Alex Willcock is Founder and CEO of Imagini Holdings Limited, an advertising technology company and makers of VisualDNA.
AdExchanger.com: Please give us a bit of history about VisualDNA. A new company? Or pivoting for new opportunity?
AW: I founded the business because I saw an opportunity to change the relationship between commerce and consumers – to put consumers in control and give them the opportunity to say what it is that they want and need and for commerce to respond accordingly. It seemed to me that eventually, users will want to be fully in control of all their data and I wanted to provide them with a means of doing this easily.
My background is retail and I had seen the compelling and often extraordinary patterns that exist in a consumer's buying behavior. I was inspired to create a visual method for people to tell the story of their life and connect them with the things that that wanted and needed.
The business began in earnest when I launched a simple site that enabled people to profile themselves and get some personality feedback. I sent out less than 40 emails to friends to promote it and in less than 6 weeks we had over 1 million users and rapidly another 4 million and now, over 20 million.
What problem is VisualDNA solving today?
We provide very rich data about the users that are visiting a given site and enable these users to be targeted with advertising and content to a degree of accuracy that is, we believe, currently unparalleled.
We understand why a user acts in the way that they do. BT products are good at understanding what a user has done, but not very good at knowing why. The implications are that we can infer user actions in a very compelling way.
VisualDNA gives publishers and data buyers the opportunity to understand exactly who is coming to which part of the site and target them on a one-to-one basis.
Our inference algorithms are able to predict the behavior of non-profiled users based on the actions of profiled users. For instance, for 100 % of users we can currently predict gender with a 69% accuracy, age band with a 48% accuracy and the 20 main interest tags of each user with a 46% accuracy.
What type of technology is Visual DNA using?
We profile users using our visual quizzes that are typically deployed as HTML microsites that are linked to from publisher sites or as HTML widgets that are embedded in publisher sites. We also have a full REST API that allows publishers to build their own front end if they wish. The API accesses our Service Orientated Architecture that includes a User Service and Feedback Service, among others.
Behind each image is a rich contextual taxonomy and series of behavioral axes, (over 40). With every click, the user produces more tags. On average, we collect 50-100 tags per quiz.
Once a user has taken a quiz we store their profile in our data store and assign a unique anonymous id to the user that we place in a visualdna.com cookie. At any time, users can visit my.visualdna.com to see, edit and delete the data that we have stored.
For example, our "Top stories for you" news site widget users our collaborative filter via our Relevant Content Service. The collaborative filter finds the users with most similar profiles that have accessed the site in the last 24 hours, collates the stories that those users have read and recommends these stories back to the user. Click through rates on our recommended content are 37% higher compared to showing the most popular stories of the day for all users.
We also integrate with major Ad Servers to enable publishers to sell impressions of specific audiences to their advertisers. For example, publishers who use DART can deliver ads to specific audiences using the Boomerang extension to DFP. It is straightforward for advertisers to buy "sports lovers" or "female, hikers" or "photography enthusiasts, family with young children", for example, and straightforward for publishers to traffic ads at these audiences.
Can you discuss a use case of how a consumer might interact with VisualDNA initially?
Consumers are invited to take a VisualDNA quiz via ads and editorial links on our partners' sites. They answer questions about themselves by choosing from a selection of images, and get feedback on what their choices say about who they are. They can also get relevant content matched to their VisualDNA, for example the product that we have launched for the LA Times filters the news stories of the day and shows the reader the most relevant articles for them. Try it here - http://latimes.visualdna.com
We've launched quizzes on a wide range of topics including Travel, Dating and Health. Our core Personality quiz consistently proves itself to be the most popular with consumers, which enables us to collect a very rich, broad set of data.
Our philosophy is that we put the user first; we always make sure that they get more than we take. How are you selling this data today? What is your target market in terms of clients?
Our clients are both sellers and buyers of data. Specifically, publishers that wish to increase the value of their inventory and buyers that wish to immerse their brands in highly relevant audiences.
We are working with a number of different partners to provide rich data on their users. We sell the products that create the data and often ring-fence that data for the partner, but this depends on the specific deal.
We have just launched a free Audience Analytics product that publishers can use to develop the content on their site and sell their ad spots as premium inventory - this is a free service. Our goal is to make this a perfect partner to Google Analytics, where GA gives you excellent data about the stats of your users, Audience Analytics tells you who they are, and what motivates them. We'll soon be providing tracking code and an inference service as well. We are also in discussion with a number of Ad Networks for to white-label this product.
We monetize VisualDNA by making the data available for real-time targeting via our API and integration into ad targeting platforms such as DFP.
Currently we are working primarily with publishers to help them collect and use the data themselves to personalise content and deliver more relevant advertising. We are now seeing great interest from Ad Networks, Demand Side Platforms, Data Exchanges and Media Agencies. Our work to date with Ad Networks has delivered some very exciting data, such as increases in CTR of up to 83% and increases in interaction rates that are over 700%.
We are in the process of opening up our data for buyers to target against and we'll be publishing more detail on this in the coming weeks.
Can you discuss your research methodology? How do you know what an image means for a user?
We're replicating the instantaneous decisions that we all make on a daily basis and, by looking at the pattern of these choices, we understand the core traits of an individual. It is the pattern that is all-important, as well as the individual image choice.
One way to think about it is to consider a consumer standing in a grocery store in front of a wide range of products. Not everyone buys the same brand and we make choices for a whole variety of reasons, such as color, font, packing material, lighting, price, brand values etc. If you then look at the assortment of products in someone's basket at the grocery counter, you will often see a pattern to their choices.
Our job is to make sure that with each question, we are placing a variety of realistic choices in front of the user. We score each image using a patented methodology and learn the patterns of choices based on the scores and tags that come from our taxonomy. I can't go into the methodology in detail as it is proprietary.
Responding to images elicits an instant emotional response that is inextricably linked to our core personality traits. We work with many different disciplines of psychology and have drawn in particular on the 'Big 5' Personality Profiling methodologies to develop our quizzes and guide the image selection process.
A quiz will capture a range of information from core personality traits like openness, creativity & extraversion, through to preferences around style and highly actionable information such as purchase intent and brand choices. We continuously analyse the VisualDNA data - from the 20 million profiles to date - to verify the consumer interpretation of the imagery. It is fascinating how consistent some images and concepts are in identifying a user 'type'.
We have proven that it is the concept of the image that is important and not the actual image. For example, when asked about the concept of freedom, one group will consistently chose examples of ‘being at one with nature' – whatever the image that is used to signify this.
We have found that the more abstract the concept, the more universal the images are. It's as if there is a primal relationship that we have with core traits. Other, more superficial subjects such as style, of course vary from country to country.
How does a publisher use your VisualDNA results? How does an advertiser use it - what might be typical targeting parameters?
A publisher can use our VisualDNA data in a number of ways. First, they can access the full profile data via our VisualDNA API allowing them to personalize a site or page as desired. Second, publishers can embed our recommended content widgets to deliver personalized recommendations to users. Our recommendation engine works with any type of content including page URLs, products, search queries and RSS feeds. We track the content that profiled users consume and then recommend back to users the content that the users most similar to them have also consumed.
Third, a publisher can sell impressions of specific audiences to advertisers. For example, a news site can sell impressions of ‘Sports Lovers' across their entire site. They can also be far more detailed, enabling audiences to be bought that display core motivations that fit a given brand's values and segments.
We have integrated VisualDNA with DFP Boomerang, to make it easy for publishers to sell audiences in the same way that they are currently selling placements. We are currently looking at other integrations.
We have a vast amount of data and can provide targeting parameters to fit a buyers needs.
Please give an example of expected lift for the publisher and for the marketer.
A publisher can expect uplifts in click through rates on content of over 30% and dependent on the Advertising inventory available, uplifts in CTR of over 300%. There are many other metrics though such as increased time on site, interaction rates and the pages viewed by profiled users over non-profiled users. We are seeing increases in page views per visit in excess of 100% currently.
What is VisualDNA's revenue model(s)? Do you share with publisher partners, too, or is the value in the reporting? Please discuss.
Dependent on the scale of customization, we provide standard audience analytics and insight free of charge and then charge on a CPM or share of revenue uplift when profile data is used to deliver personalized content or more relevant advertising.
The revenue uplift on a publisher site will come from increased page views and/or increased CPMs on ad inventory. In this way publishers only pay when our technology generates real value for them.
What's the difference between what behavioral targeting offers and VisualDNA's offering?
VisualDNA is focused helping users get a personalized web experience and more relevant content and advertising. We are able to collect accurate information about a user's interests, tastes and aspirations that are not necessarily detectable through their online behavior.
We understand the why as well as the what. We are also able to predict profiles for users who have not taken the quiz by comparing the behavior of these users to the behavior of users who have taken the quiz. Publishers can then deliver personalized content and more relevant advertising to these users too.
How do you enable tracking of a user's profile while maintaining appropriate standards are around PII (personally identifiable information)?
We don't capture any "personally identifiable information" about our Users and so we are unable to attribute VisualDNA data to specific individuals.
VisualDNA data, which includes the User's string of image responses to a VisualDNA Quiz, is only stored if the User gives us permission to do so.
Users that provide permission are cookied. Our cookies store an anonymous User ID and only permitted partners are able to access the VisualDNA data associated with that ID.
We are strong advocates of User privacy and seek to be transparent in all our dealings.
Follow AdExchanger.com (@adexchanger) on Twitter.