Saturday, September 09, 2006
AOL Search Data for Advertising 2.0
The release of the AOL search database this summer was without a doubt one of the most scary privacy revelations in recent memory. I only know one person that is an AOL subscriber and it took me less then 2 minutes to find and verify his search history. I'm not kidding, 2 minutes. I performed one generic search and found someone's history that looked like his. Then I verified that it was him by taking one of the unique search terms and seeing if anyone else searched it. Its was him. I then showed him the history and he confirmed my conclusion.
Now, I don't know a lot of personal history of my subject. Its someone that consults with my company. I'm in contact with this person far less then any co-workers, family or clients. Two minutes... Truly amazing this data was released.
What I found while reviewing the AOL data is that most people's search histories are very personal and detailed. That this data can be used as a diary or biographic time capsule for everyone. Imagine how much detail AOL, Google, Yahoo, Comcast, etc. know about their customers and worse, how little the general publics' understanding of the impact of such data. How easy it would be to discover a co-worker's or neighbors deepest secrets. "Be kind to the programmers, for they hold more power that most will ever know."
So, how does this relate to Advertising 2.0? Well, once you can get beyond the voyeuristic nature of reviewing these search histories, you will see that all searches fall into one of six categories:
- Local Search; Restaurants, shops, people
- Medical Searches; Health, symptoms, drugs
- Product Searches; Cars, electronics, services
- Entertainment Searches; Movies, music, celebrities
- General Research Searches; History, how to, educational
- And Sexual Searches; Services, fetishes, desires
With this in mind, we can begin to analyze individual people at a micro level. However, doing this automatically is by no means accurate. Many histories are not just a single person, but rather a family. I found may histories that jumped from detailed medical searches (assumed to be mom) to online games (assumed to be a child) to home repair (assumed to be dad), all within a close time proximity. In addition to multiple people per profile, there are short and long term time cycles. A short term time cycle is how a user moves through the Internet during a single day. Different people search for similar things during different times of the day. Similarly, long term cycles are monthly, annually and life time. A good example of long term cycles would be a search pattern that looks for "3 year old birthday party ideas." That search may be repeated over and over for 3 weeks and then will not likely ever be repeated again. That's very different then a local restaurant search which may be repeated every 3-4 weeks for years.
Again, what does this have to with Advertising 2.0? Well, it has to do with the whole concept of providing the right information to the right viewer at the right time. What is the 'right time?' What is the 'right information' and who is the 'right viewer?' Certainly, static ads will almost never hit the mark. Single term keywords are not the solution. And, 'closed loop' advertising campaigns, where the advertiser's knowledge of their own clients is not used, will not succeed either. What is needed is a strategy of continued and concerned review of the online opportunities available along with continued and concerned review of real-time viewer activities.
Now that I have access to the AOL database, as well as my own databases of information, I use this data to confirm the validity of my search terms. What is the thought process (as identified via the proceeding and following searches) of a user that types in a particular search term? Are there other terms or sites that are frequently identified as pre-cursers to typing in my selected term? If so, maybe buying these terms and media on these sites will provide me a higher ROI or will get me in front of my target BEFORE they see my competitor. Maybe there are complimentary terms that are not on my list or complimentary products that I can partner with. And maybe, there are different patterns that new customers have as opposed to current customers.
Has anyone else looked at the AOL data and found additional benefits, patterns, or processes? Feel free to add comments. Thanks.
Now, I don't know a lot of personal history of my subject. Its someone that consults with my company. I'm in contact with this person far less then any co-workers, family or clients. Two minutes... Truly amazing this data was released.
What I found while reviewing the AOL data is that most people's search histories are very personal and detailed. That this data can be used as a diary or biographic time capsule for everyone. Imagine how much detail AOL, Google, Yahoo, Comcast, etc. know about their customers and worse, how little the general publics' understanding of the impact of such data. How easy it would be to discover a co-worker's or neighbors deepest secrets. "Be kind to the programmers, for they hold more power that most will ever know."
So, how does this relate to Advertising 2.0? Well, once you can get beyond the voyeuristic nature of reviewing these search histories, you will see that all searches fall into one of six categories:
- Local Search; Restaurants, shops, people
- Medical Searches; Health, symptoms, drugs
- Product Searches; Cars, electronics, services
- Entertainment Searches; Movies, music, celebrities
- General Research Searches; History, how to, educational
- And Sexual Searches; Services, fetishes, desires
With this in mind, we can begin to analyze individual people at a micro level. However, doing this automatically is by no means accurate. Many histories are not just a single person, but rather a family. I found may histories that jumped from detailed medical searches (assumed to be mom) to online games (assumed to be a child) to home repair (assumed to be dad), all within a close time proximity. In addition to multiple people per profile, there are short and long term time cycles. A short term time cycle is how a user moves through the Internet during a single day. Different people search for similar things during different times of the day. Similarly, long term cycles are monthly, annually and life time. A good example of long term cycles would be a search pattern that looks for "3 year old birthday party ideas." That search may be repeated over and over for 3 weeks and then will not likely ever be repeated again. That's very different then a local restaurant search which may be repeated every 3-4 weeks for years.
Again, what does this have to with Advertising 2.0? Well, it has to do with the whole concept of providing the right information to the right viewer at the right time. What is the 'right time?' What is the 'right information' and who is the 'right viewer?' Certainly, static ads will almost never hit the mark. Single term keywords are not the solution. And, 'closed loop' advertising campaigns, where the advertiser's knowledge of their own clients is not used, will not succeed either. What is needed is a strategy of continued and concerned review of the online opportunities available along with continued and concerned review of real-time viewer activities.
Now that I have access to the AOL database, as well as my own databases of information, I use this data to confirm the validity of my search terms. What is the thought process (as identified via the proceeding and following searches) of a user that types in a particular search term? Are there other terms or sites that are frequently identified as pre-cursers to typing in my selected term? If so, maybe buying these terms and media on these sites will provide me a higher ROI or will get me in front of my target BEFORE they see my competitor. Maybe there are complimentary terms that are not on my list or complimentary products that I can partner with. And maybe, there are different patterns that new customers have as opposed to current customers.
Has anyone else looked at the AOL data and found additional benefits, patterns, or processes? Feel free to add comments. Thanks.
