For online marketers, demand-side management platforms — DSPs for short — are supposed to revolutionize how we buy online advertising. DSP advertising seems to be the the Holy Grail: the One Console to Rule Them All. The ticket to that tropical island where we marketers could sip caipirinhas and fiddle with a single online interface for an hour a day while making buckets of money.
Sadly, DSPs aren’t there yet. I don’t have a fruity tropical drink in my hand, nor does anybody else who lives and dies by performance marketing. (By performance marketing, I mean the kind of online direct marketing where you put a dollar in and you need to prove you got at least a dollar and change out.)
So what’s the problem?
In theory, there should be no problem at all, because DSPs solve every problem an online advertiser has.
With a DSP, we are told, we can put our banner ads anywhere across the Internet using a single interface. We can access every ad network at one time, using a single set of geographic targeting and frequency caps. We can even use cookie pools to retarget existing customers, or third-party data providers to precisely target our ideal new customer. And we can buy ad inventory in real time, instantly, without dealing with clueless account executives and endless price negotiations.
And what we’re finding is a sad fact: while DSPs are great for brand marketers (who really aren’t accountable for much), they often don’t work that well for performance marketers. Setup takes a long time, the interfaces are confusing, and worst of all, the return on investment often isn’t there.
Why don’t DSPs always work for direct marketers?
There are 3 main reasons that DSPs have trouble when used for performance marketing:
- Poor inventory quality. DSPs cannot access the entirety of display inventory out there on the Internet. This is because publishers are rightfully concerned that DSPs will hurt the value of their inventory by commoditizing access and exposing them to extreme competition. Since publishers can only get low prices (or sometimes no price) for inventory which they release to an ad network (and thus which can be bought through a DSP), the only inventory they provide is primarily preemptible, “Tier 2” inventory — inventory that they can’t sell for a higher price elsewhere. While display has always been a very difficult process of mining the gold out of loads of rubble, when you know the easiest-to-extract gold is already gone, many DSPs are only giving you a lot of worthless rubble.
- Poor targeting. It’s great imagining that all those third-party data providers know something we don’t, but where do they get their data in the first place? Very few websites know even the basics about you, like your age or gender, and most of those sites don’t share this information with data providers. Thus, most sites don’t know who is visiting them, forcing them to make a best guess model based on lookalikes. When you overlay behavioral or demographic targeting, you’re building models on top of models. Combine these opaque models with the usually higher price of using this targeting, and the price/performance ratio usually gets worse.
- High overhead. By overhead, I don’t just mean the time required to set up and learn to use the DSP, which is often substantial, since interfaces are still not mature. What I mean is the cost: a DSP often adds 10 to 15 percent on top of the cost of all the media and data you’re buying. When the underlying quality of that media and data is weak, for the reasons mentioned above, the higher price will hurt your ROI oven more.
What is DSP advertising good for?
I would say DSPs are good for three things:
- Brand advertising. If you need to spend a lot of money fast all over the Internet, then DSPs are great! Like Google Adwords, DSPs can suck up every dime you have and keep asking for more. For brand advertisers who are evaluated on spend volume (or reach, or frequency, or Gross Rating Points, or whatever), it’s perfect! For performance advertisers evaluated on ROI or CPA, not so much.
- Retargeting. DSPs reach all over the Internet, making them great for chasing a customer around the Internet for a few days. The danger with retargeting, however, is that it overstates its own value, since most prospects in your cookie pool have already sold themselves and would have bought with or without the retargeting. Direct marketers need to account for this, and make sure that they aren’t overattributing value to retargeting, and taking it away from the channels that actually made the customer aware of your product. (This is a complicated issue, so I’ve written more about the dangers of retargeting.)
- Site prospecting. The smartest way to use DSPs is to use them to find sites, locations, or demographics that have potential, and then test them to make sure they can deliver ROI. Once you find these, you can make direct buys with those sites, often getting better quality inventory or better targeting. Plus since you already know your expected ROI, you’re in a better negotiating position with the site publishers.
So how do we make display work?
Unfortunately, making display marketing work boils down to the same old tried-and-true methods: a lot of time, testing, budget, and commitment. Most display outlets won’t work for you, so you need to test lots of outlets and lots of creative. One thing you shouldn’t do, though, is lock yourself into long-term buys. Get in, test a lot, get out. A serious site or network should need no more than 2 to 4 weeks to prove its value. If you don’t get ROI by then, and your creative elements (both ads and landing pages) are proven, then it’s time to move on.
As mentioned above, DSPs can accelerate the prospecting process, but they are not the Holy Grail, sadly. The tropical hut will just have to wait for now….