Website Tests Make More Money

Written On Day: 57
Lump of Coal

A Christmas Carol by Charles Dickens is a very inspirational story.  Scrooge is a man who knows his business.  He operates it at full optimization.  Every aspect of the business is calculated and profits are maximized.  All that is required is one piece of coal – so all he gives is one piece of coal.  Even the merriment of those around him does not distract from his purpose.  To run a highly profitable business.  Scrooge would be a wealthy webmaster.

If you are trying to make money from a website then you need to think about what is best for the website.  What will bring in the most money.  Your website’s purpose is not to make the world a better place or to help you find the meaning of life.  It’s purpose is to bring in the most amount of money possible. That is it. 

OK, so we have established the purpose of your website.  Now what?  How do you make the most amount of money possible?

Guessing What is Best For The Website

Currently this website makes around $2 dollars a day.  Could it make more?  Yes it can.  The chances that this website is fully optimized is very low.  The website was designed based on what I thought was best.  Not a really scientific method.  Put the ads here, the content there, the header should say this – hopefully that will work.  Unfortunately, at the beginning that is all you can do.  Just guess.  But Scrooge did not become rich by guessing how much coal to give away.  He knew exactly how much Cratchit needed to survive through Christmas.

To find out what makes the most money you need to perform tests on your website.  Tests that will tell you what works and what works better. 

A-B Test

A-B testing is one method to test what makes more money on your website.  You make two pages with  slight variations and then see which one makes the most money.  For example, on one page you put the text ads on the left and on another page you put the text ads on the right.  Run one page for a week and then run the other page for a week.  Compare the results and keep the page with the best returns.

The important aspect of this test is that the content must be the same for both pages.  You are not testing the content.  Your results would be useless if your changed the content.  The content is the product.  You are just changing the packaging.  Test what packaging sells the most content.

Change the location of ads, type of ad, location of the content, the size of the header, the text of the header.  What works better a banner ad or a text ad, a large font bolded header with short text or a small font  but descriptive header text?  ‘THIS IS A GREAT POST!’  or ‘This is a great post because I wrote it’.

Running A Test In Parallel

When performing an experiment it is important to keep the amount of variables that could effect your results to a minimum.  Isolate what you are testing.  Ensure what you are testing is only effected by factors you control.

The normal method of doing a webpage optimization test is to run a page for a week.  Record the results.  Change the page.  Then run the new page for a week.  This method has a major flaw.  It is run in two different times.  There are too many external factors that could effect the results.

The problem is that between the two weeks you receive different types of traffic.  For example, during the first week your website might rise up in reddit.com and you get lots of visitors from reddit.  These are completely different types of visitors then the ones you get the second week in which you make a great post on a niche related forum.  The first week was anybody and everybody, the second week were niche related visitors.  Two different type of visitors.  This situation makes your test completely useless.

What you need to do is run the two pages at the same time.  So that the test is not effected by the type of traffic.  Each page is viewed by the same type of traffic.  You need to run the two pages at the same time.

How to Run A-B Test in Parallel

I created PHP code to run A-B tests in parallel.  Here is what I do:

  1. Pick a page on website that I want to test. Lets say tinytim.php.
  2. Make an copy of it. Now I have two identical pages.
  3. Rename the original page to tinytim_1.php and the copy to tinytim_2.php.
  4. I make a small change to tinytim_2.php. Ensuring that the pages have different ads or create specific ad channels for each page, so that you I have a way of measuring the results.
  5. Create a new third PHP page. Call it tinytim.php.
  6. Paste this code into it:

  7. <?php
    srand(time());
    $random = (rand()%2);
    if ($random==0){
        include('tinytim_1.php');
    }else{
        include('tinytim_2.php');
    }
    ?>

How It Works

When the tinytim.php is requested the PHP code randomly picks one the the two pages to display.  There is a 50% chance that  tinytim_1.php will be displayed and a 50% chance that  tinytim_2.php will be displayed.  The PHP code will pick one of the pages and then copy the page to the tinytim.php page.

The visitor does not notice anything different.  As far as they are concerned they are viewing the tinytim.php page.

After a week of running this I compare the results of the different ads.  Which one made more money.

With this method you are certain that your tests is not effected by different types of traffic.  Each of the two pages has an equal chance of being called at any given time.  On a given day if there are a hundred visitors then 50 of them will be presented with tinytim_1.php and 50 will be presented with tinytim_2.php.  The daily variations in  types of traffic does not effect the test results.  Any A-B website test method that is not in parallel – I say ba humbug.

  1. No comments yet.
  1. No trackbacks yet.
You must be logged in to post a comment.