You see a pair of running shoes on one shopping site and for a few days this pair shows up everywhere: your favorite news website, on a recipe blog, the free game on your phone, etc. It’s not an advertiser calling round to book advertising on every website you visit. It’s DSP advertising in the background, software that purchases ads at the right time, in the right place, for the right person.
This is what it means, how the auction that it’s based on works, the true cost to get going, and the things that quietly consume the first month or two of budget. For the marketer, the small business owner or the individual who was just asked to “look into programmatic” after a meeting.
What a demand-side platform actually does
DSPs, or demand-side platforms, are the software that enables advertisers to purchase digital space automatically, rather than on a case-by-case basis, from thousands of websites, apps and streaming platforms. Demand side is the buyer’s side of the deal, you’re the demand, and every website or app that has ad space to sell is the supply, which is linked through what’s known as a supply-side platform.
What no one else explains is this: A DSP is not determining where your ad will best appear. Giving thought to where your ad is likely to be effective, because you’ve got information about the person who is about to see it: what they’ve been browsing, what type of device they’re on, where they are, and what audience segment you’ve set up. You determine the target and the budget. The platform takes care of the buying, which takes place during an auction that ends before the page is fully loaded.
Inside the auction: what happens before your ad appears
It’s all made possible by real-time bidding, and it’s important to understand so that most of the mistakes that advertisers make are due to a lack of understanding that this process is occurring at all.
When a publisher opens a page with ad space for sale, the publisher’s supply-side platform broadcasts a bid request: I have one ad space, and I know this much about this visitor, who will be interested in buying this ad space? That request is sent to an ad exchange, which will then send it to dozens of DSPs simultaneously. Every DSP reviews its active campaigns and determines if this visitor fits in with what any advertiser is trying to do and, if yes, bids. The highest bid wins, the exchange informs the publisher and the ad renders. It all takes place within the time frame of the page loading.
Nobody is manually selecting to display your ad to that particular individual. That’s exactly what the bidding algorithm does for each and every visitor, at every location, every time and only when there’s a match that’s worth the spend of your budget.
What’s changed since last year
Two things have changed enough to make a difference: third-party cookies, and the direction of ad dollars.
For years, cookie deprecation has been coming soon, and browser-level identity restrictions now appear to be the norm rather than the exception, prompting nearly every major DSP to invest the past two years in contextual targeting, first-party data onboarding, and clean room-style data matching, meaning no tracking someone around the Internet. That’s not a nice-to-have question when considering DSP advertising, especially for the first time this year. It’s the baseline.
The other change is the direction of the flow of funds. Connected TV and retail media are moving money away from traditional banners, and the distinction between a retail media network and a complete demand-side platform is vanishing quickly as several major retailers have begun to develop bidding and audience tools more similar to the ones found in DSPs from a few years ago. At the same time, Google’s DV360 remains the biggest overall total spend on programmatic, and the likes of The Trade Desk, which doesn’t own any ad inventory itself, continue to make advances as the top independent choice, while more media buyers are experimenting with options other than just placing the entire budget on one.
All of the above holds true, however. It does mean that the platform that you choose and the data strategy behind it are more important than ever before, and certainly more important than two years ago.
Is a DSP right for your business?
Not all advertisers have to interact with a DSP directly, and that’s where most beginner guides fail to mention. For small budgets and in-house media buyers, a self-serve platform such as The Trade Desk, StackAdapt or AdRoll may suffice, but be prepared for a steep learning curve in month one. These tools are designed for folks who know their audience segments, know how to set frequency caps and know their bid strategy. Logically, most platforms with self-serve require a minimum monthly spend in the low thousands to have sufficient data to optimize against, otherwise, there aren’t enough impressions to learn from. It’s the same budget reality that we have with any small business when they’re talking to us about what digital marketing channels they should be engaging with, and whether or not they should be looking at DSP advertising: DSP advertising is one channel out of many and is not a replacement for a digital marketing strategy.
If that’s not the learning curve or minimum spend you are looking for now, then working through an agency or a managed-service DSP account is probably a better place to start. You give up some direct control, but you gain someone who already knows which targeting mistakes burn budget in week one. To be honest, this channel is good at rewarding those who have the time to actually watch campaign data and is bad at rewarding those who have a set-and-forget budget, and is equally bad at punishing it.
Setting up your first campaign
After you’ve selected a platform, the process of setting up is fairly similar:
- Define one specific audience first, not five. Choose the segment you know best, whether it’s past visitors to your site, a lookalike audience based on your existing audience, or people who are demonstrating intent on a given topic, and create the first campaign on the basis of just one of those segments.
- Have multiple sizes and formats of creatives ready for launch, and not as an afterthought. Ad exchanges can have great inventory, however, one banner size is all that can be used.
- Establish a testing budget that you’re willing to lose. The algorithm is learning the first 1-2 weeks, not you, to judge the ROI of it.
- Before putting any money on the line, install and test tracking either the pixel or conversion API. If not, the platform is optimizing blind, and the only way to know the difference between a good and bad campaign isn’t certain.
- Don’t make major changes to the campaign during that initial learning period. Targeting and/or budget changes can reset the algorithm’s learning and are one of the most frequent causes of initial poor performance.
None of this is nearly as thrilling as the targeting options that DSPs love to tout. It’s the difference between a campaign that improves over time and one that never gets out of its own way.
Where the budget usually gets wasted
These are a few patterns that occur repeatedly in poor accounts.
Going broad because it’s safer to have a larger audience. It isn’t. A specific audience of a few hundred thousand people who are truly relevant and interested in what you have to offer can beat a broad audience of ten million people whose interests don’t really align with yours.
Evaluating within three to four days. Most platforms need one to two weeks of real spend before the bidding algorithm has learned enough to optimize well, killing a campaign on day three is judging a car by how it drives before the engine’s warmed up.
Using the same creative for months. There can be perfect targeting and there can be a silent failure in the creatives that drive poor performance.
Ignoring frequency caps. Don’t show an ad to the same person 40 times a week, it’s increasing irritation, and it’s wasting money that could have gone to someone new.
Treating the platform’s own dashboard as the only source of truth. The conversion rate on the platform and the actual conversion rate for sales or signups you can see in your analytics may differ, and sometimes markedly. It’s generally worth the additional hour to compare that gap to the other paid channels.
Self-serve vs. enterprise: picking a platform
The self-serve level is for marketers and small agencies who prefer easy access to control over a lengthy sales process, and includes The Trade Desk, StackAdapt, AdRoll, and Basis. Most have clear, usage-driven pricing and a trial or demo period, use these before you spend a full budget.
Generally, more inventory and deeper integrations are offered with the enterprise tier: Google’s DV360, Amazon DSP and Xandr, and it often requires a larger minimum spend. An established relationship with the platform Amazon DSP, for example, is best with Amazon users or an agency to manage the account.
There is no correct or incorrect answer to this question. This is dependent on your audience, the information that you have, and whether you have somebody in-house that can handle bidding on a day-to-day basis. A platform that is great for a retail brand for campaigns that are heavy toward Amazon can be wrong for all things for a B2B software company whose customers live on LinkedIn and different industry publications.
How to know whether it’s actually working
Though click-through rate has most of the limelight because it is the easiest number to view, it is not a good measure to understand if the channel is helping the business or not. If a company has a high click-through rate for a page that doesn’t convert, then most likely the targeting attracted people who are just curious, and not buyers.
Improved metrics to monitor CPAs on an amount that you can afford, view-through conversions (VTCs), those who see your ad and don’t click on it, but still convert later, often through display & CTV campaigns, and Frequency vs Reach (FrR), indicating whether campaigns are reaching more or fewer new people. Media mix modeling is worth a look too, especially once a campaign is running across more than one channel and simple last-click tracking stops telling the full story.
If possible, the best way to determine if an ad campaign had an incremental impact is to perform an incrementality or holdout test, which involves intentionally omitting ads for a small but similar segment of the audience. This is natively supported in most enterprise platforms and can be approximated in most self-serve platforms through geographic holdouts.
Where this is headed
Several changes are already beginning to shift the landscape of what a good setup is in the coming year, including bidding leaning even more heavily on machine learning than set up by humans, the use of contextual and first-party data as the standard target, and even retail media networks becoming full demand platforms themselves. Not that any of this requires an overhaul of a total strategy, in fact, the fundamentals listed above are the same in either scenario, but it is a worthwhile review to do targeting and the data setup every few months instead of assuming what worked last year doesn’t need to change.
Getting started
Demand-side platforms are no longer an option just for brands with seven-figure budgets; they still demand the same qualities as before: defined audience, creative that is worth viewing, patience during the learning curve, and someone paying attention to the data. Start with one platform, one audience, and a budget that’s genuinely fine to spend on learning. All the other things in this guide become easier with this foundation.
We have a paid media team who can help you choose a platform and set up that first campaign the right way, without you having to learn from your paid media dollars. You can find information about the digital advertising industry as a whole in the Knowledge Hub, with topics such as ad exchanges, native content, and measuring campaigns.
Zaneek A. is a tech-savvy content strategist and SaaS marketing writer. With a sharp focus on helping SaaS brands grow smarter, Zaneek shares simple guides, smart tools, and proven tips that help businesses reach the right audience faster. When not writing, he’s testing new digital tools or breaking down marketing trends into bite-sized insights.
