Yesterday I spent some time on contextual vs. behavioral targetting. I mentioned that behavorial targeting involves gathering data on the sites a user has visited and what actions they may have taken, possibly combining it with information from external sources (including the visitors themselves), analyzing that data and coming to conclusions about that person that can later be used to serve them ads.
Google is using basic demographic data about particular sites to allow advertisers to select sites whose users have certain demographic characteristics on which to serve their ads. The data comes from comScore Media Metrix, which uses a Nielsen-like audience measurement system to determine popularity, and has collected data from its two million participants to come up with demographic information about the visitors to each site.
The problem with this type of behavioral targeting is that it is not truly user centric — it still revolves around the site where the ad will appear. I would call this enhanced contextual targeting, and it’s a good step toward where we want to be.
You can perform some data collection, analysis, enrichment and processing yourself that will help you better target the ads on your site. This should result in displaying more relevant ads and that should result in better clickthrough. Here’s how to do it:
1. Begin your research the same way you should any research — with a pencil and paper. Without over-thinking it, write down what you think are the characteristics and demographics of a person who visits your site — how old are they, what gender are they more apt to be, what is their station in life. Where are they when they surf over to your site (home, business, swimming pool, whatever). What do you think they hope to get out of your site? Why are they on your site and not someone else’s? What do you think they get from your site? Just keep writing until there’s nothing left.
2. Next go to your site logs. Start with your most viewed pages. What do they have in common? Write this down on another page. Move on to your top referrers. Visit each one and see if they have any demographic data. Find the contact email address and send an email. Explain who you are, thank them for referring so many visitors, and ask them if they’ve ever done any demographic research. You can ask them if they’ve performed an exercise similar to what you just went through. At the very least you should be able to make the same kind of demographic assumptions on their site that you made on yours. What do your site and the referring site have in common? What words describe someone who visits both sites? Sometimes this is very easy, and sometimes it is painfully difficult.
At this point you’ve probably done more research than most people and are ready to start tuning your ads. Spend some time on the major networks trying to find ads that you think will appeal to your perceived visitor demographics. Maybe you might want to tinker with your keywords to match your new conclusions.
Personally I would be dying to find out how close I was to reality with my assumptions, so here’s what I would do:
3. Enhance and validate the data . How do we find out more about our visitors? We can always ask them. The easiest way would be to pick some people who have emaied you about your site, posted comments on your blog, or subscribe to your newsletter. It is very important that you pick them at random without regard to gender, source, or timing. Send them a friendly email explaining what you’re trying to do (make your site more relevant to your visitors), and ask them if they wouldn’t mind answering a few questions.
Start off with easy questions that won’t make us New Yorkers suspicious that you’re trying to sell us something, rob our house or some other criminal angle (Where do you access the Internet from more often? Do you have a DSL line or dialup? What’s the website you visit the most?) and ease into the more complex questions that get to the heart of what you’re looking for. Any question that asks for more personal information should be offered as a range (How old are you? 18-24, 25-30, etc. or What is your income group? 0 – 25,000, 25,000 – 50,000, etc.). Before the first personal question remind them that the info is just for statistical analysis, they will remain anonymous, and no one will ever see the data. I also always offer a telephone number where they can call me and I can ask them directly. This seems to add an additional element of trust. No one has ever called me.
If you have no one to email, you might want to try a survey. People love to fill out surveys, especially if you tell them that you want to make your site more appealing to your visitors, and this will give them a chance to affect the content in a positive way.
Follow the same rules as you would if you emailed — start small and unthreatening and build. Make the questions multiple choice if you can, and ask the person if they would mind if you contacted them about their answers. Give the person a place to leave their email and remind them that you’ll only use it to followup. Then followup.
Knowing your visitors and thinking like your visitors are two keys to success.