Conversion Rate Testing - How much testing is enough to consider a conversion rate PROVEN?"

9 replies
Looking for some advice.

If you were testing conversion rates...to sell rights to a site, product, high ROI cpa campaign etc, by driving traffic to a website, product, cpa offer, or whatever, in order to PROVE tested conversion rates, what is a good amount of targeted traffic to have tested?

Say I split tested a few times, a landing page and CPA offer. Or a website and Clickbank product. After driving 100 visitors, I'm converting at 6% (example)

I'm assuming a 6% CR on only 100 visitors isn't quite enough to claim "PROVEN" conversion rates for targeted traffic.

So my question is, in YOUR opinion, what would you consider a PROVEN conversion rate? After testing on 1000 visitors, 10,000 visitors?
  • Profile picture of the author Matt Morgan
    Originally Posted by friend View Post

    Looking for some advice.

    If you were testing conversion rates...to sell rights to a site, product, high ROI cpa campaign etc, by driving traffic to a website, product, cpa offer, or whatever, in order to PROVE tested conversion rates, what is a good amount of targeted traffic to have tested?

    Say I split tested a few times, a landing page and CPA offer. Or a website and Clickbank product. After driving 100 visitors, I'm converting at 6% (example)

    I'm assuming a 6% CR on only 100 visitors isn't quite enough to claim "PROVEN" conversion rates for targeted traffic.

    So my question is, in YOUR opinion, what would you consider a PROVEN conversion rate? After testing on 1000 visitors, 10,000 visitors?
    Firstly Congratulations on the 6% that is a good conversion rate.

    I say the more the more better, the more accurate the average. 1000-2000 perhaps would be a good test range.

    But to be honest there are no maximum limit, you can even do 20,000 for more accuracy, so there is not really a 'correct maximum answer'. The sky is your limit, the more you do the more accurate.
    Signature
    -> [FREE WSO 1] Discover 77 FREE Ways To Generate Traffic!
    -> [FREE WSO 2] Discover 67 Killer Traffic Methods Which Will Crush Your Competition!
    -> [FREE WSO 3] Discover 77 Amazing Blogging Tips To Explode Your Profits!

    ...…..Now LISTEN CAREFULLY! ===> [WSO REPORTS 4, 5, 6 are >> Found Here]
    {{ DiscussionBoard.errors[2440208].message }}
  • Profile picture of the author tylerdrun
    1000-2000 is not enough either to tell ya the truth. You do need 10k at the last. But do 5k. I think that'll give you estimates enough to make decisions.

    The law of large numbers sometimes comes into play and increases or decreases the conversion rate significantly. If you've been converting at 1:250 for 1000 visitors. The site could convert at 1:100 after sending 10,000 - 20,000 visitors.

    If you are preselling, you will have higher conversions with the law of large numbers. If you aren't preselling, your conversion rates will dip.
    {{ DiscussionBoard.errors[2440281].message }}
  • Profile picture of the author xiaophil
    Originally Posted by friend View Post

    So my question is, in YOUR opinion, what would you consider a PROVEN conversion rate?
    I don't think a statistic can really be 'proven' but if I understand your intention then you may want to look into "confidence intervals".

    While everyone may have their opinions and rules-of-thumb, you can actually determine your confidence intervals scientifically.

    Using your example of 6 conversions from 100 visitors (an apparent 6% CR) I would say the actual conversion rate could be anywhere between about 1.2% and 10.8%, with a confidence of 95%.

    For accurate measurement, the more conversions the better. For example if you had 60 conversions from 1000 visitors (still an apparent 6% CR) I would say the 95% confidence interval would be between about 4.5% and 7.5%.

    Bear in mind that all this assumes a uniform sample - ie all your traffic is the same, which obviously isn't true, although if you are tracking conversions based on something specific like a keyword from a certain source then that's approaching a fair approximation I think.

    Analytics really is a science. While there's plenty of tools available for collecting data I see very few capable of doing much with it. A case in point is split-testers - it's laughable to see some of these things basing decisions on raw counts of a handful of conversions.

    This stuff is often counter-intuitive too. I think it's worthwhile for anyone serious about testing to get some basic stats skills under their belt. It will pay dividends I am sure.

    Phil
    {{ DiscussionBoard.errors[2440685].message }}
  • Profile picture of the author Neil Morgan
    While there's plenty of tools available for collecting data I see very few capable of doing much with it.
    Therein lies a good idea for an ebook that would sell in bucket loads. Or it should.

    Cheers,

    Neil
    Signature

    Easy email marketing automation without moving your lists.

    {{ DiscussionBoard.errors[2440792].message }}
    • Profile picture of the author Rich Blondi
      Go here, put in your numbers, and it will tell your confidence rate. Anything upwards 95% would be generally considered a valid test from a statistical standpoint.

      :::SplitTester.com:::
      {{ DiscussionBoard.errors[2441013].message }}
  • Profile picture of the author peterdavid
    In Average 5% of conversion ratio is good. For very targeted campaign you can get 10% plus.
    {{ DiscussionBoard.errors[2441036].message }}
  • Profile picture of the author DogScout
    :::SplitTester.com::: Is a great resource

    the testing itself should never stop. ALWAYS continue to test. Even if you get to a conversion YOU are satisfied with, the market can (and will) change. Ipso facto, so will what converts.
    {{ DiscussionBoard.errors[2441041].message }}
    • Profile picture of the author xiaophil
      Splittester.com looked interesting so I entered some values:

      Ad 1: 100 clicks 1 conversion (1%).
      Ad 2: 100 clicks 2 conversions (2%).

      Result: You are approximately 99% confident that the ads will have different long term response rates.

      I'm not knocking the tool but this result doesn't sit well with me. Am I the only one? I wonder how many people have dropped ads based on results like this.

      I still suggest learning enough to at least know what to realistically expect from tools and how to drive them instead of just plugging numbers in and blindly accepting the results without a sanity check.

      Phil
      {{ DiscussionBoard.errors[2441408].message }}
  • Profile picture of the author davemiz
    lol.... some of these answers/comments are so wrong its not even funny.

    you're giving bad advice.

    REALLY bad advice thats not true.

    theres this free software.... its called google website optimizer..... maybe you've heard of it?

    it will tell you when the test is finished. (re-read that again)

    sometimes it takes a few hundred clicks to finish a test, sometimes it takes thousands.... it ALL DEPENDS ON THE TEST.

    sometimes the stuff you're testing is very close and it needs a LOT of visitors to get the winner.... sometimes you get a runaway winner and the test is over quickly.
    Signature

    “Judge your success by what you had to give up in order to get it.”
    ― Dalai Lama XIV

    {{ DiscussionBoard.errors[2441303].message }}

Trending Topics