"They've got more than a decade of real powerful ad tech under their belt," he said. "This is very likely an opportunity for us to move much faster to build out the sophistication for ad products."
In the Q&A below he also fields questions about the targeting possibilities with Tumblr's audience, protecting the user experience, and his personal involvement in building ad products after selling his company to Yahoo in a deal that has been officially completed as of today.
Earlier this afternoon Karp took the stage and invited festival attendees to buck the trend of tightly regimented digital ad formats.
"One of the things that frustrates me about the role my community of engineers plays in the creative world is that for some reason we think we know anything," he said. "You guys have the history of making stuff that really moves us and inspires us."
AdExchanger: How do you make sure paid media placements reach the right communities of interest on Tumblr so they connect emotionally?
DAVID KARP: We knew we were not going to be able to nail the targeting in the first release. That was not an engineering specialization we'd built up yet. Tumblr was never about algorithmically recommended content. It was about human curation. We didn't have a lot of that automated or programmable targeting.
Rather than try to solve it with engineering, we solved it in a very non-tech company way, which was to go out to big brands that have stories to tell that can mean something to lots of people.
We went to the big Hollywood studios. We went to some of the biggest brands in the world who had stories to tell and ads to run that would actually mean something to most of the people that saw them.
It has been a year now that we've been building these products and this technology. Now we've introduced more granular and dynamic targeting that allows you to put something out there that will follow those communities it's performing in, or target the markets it's best suited for.
This was a big part of what we had been talking to Yahoo about, well before we were talking about acquisition -- a big part of where the conversation started for us. They've got more than a decade of real powerful ad tech under their belt. We were building from scratch.
This is very likely an opportunity for us to move much faster to build out the sophistication for ad products.
Could that involve tagging communities of Tumblrs?
Which are occasionally very hard to tag. Sometimes the overlap is so funny. Just think of photography for a moment. More than half of the content on Tumblr is photos today. But inside photo posts you have animation, cinemagraphs, photo sets, panoramas, illustration, digital art, photos of 2D art. You have people's travel photos and brilliantly composed photos. There's a tremendous amount of diversity there and Tumblr by its nature is eclectic.
Look at my blog. Some of it's my dog, some of it's my travel, some is the tech stuff I'm thinking about, some of it is dumb shit that I think is funny. Any number of things.
These are hard, interesting problems to solve. It's not just a single channel of video that's consistent every week, where you can predict who the main sponsors should be at that moment and take over the next 10 episodes.
Not that these types of problems haven't been solved before. My point is it's going to be a bit more nuanced than just tagging individual creators and saying, we're going to put this content next to the photos.
What about interest-based addressability?
We're experimenting with all sorts of tooling right now. So I don't know if I want you coming to me with tags, or with a set of example blogs, or example content. If you want to just enter it into the system and figure out where the algorithms go, but more and more we're testing it with different bits of advertisers.
How involved do you personally want to be with ad engineering?
There's no part of Tumblr that I'm not kind of obsessive about. I think I've gotten better as a CEO, to know to keep my hands out of things. At the same time I'm still a sanity check on a lot of the business stuff we're doing. Product stuff certainly. I wrote a lot of the code for Tumblr.
I haven't experienced a lot of this [ad] stuff before. This is the first time I'm going into the ad industry. The first time I'm working in a company that's this big. So I see things and say, why are we doing it like that? Occasionally the answer is, "That's how it's done." I'll say, "Let's get a couple of our guys over here to figure out, is there a more Tumblry way to do this that may be a little bit unconventional or might have some other considerations? Is there a chance to do something that's truer to who we are?"
Do you think you'll have ad engineers in house?
Totally TBD. I have a very basic understanding of how we'll work together on these things. It'll be a bit of an experiment, setting up folks on either side to collaborate on things they're doing really well, and things we're doing really well, and figure out how to leverage all the stuff they've got.
It's a really great time. I adore Marissa, and the team she's surrounded herself with. We have a few longtime Yahoos at Tumblr. Our head of sales spent 12 years at Yahoo building out their East Coast sales force. He was a big part of getting me comfortable with this. I asked him a few times along the way, "Is this really a good idea? Can you stomach the idea of working with these people again?" He said it was the best team he'd ever worked with. So yes, we are pairing up to some people who we are really going to enjoy working with.
Any thoughts on the right volume or frequency of ads per user?
We're testing all of this stuff, and we're trying to build really thoughtful relationships with those creative advertisers that are using Tumblr and we're excited to work with.
I really mean it when I say, they're going to show us the way. They're going to show us how great content and ads can be on Tumblr, and how non-disruptive they can be to everything else that's pure and special about Tumblr. We need to hear from them if we're ever not supporting them in their efforts.