Buying Programmatic Ads On The Open Web: Part 1
Understanding the programmatic ads stack, what each piece does, and the sorts of ads you can buy in the wider adtech universe — Part 1 of a 3-part series.
There is a major split in the world of advertising. It’s sometimes called brand marketing versus performance marketing, large agency versus boutique, large versus lean, etc. This manifests in many ways, but for media buyers, I think it comes down to, in many ways, for your advertising, are you largely operating in a sandbox environment or mostly buying media on the “open” web? Facebook is a sandbox, Snapchat is a sandbox, and Google search is a sandbox. They flatten the entire adtech stack from end to end. When you buy an ad on Instagram, they are the publisher. They are the buying platform. They supply the audiences. They figure out the attribution. They have ad creator tools. They handle optimization. They do it all. To be clear, all the same tools are there; it simply becomes “invisible.”
Contrast this with the open web.
When you buy an ad on CNN, you don’t go to the CNN ad platform and follow the same experience as you do with Meta. Typically that process looks more like you hire an agency because they have enough aggregated spend across their clients to access a very large DSP like The Trade Desk (TTD). They then place an order that TTD then uses their bidder to try and fill across many, many exchanges. These exchanges hook up to SSPs like Magnite, which then hook up to ad networks, aggregate publishers, and finally reach an end user.
In between, dozens of vendors do everything from supplying your audiences to helping solve attribution and ensuring that the website you’re buying on is who they say they are and the end user is a real human, not a bot. Each one of these players takes some fee or percent of ad spend. The adtech industrial complex is a giant rats nest of people hooking into one another and bolting on third-party services (see below). When it works, the scale is incredible and difficult to beat. The problem is it's so wildly complex that even many industry insiders don’t entirely understand it.
"The adtech industrial complex is a giant rats nest of people hooking into one another"... No, really, just look at this example of an adtech 'stack' for display.
The library of Alexandria could be filled with books about adtech and how it all works. The complexity is overwhelming, which is part of the reason I think that many performance marketers shy away from it. The sandbox largely just “works,” which is great, but when you only trust the sandbox, you’re opening yourself up to potentially losing a ton of efficiency as the platform is incentivized to make as much money as possible. How do they do that? Raising CPMs.
To be clear, both sandboxes and the open web are useful tools; you just have to use them in different ways. As a reader of The Marketer’s Playbook, I assume 99% of you are performance advertisers and largely operate in sandbox environments. What I’m going to do today is give a high-level overview of the adtech stack so you can effectively leverage it and expand the sorts of things you can test and offer customers. You can find an edge and rise above the zero-sum competition that sandboxes offer. You can have optionality and not be beholden to a handful of very large media companies to grow your brand or client. What you don’t want, but unfortunately, many of you have probably dealt with, is Facebook or Google banning your account with nowhere else to turn.
Not only is buying programmatically an opportunity, but it's also risk mitigation.
You can unlock not just new customers but potentially also find that opportunity to get a greater ROAS.
In this series, we’re going to walk through a few things to help you conceptually understand this ecosystem and then run through some fast tips and tricks to start off on second base in the world of open web advertising and avoid the pitfalls that many media buyers make when they move into this world. We’ll cover the following:
What each level of the adtech stack does.
What sorts of ads you can buy.
Bidding, auctions, and supply path.
DSPs you can use to buy ads.
Common pitfalls I see new media buyers fall into and how to avoid them.
Strategies to help you be successful.
But first, who am I? My name is Jeromy, and I’ve been in advertising for about 12 years at this point in my career. Most of that career was as a performance marketer working in the context of being a freelancer, running an agency, working for agencies, and being a director of digital marketing for a fintech startup. I’ve managed 10's of millions of dollars in ad spend and driven many more million in revenue. Most of that time was spent working in sandbox-like environments, primarily Facebook ads. Over time I expanded into buying open web ads and eventually moved from the agency/services side of the world into the tech side. For the past few years, I’ve been building Decibel, an audio ads buying platform (DSP) that helps make buying audio ads more like buying a Facebook ad or a Google ad. We aggregate over 90% of all digital audio ad platforms and make it easy to help folks drive performance outcomes with open web ads bridging the gaps for agencies and brands.
THE BIG PLAYERS IN ADTECH & WHAT THEY DO
I’m going to gloss past many of the “in the weeds” tools and focus on the big ones for brevity’s sake. I’ll break these up into two large categories, which I’ll call “mainline” and “vendors.” Mainline is what you would actually need to simply run an ad. It’s the minimum, and they’re involved nearly every time you buy an ad. Vendors you can think of largely as features. They integrate into the mainline companies to do various things folks find useful, like targeting ads, attribution, etc. There is also a huge caveat here that many or maybe even most players interact at multiple levels of the stack, so for clarity’s sake, the examples I give probably won’t be perfect, but they’ll help us understand things.
So starting with mainline players, we need to define the following and how they interact with each other. There are: DSPs - These are largely the UI that brands or agencies use to buy their ads. They represent advertisers in the transaction. They use what is called a “bidder” to place orders with exchanges to actually buy the ads that the end user sees. The Trade Desk is one of the biggest here and a good example. Decibel, similarly, is primarily a DSP.
Exchanges - These are where the DSPs hook in to buy the ads. They have what's called a “bidstream” which is a firehose of available ads for DSPs to buy. They get this bidstream by hooking into SSPs (below) and structuring and offering it up to DSPs to buy. These days, the exchange and the SSP are largely the same players, but this isn’t always true. An example of an exchange is PubMatic or OpenX
SSPs - The primary job of an SSP is to help publishers monetize and manage their inventory (users viewing their ads). The entire process of a bidder hooking into a bidstream to buy an ad from an SSP is called RTB or real-time bidding. There are other ways to transact and buy ads that we’ll get into later. An example of an SSP is Xandr.
Ad network - Many publishers aren’t big enough for advertisers to seek out on their own, and they’re looking for more sales, so they join an ad network. These are groups of publishers and are middlemen to make the buying process easier across more properties simultaneously.
Publisher - This is the actual media company. Everything from Hulu to Spotify to CNN are publishers. They are the media companies that the user actually sees/hears the ad on.
This is grossly oversimplified, and I’m butchering a ton of nuance here, but you get the basic idea. Now, on to the 'important' vendors you should know about.
Targeting / Data Providers: These are companies that buy and sell user data from various sources. It’s largely what people are referring to when they say adtech is creepy. Everything from your credit card, many websites, and all sorts of other folks sell user data like this. DSPs get access to this data to filter their bidding in order to make sure they’re reaching the target audience for advertisers. A big example here is Oracle which has a product called BlueKai.
Attribution solutions: These companies basically take and try and map users to devices and match them with conversion pixels you put on your site to attribute a specific ad to a specific outcome. These employ things called “device graphs,” where they basically figure out if your laptop and your cell phone are both yours. They use things like cookies, IP addresses, device IDs, and more to match these up. These are folks like Transunion.
Ad servers: These are largely built into DSPs these days, but some are still standalone. To run your ad, you need a server where the ad actually lives so it can be served. Ad servers are treated as the ultimate source of truth since they get all the data from where an ad is served in a gigantic database called log files. Every impression has a row in this database that comes back with dozens and dozens of columns with information about where that ad ran. What site it ran on, what publisher it ran with, what the user's device was, IP address, and it's down to the second of when it served. For user privacy reasons, most DSPs won’t share the log-level data, so what you see in a dashboard is the aggregated stats of what happened with your campaign. If you had 10 million impressions, you’d have 10 million rows. Adbutler is one example of this.
Fraud verification: These tools are intended to do a few things, but I’ll focus on the main ones. First, was the user that saw your ad an actual user or was it shown to a bot? Second, does the publisher your ad ran on actually exist? Third, did the ad properly deliver, or was it some nonsense like a pop-under? Did it say it served but wasn’t visible to the user (e.g. did it load at the bottom of a page and the user never scrolled down) etc.? A good example is IAS.
Forecasting tools: These look at a bidstream and the other campaign inputs and try to forecast how many impressions you’ll get, the audience size, etc. InMobi does this, among many others.
There are many, many more that do many more things, but you get the idea of how many different vendors and 3rd party tools are involved. Most are at the DSP level, but there’s a whole raft of them for publishers as well.
WHAT KIND OF ADS CAN YOU ACTUALLY BUY?
This is an excellent question, and the answer is that there are quite a lot. Too many to cover in a single blog post. Let’s quickly go through the main categories and what they are with some examples of publishers.
Display - Typical banner ads. These can be on a website or in an app.
Video - Open video. Same idea as banner ads.
CTV - This is stuff like Hulu on Roku-enabled devices. It is still video, but the context is specific enough that it’s worth breaking out on its own. OTT (Over The Top) is a larger umbrella term that includes things like Hulu on mobile phones, so for CTV, the device matters a lot.
Audio - This is usually either streaming (Spotify, Pandora, digital radio, etc.) or downloaded podcasts. There are notable areas it's expanding into, though, like in-video game, audiobooks, live streaming, and more.
DOOH - Digital billboards are increasingly being bought programmatically via DSP platforms these days.
Native - This is things like Taboola, which look like small article-like images and text that appear around news articles.
Legacy media - Stuff like am/fm radio and broadcast/cable television (referred to as linear) is slowly being sold via DSP platforms.
This gets us mostly up to speed on the fundamentals. For part 2 of this series, we’ll dive into exactly how bidding works, the various auction mechanics at play, and what in the world is 'supply path optimization.' In addition, we’ll talk about actionable next steps, like what platforms can be used to do everything we’ve been talking about. For part 3 of the series, we’ll dive into the common pitfalls I see folks make as a platform owner and someone that’s done this for nearly a decade, as well as strategies to help you get the most out of your ad dollars.