A mysterious black bird that holds out the promise of great wealth. It’s the stuff that dreams are made of, according to detective Sam Spade in “The Maltese Falcon”. It’s also the perfect way to make sure auto sales Web site Autobytel was able to uncover the riches buried in search engine marketing.
“Black Condor” is the half-joking nickname for the automated search marketing platform developed by Irvine, CA-based Autobytel to take over its keyword bidding needs. And the bird seems to be the word in building traffic to the site through search ads: Visits to Autobytel through search engines rose 79% in 2004, and page views through those engines jumped 67% from the previous year.
In the beginning—meaning 2002– Autobytel was managing about 10,000 to 15,000 auto-related keywords in-house. That slate of search terms was under the in-house management of a single employee who could only devote part of her effort to the job; to make that large task more feasible, they were being managed in “buckets” of terms rather than individually. Searchers clicking on paid listings were directed to a series of landing pages that gradually drilled down to the make and model car they were interested in to begin with.
But about a year and a half ago, the company recognized that this was quickly becoming less than an ideal arrangement. “For one thing, we were missing a whole bunch of words that we should have been bidding on, “says Mike Rosenberg, senior vice president of media and marketing services at Autobytel. “Second, by bidding in campaign buckets, we were missing out on the visibility and transparency we needed to make sure our search marketing was efficient. We were bidding on words that weren’t profitable and should have been bidding even higher on words that were very profitable.” So Autobytel went looking for some way to both scale up their search engine marketing (SEM) campaign efforts and at the same time to make them easier to handle.
At that time, the company considered outsourcing to an SEM agency and went out to interview the leading agencies in the field at that time. But they all came up short in terms of being able to accommodate Autobytel’s specific needs for strategizing an expanded search campaign.
The biggest obstacle to outsourcing its SEM was the fact that Autobytel earns revenue in two different ways. One is lead generation. The company gets income from persuading shoppers to sign up and request price quotes on cars from dealers in their area, leads that Autobytel then passes on to its participating dealers for a price. As Rosenberg puts it, “We like to call ourselves the local search engine for auto dealers.”
But Autobytel also runs a robust advertising business with car manufacturers and thus derives income from selling ad impressions as well as leads. “If someone comes to our site and doesn’t submit a lead but looks at 20 pages, because they go to a photo gallery or read a review of a car model, that’s still economically beneficial to us,” Rosenberg says. For example, the term “Ferrari photos” won’t generate many purchase conversions; Ferrari shoppers are few and far between. But someone clicking on the term may easily look at 15 page views of car pictures, making them a visitor Ferrari might pay to reach.
It also makes figuring return on investment (ROI) a more complicated proposition. Many other search marketers have only one type of transaction occurring on their Web sites. When a searcher clicks on their search ad, those single-track marketers have no problem figuring out if that transaction has taken place and therefore what the value of that keyword is to them in ROI. But since Autobytel earns revenue both by generating leads among buy-ready car shoppers (cost per action or CPA) and from ad views (sold on a cost-per-thousand or CPM basis), they needed to take both those streams into account in determining their actual and optimum search ROI.
“A lot of the SEM firms we looked at had great models for handling the transaction-based business, but none of them had the ability to reflect the twin revenue streams in calculating ROI,” Rosenberg says. “So we decided to take a crack at building our own model.”
You heard right: When outside SEM solutions came up short, Autobytel built its own automated search software.
Rosenberg is quick to allow that not every search marketer is going to be able to roll their own software platform. But the company had a handful of very capable IT developers who had worked on its automated system for forwarding leads to dealers—“You know, the really smart guys who like nothing better than diving into a complex problem and figuring it out”—and an executive leadership that was both committed to expanding SEM and open to a home-grown solution if one was possible.
Cut to two years later, and the IT department came up with a search marketing application tailored for Autobytel’s needs. Jokingly referred to as “Black Condor”, this platform relies on application programming interfaces (APIs) from both Google and Yahoo! to get direct feeds from those engines on whether a keyword gets clicked on, how much it costs and what happens as a result.
The new automated SEM bidding platform allowed Autobytel to rev quickly from 15,000 keywords to 150,000 to 200,000 terms, something that would have been impossible to do manually.
Just as important, Black Condor is able to generate landing pages appropriate to each of those keywords. That’s a big labor saving over the old manual system, in which the terms “dealer”, “Miami dealer” and “Miami Ford dealer” would have required three landing pages, all created by hand. Condor has landing three or four basic landing pages for terms such as “dealer”; when someone clicks on “Ford dealer” or “Mustang dealer”, it takes that standard page and customizes it with modules of more specific information.
“Now, Black Condor can tell when someone’s looking for a Ford or a Honda and serve up a page with that relevant information,” Rosenberg says. “A year ago, we were just serving a page that said, ‘Buy a new car.’ You just couldn’t generate all the landing pages needed for the geographical and make and model variations.” And better landing pages mean stickier sites, and presumably more conversions.
Now Rosenberg’s department calculates the desired ROI for each keyword, figuring in both its value in CPA and in page views. They enter this desired value into Black Condor and the software determines what level bid will achieve that ROI.
“The software drives towards an ROI number for each keyword,” Rosenberg says. “If we decide that we want a 35% ROI on every term, it will raise or lower the daily bid, using historical data, to get into that sweet spot. It figures these bids every day, and then we send the batch of new bids to Google and Yahoo!”
The software also automatically eliminates bids on keywords that have not shown acceptable return after a period of time that Rosenberg’s department can set. “It eliminates bids on words that are just hopelessly unprofitable,” he says. That mitigates the “death-by-a-thousand-cuts” effect that can come from spending on buckets of keywords, some portion of which will never earn a good return.
Besides optimizing keyword bidding and streamlining the production of landing pages, Rosenberg says, one of the most significant benefits of Black Condor has been freeing up its human masters to strategize about what keywords to tackle. “Now we have time to make sure our keyword list is complete and up to date for seasonal events and new car launches,” he says. “For example, if the New York Auto Show is coming up, we want to make sure we bid on terms around the cars that are being previewed there. That’s something Black Condor will never understand. Those commonsense things are better done by humans; this machine simply gives us the time to devote to doing them.”
Black Condor’s reporting also allows Rosenberg’s team to look into the details of a search campaign over a period of weeks and perhaps gain some insight into where they should be directing more effort. Simply by looking at the 25 most popular search terms in a given month, Autobytel was able to detect that keywords around hybrid cars converted very well and allotted more resources to marketing on those terms. The same applied to the keyword “sprinter”, which is a generic term for a workhorse commercial van.
“We never would have found those terms if we had outsourced our SEM,” Rosenberg says. “When you do that, you don’t get the visibility into the campaigns that Black Condor gives us. Those agencies don’t live it every day. They don’t say, ‘Let’s make a query on this term and see how it performs, or let’s sort terms this way and see what we learn.’ Doing it ourselves allows us to get valuable insight into search marketing that I don’t think you can get when you outsource. If someone does a search on a ‘sprinter’, they’re pretty motivated to buy. But I don’t think we would have found that term from looking through an agency’s reports.”
“We discovered that half of our top-performing keywords were related to trucks. But if you don’t sort the data in a certain way, you don’t find out things like that.”
Rosenberg makes it clear that he is not inherently opposed to outsourcing when it makes sense. For example, Autobytel outsources the optimization of its Web content aimed at earning higher rankings in natural search. Effective Web site optimization is more like a work of art than a math problem, he says, and it was more cost-effective to send that work out and have an Autobytel employee liaise with the optimizing agency than to develop that talent in-house.
But he is also adamant that for Autobytel’s unique SEM requirements, building a dedicated, proprietary platform from the ground up has proven to be worth the effort. “I don’t mean to disparage outside consultants, or to downplay the difficulty of building your own software,” he says. “It wasn’t easy, and we’re still tinkering with it. But it has made us a clear leader in the automotive space for search operations. And that’s part of the value that we bring to our dealer clients.”