11 {{ upvoteCount | shortNum }}
11 {{ upvoteCount | shortNum }}

A/B Testing - How Much Traffic Needed?

by TSquared
9 replies
I have only had a site up for a month and a half. Currently ranked on page 2 for my primary keyword. This is my first site. Made my first eBook sale after 227 unique visitors. I am now getting about 20 unique visitors a day.

How much traffic do I need before I start A/B testing on such a young site so I can start increasing the conversion rate? Thanks for any help.
#a or b #needed #traffic
  • Profile picture of the author JasonParker
    You should start now, but only test something big like your headline or the look and feel of your page. You can use google optimizer and it'll declare the winner over time. You may need 1000 unique visitors or less before google optimizer declares the winner.

    If you're going to use another split testing software and declare the winner yourself then there is not a real set rule of thumb. If you have one headline pulling away 3 to 1 after 500 unique visitors or so, then you can possibly assume that's the winner. But if it's only beating it like 2 to 1 then the other headline might come back and surprise you and win. In that case I'd say go about 1000 unique visitors until you determine the winner.

    It's best to let google optimizer just do it's thing and tell you the winner... After all, that software used to cost like 6 grand and now it's free. Its pretty powerful.
    {{ DiscussionBoard.errors[3758027].message }}
  • Profile picture of the author jasondinner
    Yeah, I second Jason's recommendation for Google Optimizer.

    It's real easy to set up and they let you do simple A/B splits or Multivariate testing.

    Unless your pages are converting very close to each other, it usually doesn't take very long to find a winner.

    I've had tests running where 10,000 impressions in, Google still couldn't determine a winner because they were converting just a few decimal places apart from each other.

    If it gets to that point with your test, try another variation to test as that one won't make much of a difference.

    - Jason
    Signature

    "Human thoughts have the tendency to transform themselves into their physical equivalent." Earl Nightingale

    Super Affiliates Hang Out Here

    {{ DiscussionBoard.errors[3758404].message }}
  • Profile picture of the author Paul Hancox
    With 20 visitors a day, you'll be there for MONTHS trying to get significant results for one test.

    Instead, why not measure and test some other kind of action, such as an opt-in rate?

    If you're getting 20 visitors a day, with a 10% opt-in rate, that's 2 subscribers a day, or 60 a month.

    If you can use split testing to raise that to 20%, you can get 4 subscribers a day, or 120 a month.

    You then have an opportunity to convert TWICE as many people into customers.
    Signature
    PresellContent.com - How to sell without "selling"
    {{ DiscussionBoard.errors[3758550].message }}
    • Profile picture of the author TSquared
      Originally Posted by Paul Hancox View Post

      With 20 visitors a day, you'll be there for MONTHS trying to get significant results for one test.

      Instead, why not measure and test some other kind of action, such as an opt-in rate?
      For this first site, it is only a $10.00 eBook and I didn't go with any opt-in's. However I get what you are saying and maybe can set up a two step funnel and test clicks or something of some more frequency than a sale.
      {{ DiscussionBoard.errors[3759024].message }}
  • Profile picture of the author Mirnova
    Always be testing...ummmm didn't 1 or 2 gurus say that ;-)

    Start now. But don't bother with multi-variant tests. Stick with A/B tests. And test the main stuff.... headline, offer, price, guarantees. Test the minor stuff after you're getting traffic.

    A trick is to breakup your sales process. Don't test just for the sale, instead test for the next action along your sales process. That way you can refine your sales process far faster then simply testing for outright sales.

    I know it's not perfect. But if done logically you'll make far more progress with far less traffic in far less time.
    {{ DiscussionBoard.errors[3759796].message }}
  • Profile picture of the author Stephen Dean
    Gotta agree with Paul, big time.

    You should always be testing, you don't have to wait to start.

    And it's faster to test for an action that's easier to come by then the sale. I recommend squeeze pages, like Paul suggested. It'll help you discover which headlines and appeals work best - so you can then take that feedback and tweak your sales page to match.

    That's the way to do it when your traffic is low. But you should never wait to get started

    Most testing software has an algorithm that will tell you when to stop, or should. My software is Free Split Test, free save for an opt-in. It uses Chi Square with a Yates variation.

    When you use an algorithm, you don't need to have any set rule of number of visitors or number of sales.

    Cheers,
    Stephen Dean
    Signature
    Free Coaching WSO: How to finish all your 2013 "Goals" in JANUARY with my proven productivity secrets - taken from 9 years working as a freelance copywriter. Click Here

    Occupation: Best Copywriter Ever.
    Clients:     Matt Bacak, Jim Edwards, Ryan Deiss and more.
    {{ DiscussionBoard.errors[3759846].message }}
    • Profile picture of the author jasondinner
      Originally Posted by Stephen Dean View Post

      Gotta agree with Paul, big time.

      You should always be testing, you don't have to wait to start.

      And it's faster to test for an action that's easier to come by then the sale. I recommend squeeze pages, like Paul suggested. It'll help you discover which headlines and appeals work best - so you can then take that feedback and tweak your sales page to match.

      That's the way to do it when your traffic is low. But you should never wait to get started

      Most testing software has an algorithm that will tell you when to stop, or should. My software is Free Split Test, free save for an opt-in. It uses Chi Square with a Yates variation.

      When you use an algorithm, you don't need to have any set rule of number of visitors or number of sales.

      Cheers,
      Stephen Dean
      Very true. If all things remain the same with your sales conversion rate (which I'm sure can be improved on) if you increase your opt-in rate, you sales should increase as a by-product of your increased opt-in rate.
      Signature

      "Human thoughts have the tendency to transform themselves into their physical equivalent." Earl Nightingale

      Super Affiliates Hang Out Here

      {{ DiscussionBoard.errors[3760751].message }}
      • Profile picture of the author Summertime Dress
        I use a WP theme split test plugin (SES THEME split test):

        http://www.leewillis.co.uk/how-to-split-test-wordpress-themes/

        It easily integrates with your Google Analytics and you'd be surprised how something as simple as testing different background colors can have a significant impact on bounce rate...even with low unique visitor count of 20 per day, you can easily test 100 visitors per theme in a week.
        {{ DiscussionBoard.errors[3761234].message }}
  • Profile picture of the author TSquared
    Thanks for all the advice. As a quick update, I tweaked the title only and now have 3 sales with 284 uniques. I will take the advice and start testing immediately.
    {{ DiscussionBoard.errors[3769778].message }}

Trending Topics