It can be difficult to get a straight, simple answer from most companies in our industry about what they do – they/we tend to overcomplicate things. It’s what I like to refer to “Booth Babble syndrome.” You know, that tradeshow booth that contains an ungodly amount of text that says absolutely nothing. When asking the people attending the booth, “So what they hell do you do?” you’ll get a different answer each time. You lost me at “asynchronous” or “synergistic” or “bla bla bla.”
I think a lot of companies do this because there really isn’t a good answer as to what makes them different. They just don’t have a product or service that’s unique in any way. So the only real way to sell their useless products is to layer on lots of fancy, confusing words and attempt to baffle the buyer.
Some of the worst culprits are data companies and the metrics they use to sell to advertisers. I tend to look at online data in 2 simplified buckets: look-alike and do-alike. Advertisers tend to be very confused about the difference, but it’s something they need to know.
Look-Alike Data: Why Not Use It?
Advertisers would purchase this type of data based on existing profiles of people who “do” something advertisers desire. If soccer moms in the Midwest largely purchase your product offering, you would assume that other soccer moms in the Midwest would purchase it as well, right? Look-alike data is just a guess. Nothing more. Providers of look-alike data will make assumptions based on content viewed, IP addresses, offline databases… you get the point. They then build a profile and lump individual users (computers to be exact) into categories and sell that to you based on all these assumptions. Does it work? No. Not if you’re looking for performance. Does it work if you’re a buyer at an agency and you love the sales rep (and their expense account) and you can check the box for the client that says they did a targeted campaign? Sure!
I have a close friend who worked at one of the top digital agencies in the world. He said, “I was working with their DMP, comparing data from multiple providers, and the same set of users always produced conflicting results. Of course this makes it difficult to truly make the most out of the data. The predictive models relying on quality data would produce insufficient results that would propagate throughout the system.”
Do-Alike Data: Why Use It?
Do-alike data is based on a specific user’s actions, not an extension of that user’s actions. It’s like judging a landscape by driving to it rather than looking at a kindergartener’s drawing of it (e.g. look-alike segments).. Search data falls into this camp. If you’re doing it with Google search or with a reputable search retargeter, the challenge comes when companies need to create taxonomies that throw people into large buckets. If those buckets fit your product offering well, it should work. When a search retargeter takes the keyword “1968 mustang” and throws you into auto insurance, it’s not as likely. You could argue placing ads directly onto sites that are contextually relevant is do-alike as well. It’s a bit of stretch, but much closer than methods that determine look-alike. This is what AdSense is built on. The challenge comes into play for sites that have product offerings that are very niche, and offerings other than AdSense just won’t work.
Put on a practical hat when being sold offerings. Ask the simple questions about how and where this data might come from. If the provider can’t give you a logical response that really makes sense, run. Follow your instincts. If you’re after performance, you’re not going to get it from look-alike. If you’re just after checking off the box for a “targeted” campaign, then go for it.