ChoiceStream CEO Discusses Pivot Decision And Its Embrace Of Programmatic Ads

Eric-BoscoOne of the key strengths of a business is knowing when to pivot. For demand-side audience-targeting platform ChoiceStream, that involved switching from product recommendation software to programmatic advertising services in 2011.

Eric Bosco, formerly the company’s COO, took the reins as CEO last May. AdExchanger spoke with Bosco about the company’s transition.

AdExchanger: Where does ChoiceStream fit in the ad ecosystem?

ERIC BOSCO: We help advertisers find people who are most likely to convert through our optimization system that is designed to mine data to zero in on target audiences. That’s our machine-learning and results-oriented approach.

On top of that we have proprietary things like targeting that uses a survey-based methodology. We run a website called where we drive traffic to the site, which has multiple-choice-type questions about things advertisers would care about, like "Which is your favorite brand of hotel chains?"

How do you get people to answer those surveys and how many responses do you typically receive?

The current approach is a mix of surveys that grab people’s attention. What we tend to do is run a teaser headline that will grab a consumer’s attention and then expose them to other polls. One example was based on the government shutdown where we asked, “Who do you blame for the government shutdown?” People answered it and then they got another survey question, like "What cleaning products do you use?"

We get about 150,000 to 200,000 survey responses a month. We’re trying to get half a million responses by the end of the year. And for any advertiser we offer about a minimum of 1,000 responses.

Can you give me an example of how you use the data from the survey responses?

Dunkin’ Donuts, for example, wanted to reach Keurig machine owners. Dunkin’ Donuts had a coffee flavor for these machines and wanted to target those people. So we spun out a survey, “Do you own a Keurig machine?” And then we looked for lookalike audiences. It turns out if you own a Jeep, you’re five times more likely to also own a Keurig machine, so as a proxy, Jeep owners are also a good target.

What’s your approach to mobile advertising?

Last year we finished integrating with several mobile ad exchanges, like MoPub and  Millennial’s ad exchange (MMX). Most of our optimization engine now works on the mobile ad exchange front and we’re spending the bulk of our research on trying to reach people across devices.

How are you connecting users?

It’s essentially by guessing. Let’s say you have a smartphone and a laptop and both have the same residential IP address that’s probably shared with other people in your building. So the IP address by itself is not necessarily indicative of the fact that those devices belong to you. But we’ll tie in other activities like time stamps of the sites you’ve visited and other data sources.

What types of data are you analyzing?

We’ve just started our efforts in mining the data sets to come up with our approach to this, but basically, things like the user’s IP address, the geography data of where their device is coming from, the operating system, browser types. All those things together can be a good starting point as to where you can identify whether a user is the same person across various devices.

Do you see yourselves at a disadvantage compared to mobile-first companies?

On the mobile front, I don’t think mobile-only providers will make sense in the future. What we see more and more is advertisers want to work between mobile and display as a continuous spectrum, and so the mobile-only ad networks are going to have to get their display-based counterparts.

So when you look at folks like us, who started in display and people ask, aren’t you late to the mobile party? We don’t think so because the display aspects are hard to get right and as we add mobile capabilities, I think the ability to do both at once is more compelling than just doing mobile. We’ll see if this pans out since that’s what we’re shooting for.

What opportunities do you see from the programmatic side?

We built our machine-learning system in 2011 and took it to market in 2012, which allows us to drive results in the most efficient way for advertisers. Our focus is taking this and scaling it. When it comes to the ad exchanger and real-time bidding environment, we’re still in the early stages.

In terms of the survey option, the majority of our advertisers end up not using that capability because many tend to know what kind of audience they want. When we do use our survey capability, we tend to use it to prime the pump of ad optimization.

What’s your pricing structure?

Most of our business is mainly CPM-based. We do take some business on a CPC or CPA basis, although that’s pretty infrequent for us.

Are there any advantages for you to charge on a performance-based metric, like a CPA?

From our side of the fence, the difference between a CPM and a CPA basis is the type of conversation you can have with advertisers.

If you price on a CPA basis, you might end up underdelivering, because at that price point, if the advertiser wants to pay $30 for whatever the action is, that might be too low to drive substantial results, whereas if they were willing to pay $50 they could have driven more volume. And so part of the trick is to understand how hard or soft is that performance goal for the advertiser.

What are the mistakes advertisers make with CPAs?

I think many advertisers actually have an acceptable range for what the CPAs can be. And if they have a range, as a vendor, I think you’re much better off pricing on a CPM basis since it allows you to say, “Here are the trade-offs. Do you want us to spend more or less?”

If an advertisers has a set figure, like $82, then you’re better off with a CPA basis, because then you are by definition always delivering what the advertiser wants and you’re not overcharging them for the conversion and can figure out how to grow the campaign from there.

Where are you in terms of revenue?

You’ll find a figure on TechCrunch’s (CrunchBase business profile) that’s around $60 million or $70 million dollars, but that’s relatively misleading.

ChoiceStream has been around for about 13 years now and we did a complete pivot in 2011. So the first 10 years of the company was in product recommendation software and we were competing against companies like Rich Relevance.

None of those guys are making real amounts of money though since at the end of the day, people can get that almost as a free add-on from Adobe or CoreMetrics. So in 2011 we did a full recap of the company, the founder (Steve Johnson) bought out the previous investors and we’ve been backed by our founder primarily. We’re in the process of fundraising and will most likely have something to announce in the next few months.


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>