Archive for AB testing
Deconstructing a Marketing Test Hypothesis
Posted by: | CommentsA marketing test hypothesis is a powerful and necessary part of your marketing optimization program when running tests. I am going to take you through creating a simple hypothesis.
A hypothesis clearly states:
- What you are testing
- What your control and experimental groups are
- What outcome you predict will happen (based on an educated judgment)
- What is the alternate outcome
- What you will need to track specifically in order to prove or disprove your test prediction.
Let’s look at a hypothesis more closely.
A hypothesis is clear and specific, testable, and can be proved right or wrong.
Look at these differences between a prediction, a question, and a hypothesis:
A Test Prediction: Not asking for a phone number on my registration form will increase registrations.
A prediction is the outcome you expect – more or less your educated guess of what will happen.
A Test Question: Will not asking for a phone number on my registration form increase registrations?
A Test Hypothesis: Paid Search traffic reaching my registration form that does not ask for a phone number will produce more registrations than Paid Search traffic reaching my registration form that asks for a phone number.
A hypothesis states with conviction what results you expect to see from your test, both from your control and your experimental group – it is here that you will state your test prediction. And since we know that a hypothesis has to be able to be proven either right or wrong, we only have 2 possible outcomes – either my registration form that does not ask for a phone number increases registrations over my control that does ask for it, or it doesn’t produce more registrations(or a tie).
And finally, let’s break down the above marketing test hypothesis to show specifically what it explains:
- Paid Search traffic reaching my registration form that does not ask for a phone number is the experimental group
- Produce more registrations is the outcome we expect from the experiment and what we want to track
- Paid Search traffic reaching my registration form that asks for a phone number is the comparison or control group.
If you wanted to run this test with multiple panels each having a different form field requirements (size), you could in reality replace phone number with less fields:
Paid Search traffic reaching my registration form that has less than 7 fields will produce more registrations than Paid Search traffic reaching my registration form that has 7 fields. “
You are still testing your hypothesis of that less fields will produce more registrations than your control of 7 fields, but you will determine from your testing which length is optimal to rollout with if your hypothesis is true.
With your completed hypothesis you can now execute your marketing test to your website visitors and let them prove or disprove it. If they prove your hypothesis to be true then you did a great job with your hypothesis’s prediction, if they disprove it, you still have learned something (make sure you take away lessons from each test!) – The test is not a failure, just try again after formulating a new hypothesis.
How to Document Your A/B or Multivariate Test
Posted by: | CommentsIt’s imperative to document every effort of your online testing and optimization program. Not only to see the progression of improvement over time, but to also have reference for future tests you are planning or questions that may arise from others you are working with.
How I document my A/B and multivariate testing is as follows:
First, I primarily use Excel for all my documentation efforts due its ease of use.
I create a Master Test spreadsheet that serves as a very top-level summary for quick glancing. This spreadsheet is constructed to have column headers for the following:
- Test name – giving each test a descriptive and unique name
- Test date range – documenting the start and finish date of my tests
- Test hypothesis – why I am running this test and why I believe my outcome will be such
- Results – brief description of the result
Each row is for a single test. I then link the Test Name cell for each test to another spreadsheet that is built specifically for that test (If you have ten tests in the Master Test spreadsheet each test would link to its own Individual Test spreadsheet for a total of 10 spreadsheets plus the Master Test Sheet).
Each Individual Test sheet contains multiple tabs and information, and this is where the detail will go throughout the test.
The first tab contains the test information and summary and broken into sections:
- Test name
- Test date range
- Test hypothesis
- What type of test (a/b, multivariate), and how many panels or combinations
- Traffic data (source of traffic, and current traffic stats)
- What type of metrics I will be using to determine the results and how to determine the metrics (is conversion impressions divided by sales, or impressions divided by clicks on a certain button etc?)
- A space to record final metric results (control performed as such, top performers identified individual performed as such)
- Learning’s (both as the test is live and from the results)
- Ideas for future tests based off of this information
- Next actions (will this be rolled out etc.)
- Miscellaneous notes
I also have other tabs in the Individual Test spreadsheet:
- Screenshot of control
- Screenshots of test panels or combinations (depending on how many there are). If there are too many panels or page combinations to take screenshots of, after the test is ended I take screenshots of the top performing test panels for future comparisons)
- Screenshots of Test statistics (When I am using Google Website Optimizer I take daily screenshots of the stats admins. and store in a separate folder, but the final screenshot from the point at which we end the test is stored in the Individual Test spreadsheet – just in case I transcribe something wrong I have an actual reference to go back to.
- Various other tabs as necessary for reference such as more detailed metrics information, etc.
I also keep a folder for each test (with the folder using the test name) that contains my test spec PowerPoint so that I can see all of the elements or options that we are using, analytics data, screenshots of everything-basically anything used from the conception of the test all the way through to the end.
What this enables me to do is at any point in time have a huge history of each test both from a visual standpoint and data-driven standpoint. The Master Test sheet gives me quick access to the individual tests but also a timeline of the testing I have done.
A/B Testing Calculators
Posted by: | CommentsDo you need an A/B split testing calculator to determine which sample or page is the winner of your test, to see if your test is statistically significant, or to calculate how many impressions or conversions you will need to run your A/B Test?
Here are four A/B testing calculators for you to use to determine your test results:
- SplitTestCalculator.com’s Split Test A/B Testing Calculator:
My easy-to-use online split test and A/B test calculator to determine if you have a statistically significant winner from your landing page, ppc, email, or direct mail split A/B test.
- Bulldog Solutions’ Simple A/B Testing Calculator and Advanced A/B Testing Calculator: (update: looks like they have taken it down)
Two different excel spreadsheets with video demos on how to use them. The first spreadsheet, the Simple A/B Testing Calculator has a preset 90% significance level. The Advanced A/B Testing Calculator allows you to set your significance level and provided a separate tab to calculate the sample size needed. You can also view their PDF on What’s Under the Hood of their A/B Testing Calculators. - PPC G-Test Calculator by SEOTools:
This online split testing calculator will let you uncover the results to a 2 sample test. It tells you the winner of the test with the confidence level. Be careful though as it’s not based on a preset significance level nor can you set one. But reading the test analysis you can easily see what the confidence level of your results are and decide for yourself if there is enough statistical significance. - UserEffects’ Split Test Calculator & Decision Tool:
A simple to use online testing calculator, however this one gives you an analysis of your test including an estimate of how many more visitors you might need to reach a 90% confidence level in your test.
Google Testing their Business Solutions Page
Posted by: | CommentsThis morning I went to log into my Google Analytics account and noticed when I first went to Google’s Business Solutions page I had entered into one of their Google Website Optimizer tests on that page.
It looks like they are testing only the top portion of the page and testing the following:
- Layout of the top paragraphs (3 paragraphs straight down vs. 1 paragraph on top and 2 below side by side)
- Position/Placement of the Get Started button on the page
- Background box around the button (visible vs. not visible)
- Google Analytics paragraph healdine (Increase website conversions and marketing ROI vs. Increase website conversions and marketing ROI with Analytics)
Here are screen shots from the test 5 pages displaying in their test (click thumnail for full size):
I should note that these screenshots only show the top of the actual page where I noticed the different test elements, to see the entire page you will need to visit their Business Solutions page.
A/B Testing Low Traffic Web Pages
Posted by: | CommentsSometimes the website or the page you want to test just doesn’t have enough traffic to really perform an large multi-panel A/B or multivariate test on it with multiple elements or variations. But perhaps you still want to improve upon its performance. Maybe you just recently launched a website and it’s receiving some traffic but you still really believe that you could influence the current traffic better to get better results. Maybe you aren’t getting 100 orders per week, but you are getting 10, and you believe that you should be able to get to 20 orders with improvements to the page without an increase in traffic.
Lower traffic shouldn’t stop you from still performing a small A/B test as you can still learn a great deal of information with the web traffic that you do receive. But the key is to test the high impact elements on the page that are easily noticeable either consciously or subconsciously by the visiting traffic rather than the minutia that would allow you to see incremental improvements with higher traffic pages.
High impact elements will vary according to your page design, but think along the lines of your:
- Main Headline
- Page Background Color
- Main Hero Image
- Drastically reducing or increasing your registration form fields
- Your Offer
Of course with less traffic you will only be able to test a few variations due to the fact that you don’t have the traffic to get statistically significant results with lots of variations. However, this shouldn’t stop you from creating a list, sorted by priority of impact to continue testing over time. Maybe you can’t test 6 different page variations right now in the first round of testing, but it might be viable for you to test 2 page variations, and then continue testing your other ideas with the winner of the first round of tests.
Want to learn more about A/B testing your low traffic pages? Check out How to Do A/B Split-Testing on Lower Traffic Sites with Bryan Eisenberg at Dr. Ralph Wilson’s Web Marketing Today blog.
Good luck and remember although you might not be able to test everything you want to right now, anything you can do to statistically improve your results is a good thing!






