How long do you test for?

by mguy
5 replies
Say you are selling a book and you get organic traffic. You get some sales but you think you can do better.

How long do you test things for? For one, could change the one cover page for a month then change and compare to the next month? Is that reasonable?
#long
  • Profile picture of the author lukeblower
    Why not put up two or three covers and see which one does best. If you do one at a time it's going to take ages.
    {{ DiscussionBoard.errors[9062582].message }}
  • Profile picture of the author Luke Dennison
    I would say one week each. Or if your traffic is intermittent, veiw results after every 1000 visitors or so.
    {{ DiscussionBoard.errors[9062589].message }}
  • Profile picture of the author TheCodex
    Don't test with organic traffic, test with paid traffic...

    As for the amount of clicks? It's up to you, but there needs to be a statistically significant amount of clicks to account for variance. You'd also be wise to day-part your traffic so you can gain a better understanding of your audience's buying patterns.
    Signature
    Internet Codex - Redefining Internet Marketing
    {{ DiscussionBoard.errors[9065745].message }}
    • Profile picture of the author JohnMcCabe
      Originally Posted by TheCodex View Post

      Don't test with organic traffic, test with paid traffic...

      As for the amount of clicks? It's up to you, but there needs to be a statistically significant amount of clicks to account for variance. You'd also be wise to day-part your traffic so you can gain a better understanding of your audience's buying patterns.
      This is the ticket. Running tests like you (the OP) propose won't likely yield anything useful. Too many variables that can change from month to month.

      Take Kindle sales, since you want to test an ebook.

      If you ran your first test in January, and the second in February, it wouldn't be accurate. In January, people with new reading devices are actively hunting for content to fill up their shiny new Kindle or tablet. In February, this goes away.

      You need to learn how to do A/B split testing. Or even A/B/C (or however many letters you have the patience for). Then you just run your test until you have significant results. Be prepared for some of those results to show that there is no significant difference between two alternatives.
      {{ DiscussionBoard.errors[9067427].message }}
  • Profile picture of the author Jtraits
    I'd say 5-7 days. But also, set some criteria. If the results pass the criteria you've set up -preferring criteria before you start and don't change them- , go on
    {{ DiscussionBoard.errors[9069361].message }}

Trending Topics