Split Testing Local Search

The best team wins nearly every game and the faster runner wins nearly every race, so are we surprised that the best ad gets nearly every phone call?

For similar sized ads; the standard deviation of the calls received from yellow page ads is greater than the mean.

-Paraphrasing Dr. Dennis Fromholzer, CRM Associates

That quote means that 2 same size yellow page ads for the same type of business are likely to get wildly different call volumes. One of the ads will be just like the better team and the faster runner and win nearly each time the product is searched. Should we expect this to be different on the web? I think not.

Eighty-five percent of respondents agreed that the quality of a business owner’s website is an important factor in earning the consumer’s trust. Over 75 percent of respondents said they were more likely to make a purchase from “an unfamiliar business with a quality website,” than “a poor website from a known business.”

-This is from a Study by Nielsen and WebVisible as reported by Peter Krasilovsky.

The lament of the loser


“I was getting the clicks, but I wasn’t getting the calls” or “Yellow Pages doesn’t work for my type of business.”

I always felt terrible when I heard this and I work hard to inform anyone and everyone around me as to the importance of good content. But the challenge is always the same… how do you define it? What does it look like? The only way to consistently improve the content of any form of advertising that I am aware of is to split test it.

It is has always been much easier to focus on coming up higher in the search results or getting a bigger ad; but there is no doubt the best bang for your buck in advertising will be accomplished with good content.

Local advertisers have been relegated to a place where marketing tools like split-testing just weren’t available to them. After all, do you think the publishers of any Yellow Pages, are going to spend the extra millions to publish and distribute 2 directories so the ads could be split-tested? Not likely. (Side note: This might not be a bad investment for yellow pages, as they would own the copyright on almost every small businesses best yellow page ad and could prevent that ad from being published in a competitor’s book).

The Times They Are A Changin’


I am thrilled that this is not true for the local advertiser on the web even if they depend on phone calls; and can now have this data all reported in one analytics program.

Such a local site can be measured with nearly the precision and ease, previously afforded only to e-commerce sites. It can be done without creating dynamic pages and except for the cost of the phone lines and calls… it is free.

It could also be used to measure conversion for keywords. I have just set up a test that will measure the difference in conversion rates between search phrases that include geo-modifiers and the same terms when the geo-modifier is missing. That’s cool… when you are a local nerd like me.

How To Split Test A Website When Measuring Conversion In Emails and Phone Calls.


What follows is a step by step tutorial for setting up split testing for a website when conversion is measured in contacts through an email contact form and telephone calls. In order to use it, you must have or set-up an account in Google Analytics and have or set up an account with Mongoose Metrics. I have no affiliation with either service.

  1. Have or create 2 hidden pages on your site.
    • One for contact form submissions and one for telephone calls. Only the one for contact form submissions will ever be seen by the customer, this one should say something like “Thank you for contacting us, we will return your message within 24 hours. The other one doesn’t need to say anything as it will not be viewed by the customer.
    • Both need to have the code from Google Analytics. The same code will be on all your pages. Not to worry, we’ll set up filters for that in GA.
  2. Replicate the site twice and put each copy in its own subdirectory
    • On each of the replicated pages use the meta tag “noindex, nofollow” to prevent your content from being indexed multiple times. (Make one copy; add the tags and then make the second one to save you from having to add the tags twice)
    • The name of the subdirectory you choose will be viewable in the url (www.example.com/subdirectory/actualpage.html), so choose something that relates.
  3. Have or create an account in Mongoose Metrics and purchase 2 phone numbers.
    • Consider waiting awhile after purchase to make sure the lines are clean.
    • In your mongoose account, map the tracking number to the actual phone number and configure the tracking to the url of the hidden page of each replicated site.
    • On each of the replicated sites, replace the real phone number with the new trackable number on all pages. Make sure you have one unique phone number per “site”. And that the url’s are mapped to the corresponding phone number in your mongoose account.
    • In effect, you now have 3 sites, each with a unique phone number, and unique confirmation pages. But all on the same domain. We just need to prepare Google Analytics for our first test.

  4. Go to Google Analytics and set up a new profile. Put the radio button on “Add a profile for an existing domain” and use the pull down menu to choose your site.
  5. Name the profile and add a filter to include only traffic that comes to the pages within your first subdirectory. Here’s the set up:
  6. Filter Type: Use Only Traffic From A Subdirectory
    Subdirectory: ^/mysub/ (inside the lines put the name of your subdirectory)

  7. Find the profile for your original site, and create a filter to exclude the traffic from pages in a subdirectory
  8. Filter Type: Custom Filter
    Exclude
    Filter Field: Request URI
    Filter Pattern: ^/mysub/

    These pages in the subdirectory will now be tracked in Google Analytics like they are on their own domain. Which is exactly what we want.

  9. Repeat steps 4, 5 & 6 for the pages in your second subdirectory
  10. We’re now ready to set up the goals in Google Analytics

  11. Find the settings column in your first “test’ profile and click edit to add the goals:
  12. /mysub/emailformthanks.html
    /mysub/ijustgotaphonecall.html

  13. Repeat this step for the 2nd profile.

That’s it… now you’re ready to go. Split test a redesign, keywords, copy… whatever you like. And since both Mongoose Metrics and Google Analytics allow you to tag urls you can use this set up to test conversion on almost any form of internet advertising.

Local websites, typically don’t get enough traffic that these tests can be done overnight. So be patient, over the course of months, I believe they will prove invaluable when making decisions about where to spend money, how much and changing aspects of the website.

None of us are good enough that we are going to develop the very best combination of content on our first try. So, track and enjoy.

This entry was posted in Analytics, Local Search. Bookmark the permalink.

7 Responses to Split Testing Local Search

  1. Pingback: links for 2008-04-05 » eWhisper.net

  2. George Bounacos says:

    Nicely done, and I agree that we need to bump up the level of testing in local SEM.

  3. David Mihm says:

    Tim, I just wanted to make one comment on your YellowPages analogy:

    “that 2 same size yellow page ads for the same type of business are likely to get wildly different call volumes. One of the ads will be just like the better team and the faster runner and win nearly each time the product is searched. Should we expect this to be different on the web? I think not.”

    One other factor to consider is PLACEMENT — my clients’ experience has been that placement within the Yellow Pages (i.e. at the beginning of a particular section vs. middle or end) is even more important than the ad itself.

    And while I think that the web is a bit more democratic than that, since it’s SO easy to do multiple searches and research multiple companies before calling or emailing, you’re still going to really hit the sweet spot with a great-looking website that ALSO ranks well.

    Great post about testing–this works particularly well with PPC; more difficult to do organically, obviously, since it’s tough to hold all factors equal with duplicate content :)

  4. It’s true David… you need to be found/seen before the great ad/website gets it’s chance to shine!! But this also when you need it to come through the most!

  5. Pingback: Mongoose Metrics Web Analytics Blog » Blog Archive » How to A/B Split Test With Tracking Phone Numbers

  6. Pingback: Loci 2008 - Will Scott’s Best of 2008 in Local » Understanding Google Maps & Yahoo Local Search

  7. Rex Dixon says:

    Would you have any of these local tests to share with our community – A/B Tests? If you have the time to do so, we’d love to see an upload of such a test: http://www.abtests.com/ – also – follow us on twitter @abtests