Are you testing your online messages? Your ads? Your post-click experiences?
Without testing, we leave conversions on the table. Because testing
allows you to incrementally improve your response and conversion rate.
Higher, higher and hopefully even higher.
We need to test, we must test, and sometimes, our tests even end up
proving our assumptions wrong. But how do you choose the most
effective method for your business?
You probably hear a lot about multivariate testing and A/B
testing. Both have their place, so you just need to determine
which is right for your business.
Here are a few pointers to get you thinking:
Product: MVT is great for a retail site experience. If you are
selling $20 widgets, then MVT may be a good fit. It will let you test
various elements on a page of your site (such as the checkout
page). If you have a complex, high-ticket sale, A/B might be a
better way to go, because it lets you test post-click experiences, rather than just pages.
Traffic: You need a lot of traffic to really get the most out of
an MVT investment. If you have a high volumne of traffic, and are a
retail site, MVT is starting to look pretty good. However, if your
traffic is lower, perhaps more targeted, and your sale complex, you
should be leaning towards A/B, which lets you test experiences (apples
and oranges even) against each other and get fast results with lower
traffic volume.
Conversion: What is a conversion for you? If it is a check out in
your shopping cart, then MVT might a good way to test the elements of
your shopping cart pages to see what the best mix is to increase sales.
But A/B is just as good for testing conversions (that's the whole
point, right?), especially in the apples to oranges type of tests, or
in the lower traffic/complex lead gen environment.
At the end of the day, regardless of whether you choose MVT or A/B, in
order for your testing to be statistically significant, you need to
have a relatively low margin of error to draw conclusions. If you
are driving minimal traffic, using A/B will yield results, which you
can then build on for future tests. And tests that yeild results will yield more conversions, whatever they may be.
-Susan Delz
