What Is The Biggest Misconception About The Use Of Data For Ad Targeting?

One QuestionOften, a question doesn't have an easy answer in the digital advertising business. This is a column devoted to an answer to a single question - and providing a bit of space for it.

Today's participant is Hooman Radfar, CEO of Clearspring, which provides content sharing platform AddThis. He recently answered the following question during a conversation with AdExchanger.com...

AdExchanger.com: In terms of data, the use of data and ad targeting, what would you say is the biggest misconception out there?

HR:  Where do we start?  All joking aside, there are so many misconceptions.  This is a relatively new and rapidly growing area.  Things are changing fast.  That said, I would say the single biggest misconception is thinking that one particular data type is the answer.  And here’s the punch line – it’s not.

I remember when search retargeting first came out.  People said search is all you need – it works for Google, right?  Search is a powerful signal of intent.  But all things being equal, if you were to say, “I'm only going to use search data - I'm going to eliminate all other data that could give me indication on purchase signals – that just doesn't make sense.”

So you have these companies, whose entire thesis is built around a particular data type. Advertisers and agencies need to think longer term.  I'm not arguing that the entry positions of single-data point companies doesn’t make sense, or even that their data is not incredibly valuable.   Everyone is evolving.  But to just hinge your long run targeting methodology on a single data type is simply myopic.

You can apply this same line reasoning to social targeting.  Using the social graph for targeting, which is powerful data source, certainly works.  But the sole use of social graph-base targeting means you're missing out on the powerful signal afforded by search intent, for instance.

Today, when you look at data platforms, advertisers and agencies just really need scale and powerful signal to start.  There are very few companies that have this high-value data at massive scale.  So this formula works for today.  Like all industries, however, we have to evolve.

We will move from a world of single data types to the vision that Gil Beyda described, Data 2.0.  In that world, multiple data types will be scored over time and across campaigns to create high-performance audiences.   In the future (read: near future) we won’t be talking about search, behavioral, or even social – we will be talking about what works – period.

Instead of data being a one-man show, we need to treat targeting more like a symphony, using various types of data to create a unique and perfect harmony in a campaign.   When appropriate, you will use data on attributes that apply to individuals in the graph, whether it be the person's intent, or interest.  In other cases, you may apply knowledge of their affiliations with one another.

To me, that's the future.  And it kind of makes sense because it’s a simple perspective -  – use the data that performs best for you in a given campaign at a given time.

So, I think that is misconception number one.

A second and related misconception is that Audience Targeting - whether it is search targeting, social targeting or all of the above - is the panacea for targeting.   That is simply untrue.  Context matters.  I am sure that the folks in Mountain View would agree.   Again, you have to look at these things in concert - there's a violin, a cello and all these different instruments to play.   There is of course, the problem of reconciling attribution and payment, I am sure someone smarter than me will figure that out.

Today, we’re experiencing a massive surge in interest in audiences.   But the people that win in this game, will not only leverage multiple data sources to build the best audience, and also leverage other techniques like contextual targeting to deliver the best result.

Follow Hooman Radfar (@hoomanradfar), Clearspring (@clearspring) and AdExchanger.com (@adexchanger) on Twitter.


  1. Good piece Hooman. I think there are a number of points about context that play off your paragraph on it:
    -sometimes context matters and like your data argument, eventually we'll get a sense of what the relative value is.
    -contextual buys are historically made as a proxy for audience, and in a lot of cases, it's not a great proxy, so the push to audience targeting implies that in some cases, data outside the context of the ad is a more effective proxy. Buyers want what efficiency of media against a target. Said another way, we're all trying to prove Wannamaker wrong.
    -buying quality context in some categories can limit spend and be expensive, so audience targeting is a viable extension (i.e. auto, biz traveler)

    again, good piece.

  2. These are great points, Hooman. I agree that data isn’t just a one-man show. Data has always played a significant role in developing ad campaigns, but as consumer buying habits change, so does the advertiser’s approach. Brands are certainly dedicating more and more resources to data optimization and ad targeting, but the technologies they seek to support this process must consider personalization in addition to all data types and contextual targeting. To your point, in order to be truly effective, you have to take the whole symphony into account.

  3. Nate, thanks for the comments. You really hit the nail on the head. Many people use context as a proxy for audience. That is simply not correct. Just because I am reading a story about a particular news item, it does not necessarily mean that I am interested in all articles pertaining to that topic. Context is important IMHO because it is the current 'frame of mind' that the user is in. My thinking is that by leveraging the right audience with the right context, you can actually get some order of magnitude changes in lift.


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>