In 2007 we made our first ad tech investment, which was a company in New York called Collective Media.  Since then we have been one of the leading groups in the country for funding ad tech companies, with a portfolio that at one point contained 16 active investments in the sector.  I wrote this blog post to explain what all this stuff is about in simple terms, or at least as simple as I can make it.

The key observation that spurred all the VC investment in this sector is that digital media enables a new type of media buying.  In the “old days” most advertisers bought media against specific content (i.e. “I want to run Honda ads on ESPN.com”).  Today advertisers are focused on buying audiences, such as “males, age 18-34, who are in the market for a new car”, wherever they appear.  Audience targeting produces better results than content-based targeting, although it requires a lot of know-how to correctly plan and execute the campaigns.

The easiest way to think about audience targeting is to look at it like a Chinese menu:  in order to execute an audience campaign you need to pick one data option from Column A, one algorithm option from Column B, and one media fulfillment option from Column C.

There are no strict rules for constructing the perfect audience campaign, but there are guideposts.  I think marketers should start with Fulfillment (Column C) and go from there, because everything stems from the type and cost of media impressions you are buying.

The reason that agencies and marketers use a rep network (like Gorilla Nation) or a network with proprietary inventory deals (like Collective) for fulfillment is that you get high quality media, you can run ads with embedded video, you can plan campaigns in advance, and you are guaranteed to fill your campaign.  All of these things appeal to brand marketers.  Brands allocate budgets annually and they are accustomed to buying media in bulk to get a better price.  The DSP bucket (DSP stands for Demand Side Platform) is kind of like an electronic stock exchange for buying media.  The exchange consists of remnant inventory, which is left over after the direct sales and ad networks go through it, so it includes some less desirable sites or locations on the page.  However there are certainly good deals in there and it is significantly less expensive.  This is enticing for a subset of marketers, particularly those in the direct response field (lowermybills.com, expedia.com, bankrate.com).  Advertising agencies can also fulfill campaigns by themselves in the DSP market, which means they can capture the money that otherwise would go to a fulfillment provider.  Agencies are making a lot of money this way – on a percentage basis they make more money from their DSP than they do on any other type of media buying.  This has helped to rapidly accelerate the growth of the DSP segment.

In  Column B you have the targeting algorithms, which are the secret sauce to making data and media perform well for marketers.  Almost every algorithm involves some sort of retargeting.  A common example of retargeting is when you go to Zappos and put a pair of Nike shoes in your check-out cart, but you leave before you make a purchase.  If Zappos uses retargeting, you will see Zappos ads with an image of Nike shoes following you around the Internet for days.  The same thing happens when you go to cars.com to research a new car or go to Orbitz to book a flight.  There is retargeting based on almost every element of user behavior, both online and offline.

In column A you have all the data companies.  These companies either have relationships with publishers that allow them to gather large amounts of consumer data via cookies or they have direct relationships with consumers.  The data by itself is not valuable, but after it is fed into an algorithm it can produce a unique signal for buying media.

Some trends worth noting:

1.) In the last few months almost all of the data companies have stopped selling their data to third parties, and instead they have begun to build their own algorithms and run their own campaigns.  The algorithm vendors made a similar move a long time ago.  It is clear that the future of the business is vertical integration.

The primary reason for this is that the market for media is 100x larger than the market for advertising data.  By turning themselves into full service ad networks they can capture a larger amount of money from clients.  Media is also the currency of the industry – at the end of the day advertisers and agencies buy media, and they expect the data and targeting to be included in the price.  If companies are not having that conversation with a marketer directly they are at risk of disintermediation.

2.) As the data sources dry up it will be harder to execute campaigns on a DSP platform.  I expect that some people will default to a “data blind” approach to buying media. The easiest way to explain this is to use an analogy from the board game Battle Ship. Either you can buy data that tells you where the boats are (aka audience targeting), or you can carpet bomb the entire ocean.  You will turn up boats either way, but carpet bombing is much more expensive.

3.) One last general point, the ad tech industry matured very quickly and a lot of the innovations from 2007 and 2008 are now widely deployed.  It takes 18-24 months of heads-down development before a new product can reach parity with the existing solutions in the market, and that can be discouraging for many VCs and entrepreneurs looking to start ad tech companies and get to market quickly.  I expect this will slow down the pace of VC investment in the sector over the next five years, at least compared to 2006-2011.

I hope this is helpful for folks who read my prior posts on the advertising industry and couldn’t make heads or tails of it.  This is just one of the sectors we are focused on but I have tried to cut through the clutter.