Iterating Display Advertising, Optimal Part 2

Iterating Display Advertising, Optimal Part 2

In part 1 of the story, I started talking about starting CPM Advisors (which later became XA.net and then finally Optimal). From 2009 to 2011 we mostly focused on “display” advertising – ‘banner ads’ – on Yahoo!’s Right Media Exchange (RMX), Google’s Adwords and various other exchanges. The goal was a completely self-serve interface for buyers that allowed media buys to be made with a credit card across multiple display exchanges, where our system would optimize the ad allocations and get the best results for the customer. In no particular order, some of the things we encountered:

  • Icky tech choices. We wanted the app to feel integrated and stateful, and at the time Ajax-like UIs were nascent, so we chose Flex/Flash. It was initially a pretty good user interface and felt very different from 10-page ad creation flows that had previously frustrated me and our customers, but it bogged down and slowed development down since you had to recompile SWF files (!). At one point in 2010 I talked to a developer advocate for Flex/Flash at Adobe about promoting our app as someone doing something interesting with Flex. He liked it but thought that developers were unlikely to be excited since “most of them hate advertising”. Ouch!
  • Being called a “DSP”. Or “Demand Side Platform”. Yahoo! coined this term and conveyed it to us in a partner update meeting we had with them. They’d been looking for a name for tech companies who were somewhere in between an agency and an ad network model. It stuck.
  • Credit card shenanigans. We were lucky not to get any really big fraud hits on credit cards, but we learnt a lot more about chargebacks than we ever wanted to – and for a small startup a few thousand dollars can be a big deal. The main learnings here: a) it’s almost impossible for a tech/media company to be able to win a chargeback case with no physical product and a “terms of use” or “service order” and electronic signature to show, b) when someone pays with a credit card, the money isn’t really yours for up to 6 months after they pay, c) when you start ramping up and charging a lot on credit cards, get ready for your processing terms to change overnight. At one point thanks to our friends at PayPal, overnight we went to having 15% of all the non-Amex money we took in be held by PayPal for 120 days. This sucked big time, and the only fix would be to rip out our PayPal Payflow integration and use another processor. Luckily we were able to strike a separate deal with American Express directly, encourage users to use Amex, get a better rate AND still be able to use PayPal. This seemed bizarre.
  • Impressions fraud on exchanges. AdECN (while it still existed) and OpenX were targeted for impressions fraud by various unscrupulous folks running hidden 1×1 iframes. RMX had some companies selling ad inventory on it that seemed to manufacture impressions from thin air. Before we built our own view- and click-tracking system from scratch, we partnered up with the folks from IPonWeb who’d built a lot of the backend of RMX, in order to scalably track traffic performance at the line-item and site level. By having RMX macros we could insert into ad tags that let us dynamically track traffic, and see suspicious IP addresses, we were able to find potentially fraudulent traffic far more easily. But still, the amount of work this was was part of what convinced us to pivot the entire business to social media ads on Facebook/Twitter/LinkedIn.
  • Breaking the Doubleclick Exchange. One day early on we got a call from our Doubleclick Exchange rep, (after the Google buy but still on the old DCLK tech) saying that we needed to reduce the number of campaigns we were running because we had helped to crash their servers. We’d set up lots of campaigns, one per website, to (1) figure out which sites worked best, and there was no way to see performance at a site-only level unless you could restrict your campaign and conversion tracking to that site only (no site-ID macros!) and (2) try to figure out what all the “anonymous” websites were. We were able to do enough of this to figure out which sites actually performed and in some cases to find out that the same sites (many of which performed very well) were listed multiple times as both anonymous and attributed (in the latter case usually with a very high floor CPM). But what was great for many publishers was that we’d discovered the good quality inventory and were able to spend a lot more, and to get more of the good inventory we were willing to pay well in excess of those floor prices.
  • The Lumascape Before There Was a Luma. Terry Kawaja was doing the rounds with his logo-laden slide of the adtech universe when he was at GCA Savvian; it was brilliant marketing. Putting all those logos on a single page scared a lot of people out of our space. The good news was some of them were potential competitors, and the bad news was that a lot of them were acquirors or investors who now had their own version of the famous jam “decision paralysis” study.
  • Invite Media gets acquired for too little, too early. When Invite Media got acquired in 2010 for (apparently) somewhere between $70 million and $80 million, since some investors thought they were the market leader in a nascent space (DSPs) they then thought this set a ceiling on possible exits. This wasn’t great for us.

We hadn’t raised a lot of money, and were able to stay close to profitability for the first few years, albeit growing more slowly than many of our competitors who’d raised upwards of $5 million and $10 million. One of our investors, Mike Walrath, reassured me that fundraising and success were not the same thing. Our first year saw over $1 million of revenue which we grew to over $8.7 million in 2010, winning a San Francisco Business Times award for being one of the 10 fastest-growing private companies in the SF Bay Area, on the backs of (almost all) display advertising revenue. (I like to say we were one of the 10 fastest growing companies that applied… cause obviously some darn big/fast growing ones didn’t!)

By the time of that announcement, we’d already started the shift to social as you could see from that press release.