A/B testing

since we all know this is important.

[Roxster]

Because A/B testing is so important, and improving sales conversions is a large goal I think it should be discussed a little more formal.

I recently purchased a scan alert logo, and I do not reconize any sales increase of proportions that they had suggested. However I am not pro/con'ing hacksafe certs. I would like to better test any logo or graphic's ability to slightly change performance.

My current method for measuring a sales/visitor conversion is:
I use my logs in Clicktracks and count all visitors against the target page; for me is the final reciept that the customer prints out. I divide number of sales by number of visitors and multiply by 100 to get a percent. My percent floats between 1.5% to 2% and can be rather seasonal.

My question: Is there a better way to do an A/B test than to measure conversion after a change in the page?

[KeywordROI]

There are a number of services available that will allow you to dynamically change the front end of your site. You can apply these A/B changes based a number of factors to some or all of the site's visitors.

[jsinger]

Our market "research" mostly consists of copying successful, high volume commerce websites. Amazon, Wal-Mart.com etc . Pretty primitive, eh? But I'll bet that 90% of our competitors don't even go that far.

The big sites have the traffic, money and brains to do A/B testing correctly.
--
BTW, when Macys, Penney's, Target and Wal-Mart start posting the Brick/Mortar equivalent of Hacker Safe logos in their stores, I'll try it online.

Do you see any B/M store with a sign: "Our Clerks Won't Steal Your Credit Card Info"?

[paladin]

I was asked by PM to give a few examples. Searching Google for "a/b testing" (withough the quotes) gives a few.

Most are JavaScript based, the one that reads "Non-Intrusive A/B Testing .... A/B tests without IT's involvement...." is DNS based so you do not have to change any of the physical files on your server or database.

Mods - PM me if it is OK to post the URL to that site here

[mreining]

There are definitely split testing tools out there for small businesses. However, be careful with using tools that use "javascript." Yes, javascript is easy to add but it cannot be read by search engines.

So, if the content that you want to test in your A B tests is relevant and important for SEO or PPC (Google now has an ad ranking bot) then I would highly recommend against using a split testing tool that uses javascript.

Following what the big guys do is an interesting strategy. We have so far just used split testing to improve our sites. We find an idea and then test it. It works for us.

[edited by: Woz at 1:52 am (utc) on July 24, 2006]
[edit reason]
[1][edit reason] No promos please, see TOS#13 [/edit] [/edit][/1]

[sniffer]

Id be interested to know: Do you prefer to split test simultaneously, or do you test one page and then change it to another after a period of time?

[zulu_dude]

Wouldn't it be possible to do this sort of testing (i.e. whether a logo makes a difference to order completion) by writing a simple PHP/ASP/whatever script?

You could display the logo 50% of the time and not display it the rest of the time. Then just have an extra variable in your URL so that your analytics software can pick it up.

In this way you would effectively eliminite outside variables such as time of day, day of the week, seasons, etc (assuming you get a decent number of orders each day).