How to Conduct Split Tests in Marketing
Choose B2C or B2B., Establish control groups., Identify test material., Design material side by side., Setup an email capture page or complimentary contact form., Launch the campaign., Track progress proactively., Tally results when test expires...
Step-by-Step Guide
-
Step 1: Choose B2C or B2B.
Too many businesses make the mistake of using a B2C approach when engaging in B2B split tests, which ends up throwing campaigns into disarray and leads to overspending.
Interacting and catering to the end consumer is not the same as establishing relations with another enterprise.
A whole different set of testing objectives and strategies must be established before control groups are formed in order to purport an accurate test, the reason your marketing head should decide if B2C or B2B should be the focal point of the test. -
Step 2: Establish control groups.
Otherwise identified as 'A' and 'B' groups, control groups are the focal point of all split tests.
These can be two landing pages, two blog posts, two contrasting sets of females, banner advertisements or anything indicative of the current business's interests.
It's vital that these two groups have few, if any, similar characteristics as the accuracy of the test relies on having control groups which will yield different end results. , In marketing nomenclature, this could involve showing advertisement 'A' during peak business hours, and advertisement 'B' during weekends; sending offer code 'SAVE' through Facebook advertisements, and offer code 'SAVEMORE' through Twitter hash tag optimization.
The more specific the test, the higher the likelihood your actual marketing campaigns will increase in effectiveness and ROAS (return on ad spend). , Remember, you want each group to see campaigns which provide one similar end result, but offer two separate paths in getting there.
For example, if designing two distinctly different banner advertisements, have your illustrator or creative writer swap similar keywords to gauge reaction but lead the potential buyer to the same landing page.
Because 70% of A/B split tests fail historically, affording the time for this extra step will increase your success rate. , When your campaigns have concluded, you'll need to send surveys or have permission-based phone calls placed to judge the response rate based off carefully designed questions.
Make sure any forms or email captures are permission-based. , During the launching phase, you'll benefit more by submitting these two distinctly different campaigns through the same medium (i.e.
Facebook, Bing) since you'll essentially want the same individuals viewing two different ads.
You'll run this campaign for a predetermined amount of time; 30 days is sufficient, however, if you have a large control group.
Make sure some method of tracking campaigns has been activated; this could be some proprietary in-house method, or your standard Google Analytics. , View your campaigns' ongoing progress from a distance, perhaps jotting notes and scoring leads as they come in.
Avoid ending the campaign before it's due time as you'll miss information and discount potential sales or interested parties from your control group. , You will have several pieces of data, including analytical data and any sales or email responses generated from your testing material; print several copies of these results, and rank the data by order of importance.
So, if you wanted to test a control groups response to your advertisement by time of day, pull only those results from your analytics and mark them '#1'
country of responder '#2' and so forth. , Although your surveys should be written for the layman, it's vital your questions directly correlate to information you wanted to collect in Steps 1 and
2.
Send these to those who opted in during your campaign; you'll tally your results based off responses received back in multiples of 10; so, if 102 sent replies to your survey, it would be much simpler (for purposes of grading) to count 100 of those, discarding two which seem to be the least complete; if 80 send responses back, you can use all of them since 80 is divisible by
10. , The marketing head will probably want your final report; therefore, compile all surveys into one report, all analytical data into one report, and present this in either paper or PowerPoint format for easy dissection. -
Step 3: Identify test material.
-
Step 4: Design material side by side.
-
Step 5: Setup an email capture page or complimentary contact form.
-
Step 6: Launch the campaign.
-
Step 7: Track progress proactively.
-
Step 8: Tally results when test expires.
-
Step 9: Send out surveys.
-
Step 10: Compile surveys and data into one report.
Detailed Guide
Too many businesses make the mistake of using a B2C approach when engaging in B2B split tests, which ends up throwing campaigns into disarray and leads to overspending.
Interacting and catering to the end consumer is not the same as establishing relations with another enterprise.
A whole different set of testing objectives and strategies must be established before control groups are formed in order to purport an accurate test, the reason your marketing head should decide if B2C or B2B should be the focal point of the test.
Otherwise identified as 'A' and 'B' groups, control groups are the focal point of all split tests.
These can be two landing pages, two blog posts, two contrasting sets of females, banner advertisements or anything indicative of the current business's interests.
It's vital that these two groups have few, if any, similar characteristics as the accuracy of the test relies on having control groups which will yield different end results. , In marketing nomenclature, this could involve showing advertisement 'A' during peak business hours, and advertisement 'B' during weekends; sending offer code 'SAVE' through Facebook advertisements, and offer code 'SAVEMORE' through Twitter hash tag optimization.
The more specific the test, the higher the likelihood your actual marketing campaigns will increase in effectiveness and ROAS (return on ad spend). , Remember, you want each group to see campaigns which provide one similar end result, but offer two separate paths in getting there.
For example, if designing two distinctly different banner advertisements, have your illustrator or creative writer swap similar keywords to gauge reaction but lead the potential buyer to the same landing page.
Because 70% of A/B split tests fail historically, affording the time for this extra step will increase your success rate. , When your campaigns have concluded, you'll need to send surveys or have permission-based phone calls placed to judge the response rate based off carefully designed questions.
Make sure any forms or email captures are permission-based. , During the launching phase, you'll benefit more by submitting these two distinctly different campaigns through the same medium (i.e.
Facebook, Bing) since you'll essentially want the same individuals viewing two different ads.
You'll run this campaign for a predetermined amount of time; 30 days is sufficient, however, if you have a large control group.
Make sure some method of tracking campaigns has been activated; this could be some proprietary in-house method, or your standard Google Analytics. , View your campaigns' ongoing progress from a distance, perhaps jotting notes and scoring leads as they come in.
Avoid ending the campaign before it's due time as you'll miss information and discount potential sales or interested parties from your control group. , You will have several pieces of data, including analytical data and any sales or email responses generated from your testing material; print several copies of these results, and rank the data by order of importance.
So, if you wanted to test a control groups response to your advertisement by time of day, pull only those results from your analytics and mark them '#1'
country of responder '#2' and so forth. , Although your surveys should be written for the layman, it's vital your questions directly correlate to information you wanted to collect in Steps 1 and
2.
Send these to those who opted in during your campaign; you'll tally your results based off responses received back in multiples of 10; so, if 102 sent replies to your survey, it would be much simpler (for purposes of grading) to count 100 of those, discarding two which seem to be the least complete; if 80 send responses back, you can use all of them since 80 is divisible by
10. , The marketing head will probably want your final report; therefore, compile all surveys into one report, all analytical data into one report, and present this in either paper or PowerPoint format for easy dissection.
About the Author
George Ferguson
Brings years of experience writing about cooking and related subjects.
Rate This Guide
How helpful was this guide? Click to rate: