In-image ad platform GumGum has spent the past year building up its sales team, following last October’s $7 million VC round, which brought the company a total of $11 million raised since opening its doors four years ago.
Over the summer, Los Angeles-based GumGum brought in a new sales head: Robert Elder from Independent Television Networks (ITN), where he was VP of digital sales. The idea of in-image ads has taken off, especially as an e-commerce play, but for the most part, GumGum’s CEO Ophir Tanz is resisting that pull, hoping to appeal to its stable of premium publishers — Gannett, Hearst, Time Warner, EW Scripps, and others — saying that the company is mostly focused on branding campaigns, and has relegated direct response to a lesser extent.
AdExchanger: Gumgum is generally regarded as photo-based ad network. Is that how you see yourselves?
OT: We see ourselves as an image-advertising platform. But you’re description isn’t wrong either. Gumgum is about four-years-old now. We do work with thousands of publishers, and we specialize in helping traditional publishers figure out, programmatically, how to derive revenue from the billions of images they publish a month. On the demand side, we gather the data and sell highly targeted placements to brand advertisers.
What is the nature of the deals you have with publishers?
Our value is that we provide incremental revenue for major publishers. We place a one-time, one-line piece of code that the publishers install on their sites. There’s no additional work that the publishers need to do. From there, we figure out what the images are about, then provide ad-matching, serving and sales for the display ads.
Does the publisher ever handle any direct sales for the in-image ad placements you set up?
There are a number of large publishers that we work with and that’s why we don’t like to call ourselves an “ad network,” though technically, we are related to that model. But it’s not like any can come to the site and become a Gumgum publisher. You have to have a certain amount of monthly traffic, as well as site quality that is measured by hand.
Given that client profile, many of our publishers do come with very capable sales teams. So in many cases, we do give them the opportunity to sell directly into the in-image unit, while we do the serving and ad matching.
And what sort of conversations do you have with the agencies and marketers on the demand side?
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
We work with buyers at the keyword level understanding of what a particular image is about. So the agencies and marketers buy keywords from us at scale. It’s nice, because it’s a language that they’re used to speaking anyway. We do all kinds of semantic, facial recognition and skin detection to find the right placements. We do image-clustering so we know if we find an image that has a high probability to react to a certain keyword, we can distribute that all across our platform, even if it’s been cropped or altered in some way.
You place a lot of focus on the human quality of the ad sales. How much of the ad placements are algorithmically placed?
Everything we do is 100 percent algorithmically placed, though we do have people reviewing the images. We only deploy those people for highly sensitive campaigns, such as a diaper ad. There’s a lot of talk about the trend of audience buying, but at the core, I think people enjoy buying contextual ads more than audiences. For one, you’re actually able to verify that you’re getting what you pay for. And when you’re buying strictly audience, you don’t really have the visibility. You have to take it on faith in many ways. A lot of buyers have told that if they could, they would just buy contextual. But there’s not enough of those kinds of placements available to make it worthwhile.
If you’re an agency, and you’re buying a packaged goods marketer, if you can show them an ad with a given concept next to content that reflects well against that ad, it’s really easy to argue the value of that. Whereas the data/audience side, you can look at conversions, but there’s a lot of concern about what’s actually happening after those ads are placed.
So is there no value in behavioral ads for in-image ads?
Oh, there is. It depends on what the advertiser is after. There are some advertisers who are only interested in getting a user to pull out a credit card or in getting some sort of lead. But for the majority — and most of the money that’s spent on advertising — the goal is brand marketing, with the goal of generating awareness. But if someone is in the market for shoes, and if you can show that after every third time they’ve seen the ad, they buy a pair of shoes, it’s hard to argue that’s not an effective use of a marketing budget. I’m a big believer in both. And we do some of that behavioral targeting, as well as geo-targeting.
So how do you know if your brand advertising placements are working?
The click-through rates on our pages tend to be 10x or 20x versus the industry average. That’s not surprising, considering the level of banner blindness out there. But we’re putting ads right in the literal line of sight as the content — so it’s impossible to ignore. When we do our job well, we create an additive experience, because someone is looking directly at an image that they’re interested in. But we’re not trying to be the best solution for every advertiser out there.
Gumgum does do some direct response advertising, not solely branding. How do the two balance out in your marketing mix?
We know where DR works well, but we’re also philosophically opposed to the way a lot of the ways that stuff gets bought and placed. We don’t just want to go out there and get the conversion. The reason is that it’s deleterious to the publisher and user experience. We can’t go to The Boston Globe’s Boston.com, one of our publishers, and start running a bunch of crappy remnant ads. We’re much more focused on the branding side. And we do work with a lot of automotive marketers who are looking for leads. But that’s about as far as we go.
By David Kaplan
