps running.
Next installment: The Prebid Auction in One Second or So: What actually happens between a page load and an ad appearing.
Want to stay updated on future posts? Follow me on X @Scientia_CTO.
Spotted any inaccuracies? Slide into my DMs @Scientia_CTO