Chapell & Associates' president talks with WhenU's CEO about providing consumers value, branding, privacy and more.
Alan Chapell: I remember reading an article you wrote last fall that discussed, in part, how far we in the online space are from delivering on the promise of giving consumers something they actually find of value. And I think to some extent, concepts such as "consumer in control" and "relevance" are way over-used in some circles. How does WhenU ensure that there's real value provided to the consumer?
Bill Day: Well, traditionally, value to consumers has been measured through clickthrough rate. The Google breakthrough in the search space was to look at clickthrough rate as an arbiter of success-- if the user is clicking, that's a vote of confidence. On one level, that works for us too. So we've been very focused on driving clickthrough, which is about showing fewer impressions and making each impression count.
When I'm speaking about fewer impressions, I'm talking specifically about the frequency caps we put in place in late '04, which was a novel concept back then. The idea that you had perishable inventory, and not unlimited inventory, was a new concept, but it forced you to think about how to treat the user right. And the second part of that was the TrueRelevance technology we rolled out really at the start of the fourth quarter last year, which has allowed us to offer a "push" approach that makes comparison shopping effortless for the consumer, resulting in tripling of clickthrough rate in some categories to above 15 percent with consistently high conversion rates-- that's proof-of-concept that consumers find real value in these ads.
But there's also an underlying value proposition for the consumer aside from receiving relevant, money-saving offers, because WhenU software comes "bundled" with content or software that the consumer is able to access for free because of WhenU's ad support. WhenU continually reminds consumers of this value exchange by putting the logo of the partner download right on the wrapper of the ads (for instance, "WhenU Save… in support of Radlight Media Player").
Chapell: When you talk about clickthrough rate, I think direct response. Do you see a branding component to what WhenU does, or are you strictly direct response?
Day: I do see a branding component. But I think the truism is that even when people say they're buying on a brand basis, they are still buying based on performance. So we always look at clickthrough rate across those campaigns and look for ways to optimize and balance what the marketer wants and what the consumer wants. WhenU's "Shaper" technology has been used for branding very effectively (we can "drive" an SUV across the screen, for instance) but we still want you to click.
Chapell: What's your privacy model-- not simply your privacy policy, but what's your model for ensuring privacy?
Day: We feel pretty ardently that the right way is not just to focus on a privacy policy, but to view privacy as part of your overall strategy, or from a technical standpoint, part of the architecture. Our goal is to capture as little data as possible, and if you can target effectively without capturing any information, then all the better.
WhenU has succeeded in creating an architecture that doesn't require the transfer or collection of any personally identifiable information. Similarly, it doesn't require a lot of other anonymous behavioral data that is typically collected to enable targeting. We don't use tracking cookies or PIE (Persistent Identification Elements); there is no server-side profiling, anonymous or otherwise-- which is a fundamentally different approach in our space. We're zealots about this, frankly. It can sometimes be much harder to do things this way. But it keeps us very clean. From a consumer standpoint, this leaves us very well positioned. That's what we mean by a privacy model: do you need to transfer, collect or track user information to target effectively? Can you give more while taking less? I don't think most people really start by asking these fundamental questions.
Chapell: I've always thought of WhenU as less of a BT company, and more in line with a real-time contextual targeting company. Do you agree with that assessment, or am I just splitting hairs here?
Day: We do both. I don't think you’re splitting hairs. I think the marketplace has latched onto the idea of 'behavioral targeting.' One can logically extend behavioral to include behavior at that moment. So contextual is a subset of behavioral. And it ranges up to behavior over the last 30 or 60 days.
What's interesting is that the half life of behavioral data online is very short. If you don't have a behavioral model where you can act upon the data in the first 24 hours, you lose a lot of potential clickthroughs. It isn't like the offline world where you can build a model and categorize people as 'frequent travelers' and have that be actionable over a three- to six-month time period.
So I'd argue that the behavioral model online, it may be 24, 48 hours max-- it's a relatively compressed model. It argues for having a real-time component to what you do, and how you update your scoring on the user-- because users change. A fair amount of our business right now is done contextually. But we also run behavioral campaigns, though all of this is done in a relatively compressed time frame. The longest we'll run something is about 30 days.
Chapell: A few companies in this space have started talking about the concept of providing the end user with some level of control over the profiles being built on them and/or the types of ads that they'll see. Is WhenU currently thinking about doing this? What are the challenges around providing this level of access to consumers?
Day: I'm not yet convinced that most people want to be that actively involved in managing their personal profiles or the types of ads they'll see. I think you can split this into two different things.
The first piece is allowing users to actively participate in improving the model. What TACODA and others are exploring is the idea of sharing the data. They're saying: here's what we have collected, or why you're seeing this ad.
On the first one, we're actively experimenting with it right now. I don't think the majority of users will actively get involved in fine-tuning their ad targeting. There's some that will, and there's an advantage for us and for them in doing that. We've been testing the ability for users to do very simple things, such as: "Even though it looks like I'm interested in travel, I'm not-- so don't show me any ads about travel." Those are things that are exciting to us. On the other hand, not every user is going to use this, and you have to have a very small, casual model.
The other thing is just sharing the information. We have the ability, but we don't do it right now, because it wasn't built in the model. It's not in a format right now that we could put in front of consumers -- the data is stored on the consumer's machine -- but a subset of this could be done pretty easily if it became something that consumers wanted. Consumers may be more interested in seeing data that is actually being harvested from their machines-- but since WhenU keeps the data on the user's machine rather than collecting it on our servers or sharing it with third parties, there may not be much value-add to the consumer in being able to view that data, since it's not being sent anywhere.