Tuning AdWords, Part 3: Ad Variations

23 Jun, 2008  |  Written by  |  under AdWords

One of the things I love about AdWords is that it tends to give me information I need to make good choices.  I just looked at an ad campaign a client originally set up a year ago, and it’s done pretty badly for them over time.  They seem to have good keywords, and the landing page was chosen reasonably well, but they’re still not getting many conversions.

The culprit?  Bad ad text.

These guys have just one ad variation set up, so it’s tough to say the ad text is definitely at fault, but it’s my first guess.  Luckily, AdWords makes it easy for us to test ad text performance, using “split testing”.  The idea with split testing is that we write two very different ads, we show them each 50% of the time, and whichever gets more clicks is better.  (In reality, tracking conversions is better than clicks, but we’ll keep it simple for now.)

AdWords also has a system where it will rotate the ads for you, figure out which gets more clicks, and run that one more frequently.  This is good for “fire and forget” advertisers who want to set it up and walk away.  If you’re spending $50/month, this is fine.  If you’re spending $500/month or more, it’s worth doing your own testing.

In either case, it makes sense to create perhaps 8 ad variations before launching a new ad group.  Here are some ways I vary ads, to see what resonates well with searchers:

  • Ask a question
  • Make a promise
  • Challenge an assumption
  • Vary the headline
  • Mix the headline and body rows up
  • Change the URL (with/without “www.”, change capitalization, etc.)
  • Get specific:  Give your price, timeline, etc.
  • Be personal.  Start your introduction process.

Once you’ve set up some variations, you’re ready to test.  If you’re a $50/month “fire and forget” kind of advertiser, head over to ‘Edit Campaign Settings’ and pick “Optimize: Show better-performing ads more often”.  Now put a reminder in your calendar software to come back every month to review AdWords.  Kill the ads that Google has stopped showing, and create variations around the ads that are working well.

If you’re a $500/month advertiser, you have a harder, but more interesting, road ahead.  Set up your ad variations, and pause all but two of them.  Now head to ‘Edit Campaign Settings’ and pick ‘Rotate: Show ads more evenly’.  Google will now show just those two ads, each 50% of the time, while the rest of your variations wait their turn.

Our goal is to get perhaps 30 clicks so we can figure out what’s performing well.  Guess how long it’ll take you to get that many clicks, and set a reminder in your calendar software to come back in a day or a week or whatever.  (I can’t tell you how many ad campaigns I’ve seen where people meant to come back and forgot, only remembering when they got the credit card statement and realized they’d spent WAY more with AdWords than they meant to, on poorly-converting clicks!)

Once you’ve got your 30 clicks, you probably have a clear winner from the first A/B.  Whoever lost, go ahead and delete.  Whoever won is your new “control” ad, the one against others will be measured.

This is important: The “control” ad now has a history with Google, and will be shown higher than a new ad, so you have to make a copy of this ad for the next round of testing.  Pause the “control” ad, make an exact copy, and un-pause one of your other variations.

Now just come back at intervals when you have about 30 clicks, deleting losing ads (and their old copies, if they were a control ad), pausing control ads, making new copies of control ads, and bringing new variations online.  If you’re making money with your ads, great!  You may hate the 50/50 split approach, but you can get more advanced with this and create multiple copies of the control ad so your test is only getting 25% of impressions or whatever.

As you iterate through this process, you’ll probably come up with new ad variations to add to the pool, and you may want to continue iterating likc this forever for highly-converting ads you’re spending a lot on.  If you do really well, you may notice other advertisers stealing your copy.  Sorry, but remember that’s just flattery!

One Response so far | Have Your Say!

  1. Cliff Bernstein  |  July 1st, 2008 at 1:04 pm #

    Hey, Glenn, nice write up. I’ve been aware of the A/B split testing strategy, but hadn’t thought about the fact that in the next cycle, Google will initially give more weight to the ad that just ran. Nice strategy for dealing with that.

    Yeah, it may be flattery with a competitor “steals” someone’s ad copy. However, it can also be lost potential profits if the competitor’s ad copy get better. But, hey, that’s the nature of the beast…

    Cliff Bernstein - Gravatar

Leave a Reply