March 23, 2004

SEM case study: SoftwareCEO's results from search engine ads, how we did it, and what we have left to learn

by Bruce Hadley, SoftwareCEO

About a year ago, we started investing some money in search engine marketing (SEM) through Overture and Google AdWords.

Like many entrepreneurs, we couldn't resist the combination of results-oriented marketing—you only pay when someone clicks on your ad—and an easily controllable budget (read: cheap). Heck, we started out at Overture with a monthly cap of $100, and back then they were giving new advertisers a $100 credit just for signing up.

As the old joke goes, that seemed like such a good price that we couldn't afford not to buy.

But, our do-it-yourself approach yielded mixed results—which, we suspect, is the experience shared by most.

Our rookie year
For the first half of 2003, we muddled along. We continued to budget about $100 per month on each service, and we tried to pay attention to the keywords and phrases that were attracting clicks. We let the inactive ones die, and we occasionally got into bidding wars for the ones that appeared to be hot.

Roughly midway through 2003, we gave up on Overture entirely, for two reasons: Their clumsy user interface made our account there unreasonably difficult to manage. However, to be fair, remember that this criticism comes from the perspective of amateurs; we had neither the time nor the inclination to become Overture experts.

Second, we knew from the analysis of our Web traffic that the vast majority—about 85 percent—of our visitors was finding us via Google.

So, for the second half of 2003, we increased our Google AdWords budget to $200 per month, and tried to focus our spare energies there.

With our very limited expertise and a list of about 100 keywords and phrases, we did OK. Not great, but OK. In November, for example, we got 1,398 clicks on our Google ads, and a click-through rate (CTR) of 0.38 percent. In December we got 1,258 clicks and a 0.29 percent CTR.

Then we decided to get serious. We sought professional help.

We called in the professionals
We did Google searches on "SEM" and "search engine marketing," and JumpFly, a firm from Carpentersville, Ill., kept popping up near the top. We figured that if they knew how to get that kind of result for themselves, they probably knew how to do it for their clients.

Plus, we liked their pricing: $1,695 for setup and monthly management of our AdWords and Overture accounts for six months; if we decided to continue, it would cost us $100 per month per account to have them manage everything.

(You don't have to sign up for a six-month project like we did; JumpFly will set you up for $495 and manage both AdWords and Overture $200 per month, with no commitment beyond the first month.)

And, we like their attitude: We told JumpFly partner and founder Brad Goldstein that we were going to report on our results, good or bad, in our newsletter—meaning, if we weren't impressed, there'd be about 50,000 readers who'd be similarly unimpressed. Goldstein didn't flinch.

OK, fast forward for a moment: JumpFly overhauled our AdWords and Overture accounts and took them live at the start of 2004. In January, we got 4,207 clicks from AdWords alone, and a CTR of 1.30 percent. In February, we got 8,055 clicks and a CTR of 1.52 percent.

So, yes, we'd say it's working: Our quantity of clicks tripled, and our CTR nearly quadrupled, while the spending limits at AdWords and Overture remained the same.

The initial evaluation
All right, how did JumpFly pull this off? What had we done so wrong? What did they change? What's next? And how can you mimic this success for your software company's site?

Founded in 1998, JumpFly now has six employees and focuses exclusively on SEM; we started talking with Goldstein in December 2003. He looked at SoftwareCEO, listened to what we were trying to accomplish, and reviewed the accounts we'd set up at Google AdWords and Overture.

"I saw a Google account set up the way most people would set them up," he says. "It wasn't poor, it just wasn't based on years of experience. Whenever we take over client-created accounts, we can almost always get them three to four times more traffic for the same budget."

Making a match with keywords
Our biggest failing, in our do-it-yourself phase, was a total lack of regard for the type of keyword and phrase matches we'd pay for each time someone clicked on our ads.

You can select four different types of SEM matches, Goldstein explains:

#1. Exact match. With this option, if we pay for the phrase "software business," someone who searches for "info on software business" will never see our ad.

#2. Phrase match. If, again, our desired phrase is "software business," our ad will show up when someone searches on "info on software business" or "software business tools," but will not show up when someone keys in "business software."

#3. Broad match. As the name implies, this will show our "software business" ad to anyone who searches on any combination of those two words, in any order.

"Google offers broad match, Overture does not," Goldstein says, "except for their secondary partners—Alta Vista and HotBot, for example, but not for Yahoo and MSN.

"At Overture, because they don't offer broad match, you need to pick and choose each term; you might be the only bidder on a specific term. Over at Google, on the other hand, you might find you've got lots of competition for the same term because lots of people are using broad matching.

"For example, one of our clients is NCvacations.com; they handle vacation rentals in North Carolina. At Google, they were competing with everyone who broad matched on "North Carolina rentals," while at Overture, there were lots of specific terms that no one else was bidding on.

"Take the phrase 'North Carolina waterfront vacation rental': at Google I'd be buried with 20 other people who're broad matched on ' North Carolina rental' or ' North Carolina vacation,' but at Overture I'm all alone, or maybe competing with one other guy. I'm paying a dime at Overture, but I'm the top search result for that specific, very targeted phrase.

#4. Negative match. This has the effect of suppressing your ad for the terms you select. It isn't used often, and JumpFly didn't use it for our site, but it can come in handy when you know there will be lots of searches on something similar to your phrase, and you want to filter them out.

"Negative match can be relevant for some types of businesses," Goldstein says. "For example, one of our clients is Paul Morad, a realtor in Valencia, California, who serves the Santa Clarita Valley.

"For him, we created negative matches for 'Santa Clara University,' ' Santa Clara school,' and anything having to do with the college, because we knew there were thousands of searches from students. If Paul Morad's ad showed up on all those searches, it would kill the CTR for his term."

At SoftwareCEO, before hiring JumpFly, we had all our keywords and phrases set for broad match, just because we didn't know any better.

Now, that is not to say that one type of match is preferred; we're only suggesting that you need to display a little more intelligence about choosing the type of match than we did.

The best match type for you will depend entirely on the kind of results you're trying to achieve, and how well your ad speaks to the term that the user just searched on.

"With 'software business' set for broad match, someone looking for 'business software' would see your ad," Goldstein says. "If you put it in quotes, thus setting it for phrase match, it won't show up, which means it improves you click through rates—and at Google that also improves your positioning.

"Google is unique in that they take into account your click through rate as well as your bid amount. Even if you've bid less than your competition, you can get a higher showing.

"In fact, I suspect that's why Google doesn't show the prices on keywords: They don't want their support reps to go crazy with calls from people wondering why their ads aren't showing even though they've put high bids on them."

Mixing and matching for maximum clicks
For most of its clients—including SoftwareCEO—JumpFly starts out with an equal mix of broad and exact match terms; in other words, they put "software business" in our list of keywords twice, once as a broad match term and once as an exact match term.

"We set up most everything up as broad and exact," Goldstein says. "That's our strategy, and it's straight from Google's mouth. What happens is that the click rate for the exact terms goes up, and that helps your showing with the broad terms."

For example, as we go to press, our CTR for the phrase "software services" on exact match is 1.4 percent; on broad match it's only 0.7 percent. "If we'd only set it up as broad, then your overall click rate would be somewhere in the middle," Goldstein says.

At the same time, some terms were set for phrase and exact match only. For example, Goldstein didn't go for broad match on "software forums," "because you don't want someone who's looking for forums software."

Even though we're pleased with the traffic, Goldstein says SoftwareCEO's click-through rates are a little lower than what he sees with his other clients. "The normal CTR is 2.0 to 4.0; yours is a little harder, because it's broader market. You don't have an exact, easily describable product that you're selling."

As we go to press, our CTR on Google is holding steady at 1.5. At Overture, where we use very narrowly-defined keyword phrases (and far fewer of them), our average CTR is 0.93.

Cranking up the volume
In addition to tweaking the accuracy of our matches, JumpFly greatly increased the sheer size of our list of keywords and phrases.

We had just shy of 100 in total before bringing JumpFly onboard. We now have 460 unique terms at Google, but since JumpFly runs different match types on many, our total list is 920. At Overture we have about 300.

"We went through every page on the site, then brainstormed with people here to figure out what might work," Goldstein says.

"Then we used Overture's search suggestion tool to expand on that; it's a great tool to find out what people are actually looking for. It doesn't give you Google results, but it gives you Yahoo and MSN, and it gives you a pretty good idea of what's going on on the Web."

By the way, anyone with an Overture account can access the search suggestion tool from the "Manage Products" page; but if you'd like a quick search, here's a direct link.

What's the right number of keywords? As many as possible, Goldstein says. "If you choose just the 20 most relevant terms, those are also the 20 most popular, which means they'll be the most expensive, because your competitors are also bidding on them."

If you focus only on the obvious and the "cream," you'll get a ton of fat and not much protein, Goldstein says. At the same time, it doesn't make sense to choose terms so obscure that it will never occur to anyone to search for them. JumpFly's general rule of thumb: Don't choose terms or phrases that generate fewer than 50 searches a month.

To every such rule, there's an exception, however. "If you're a realtor or lawyer or doctor, for example, we'll slap in 60 terms specific to cities, even though they might get a lot less than 50 searches on each," Goldstein says.

"These are incredibly valuable if someone does look for them. For example, for a client who's a personal injury attorney in Florida , we've got lots and lots of terms that consist of the name of a local city in Florida plus the name of a particular injury.

"If someone does look for something that specific—say, 'Orlando lawyer whiplash lawyer'—they are a valuable lead, and a lead is worth potentially tens of thousand of dollars."

Writing ads that pull in the right clicks
Before we hired the professionals, we were smart enough to create SEM ads that were reasonably specific to the terms being searched on, but that only scratches the surface.

We hope you know by now that it's usually a mistake to send someone directly from an SEM ad click to your home page. If they've searched on "dental practice management software" and clicked on your ad, don't take them to your generic home page where you tell them about your company. Instead, take them directly to the page that spells out the benefits (not features) of your dental practice management software.

OK, we'd done that to some extent; we created a different ad for each of our nine ad groups. Depending on the user's search, we took them to one of our nine Topics pages, which at least showed we were paying attention their subject area interest. For example, if you keyed in "software programmer salary," we took you to our Human Resources page rather than our home page.

But that's all the further we went, and that wasn't nearly far enough.

JumpFly created multiple ads for each ad group. "We changed the titles for your ads, and rotated them to see which pulled the most clicks," Goldstein says. "Where you had a headline of 'SoftwareCEO,' I found the CTR with different headlines was two to three times better.

"But the ad that works best for you, and works for many of our clients, is a Google rotation. You put a special piece of code inside curly brackets, and it will automatically insert the word or phrase they're searching for, as long as it's no more than 25 characters; if it's over 25 characters, it goes to a default title that you've set up."

Thus, someone searching on "software programmer salary" would be presented with our ad with the headline, "Software Programmer Salary." Unfortunately, Google doesn't tell you how to do this. "That bit of code in the curly brackets is something they only share with their big clients," Goldstein says.

"We meet with a rep from Google at least once a year. They'll give us the tips and tricks they're using with their big clients, the ones that are spending hundreds of thousands of dollars a year, and we apply those to our small clients."

The ad headline trick Goldstein refers to is indeed undocumented by Google, and we don't want to give away all of JumpFly's secrets, but we're willing to share this little jewel with our subscribers. Write your AdWords headline as follows:   {Keyword: Default Headline}

...where "Default Headline" is the headline you want to appear if the user's search phrase is longer than 25 characters. If his/her search phrase is 25 characters or less, then the phrase itself will appear as the headline. Note: You must include the curly brackets.

More SEM ad tips
Other than the maxim that your ad headline should match the user's search terms, Goldstein offers the following tips for Google and Overture ads that pull clicks:

"As with any ad, you have a limited amount of space," he says. "Pull the big impact words in early. If you have something free, something guaranteed, stick that in the ad."

Keep in mind, however, that you can only use the word "free" if the Web page that you're sending people to also says "free." As much as they can, Google and Overture are trying to prevent bait-and-switch tactics.

But, the guidelines for all SEM ads are a little vague, Goldstein says. "What gets through depends on their review process, and the review depends on which human at Google or Overture is reviewing your ad that day. One time it might get through, the next three, not."

Google doesn't permit exclamations or superlatives in ads, such as "best," "most," "greatest, "#1!," and so on. Comparative terms, however—e.g., "great"—are allowed.

Once again, enforcement is sometimes sketchy: "I did use 'most trusted' the other day, and got it through," Goldstein says. "Basically, if you get caught and they reject it, you just change it and resubmit it."

Other suggestions: "Capitalize the first letter of every word; Google says it improves CTRs, and we see in our own research that it seems to hold true," Goldstein says. "You can't control the colors and bolding; the words your prospect is looking for will show up as bold, which is another reason that the curly brackets trick works well."

Bidding, budgeting, and finding the right balance
Keep in mind, it's the combination of the right search term plus the right bid amount plus the right ad that brings in customers. So, Goldstein says, it may in some cases pay off to set your bid very high if you can use your ad to weed out the non-buyers.

"For example, the bid for the term 'car dealer' is $1.10 right now at Overture," he says. "That means you could be #1 for $1.11. Now, Overture may or may not approve it, but if your ad headline is 'Minnesota Car Dealer,' people outside Minnesota aren't likely to click.

"As long as you're willing to pay for the position, Overture will keep you alive. Only if your clicks are really, really low for a long period of time will they take your ad down. I've had ads up for a year at Overture that have gotten a warning for low click index, but they're still up there."

Bottom line: Try everything. Keep what works, and throw out what doesn't. One of the beauties of SEM is that you can see results in real time, and can make changes very quickly. "If you have four or five different ads, rotate them all, and see which ones work," says Goldstein.

As we said at the outset, one of the attractions of SEM is that your spending level is always under your control, and the notion of worldwide exposure and targeted clicks is very attractive to those of us with shallow marketing budgets.

But SEM budgets are more complicated than your monthly allocation, because you're looking for the magic combination of exposure (impressions) and prospects (who click on your ad) and customers (who convert to sales). Even if you're only spending $100 a month, if you're getting no sales from the exercise, that's $100 too much.

"Budgeting for SEM puts a lot of people in their graves by doing it incorrectly," Goldstein says. "For example, we now have your AdWords budget set at $30 per day. It might go higher per day, but it won't be more than $900 a month; Google promises that."

Here's the hard part: The budget ceiling we gave JumpFly to use at Google AdWords was $400 per month; so why is he so glibly willing to break our budget by $500? Answer: He isn't.

"If you want your ad to show all the time, you have to make sure your budget guideline is considerably over what you're really going to be spending," Goldstein says. "We're monitoring it for you, and if we see you're approaching the $400 limit you set, we lower your bids.

"We have software that checks your account every hour, 24 hours a day. If someone was bidding 17 cents, we'd take you to 18, if they then went to 20, we'd take to the next best position. Or, if your budget was maxed out, we might lower it.

"Our software will see who's bidding what on all terms, and carry out our specifications, whether it's for a dollar amount or for a particular term.

"When we started, we had your bids at 10 cents, and we were spending 7 cents, so we lowered your bid to 9 cents, and you're now actually spending 6 cents. We did that in order to stay within your budget; lowering the bids decreased your traffic a little bit, but you're still getting a strong CTR."

In other words: As long as you monitor your activity closely—very closely—bidding higher than you're willing to spend will have the effect of building traffic; once the volume is there, you can lower your bids to stay within your actual budget.

Most amateurs will start out with a few keywords and terms, bid high on those terms, and set a tight daily budget because they're nervous that things might get out of control—but that's exactly the wrong approach, Goldstein says.

"You can put your spending limit at $15 a day and set your maximum bid at 50 cents, and you'll get 30 clicks," he says. "But then Google will show your ad intermittently throughout the day, and you'll wonder why you never see it.

"We think it's better to start with a bid of 5 cents, across lots of terms, to see how much traffic we can build, and find out where it's coming from. Then we'll increase the bids to match the budget.

"If we increase your bids, you'll get more traffic, but you'd be paying more per click, which is going to increase your cost per lead—is it still worth it to you? That's what it ultimately comes down to: What's a lead worth to you?

"I don't know the answer to that one, of course, but you want to find that peak point. You're the one who knows what kind of sales you're getting. If you tell me you're getting great results, and all your leads are coming from this, then we know we should continue to increase what you're spending on SEM."

Tracking clicks to purchase orders
Unfortunately, this is where the SoftwareCEO case study falls apart; we'll have to leave you somewhat in suspense until we complete the second part of what every serious SEM campaign demands: conversion tracking.

You see, we know we're getting far more site traffic, thanks to SEM, and we know that in general we're getting more conversions; in our case, that means paid memberships and seminar registrations. However, we can't tell you exactly what percentage of the increase is directly attributable to SEM.

Partly, we have an excuse: Because a third-party ASP system handles all our registrations, we are currently unable to track the conversion from a click on our ad at Google or Overture.

This is frustrating for us, because both Google and Overture offer free conversion tracking; all you have to do is put their HTML code on a page on your site that users see after they complete your desired action. It could be a trial version download or a software purchase; as long as it's a unique page with that code—a "thank you" page, for example—the conversion will be reflected back in your SEM reports.

In our case, because we don't control the registration and purchase process, and because none of those pages live on our server, we can't complete the circuit—at least not yet. We're working on it, and we do have our conversion tracking in place, we'll let you know in a future report exactly how well our SEM effort does in terms of sales.

Interestingly, Google and Overture both introduced regional tracking not long ago. "It may not matter to most software companies, but for some of our clients it's great," Goldstein says. "You can look at conversion rates by geographic area, or you can show just the U.S. and Canada if you don't want to deal with foreign inquiries."

The differences between AdWords and Overture
Do you need both Google and Overture, we asked? Is one better than the other? "It 100 percent depends on the client as to which is better," Goldstein says.

The most obvious difference between Google and Overture is the size of your ads. At Google, you're limited to just three lines: A 25-character headline and 35 characters on each of the next two lines. (All character counts include spaces.)

At Overture, you get to run a 40-character title and a 190-character description. "With that much space, you can say a lot about your company and really target your leads before they click," Goldstein says.

Another point in Overture's favor, which ties into the ad length: Overture tends to be more forgiving about low CTRs over a long period of time, says Goldstein.

"At Overture, you might have bid on a really broad term describing your business, but thanks to your very specific, 190-character ad description, you're getting a low CTR. That's OK, as long as you know you're getting clicked by qualified leads. At Google they'll pull your ad for a low CTR."

You'll also see pretty significant keyword price differences between Google and Overture, but cheaper doesn't necessarily mean better.

"In SoftwareCEO's case, the price variable is fairly pronounced," Goldstein says. "You're getting tons of traffic from Google at 7 cents, where at Overture it's a dime. I think you still have to have both, but most of our clients aren't getting the traffic you're getting.

"Most of our clients don't get leads as affordably as you; your site is an exception to the norm, because there are just not a lot of people who do what you do."