Use the scientific method to optimize your SEM campaigns
Published: April 4, 2013
Author: Dave Yoo
There have been numerous occasions when colleagues check out my LinkedIn profile, notice that I have an undergraduate degree in Molecular and Cell Biology from UC Berkeley, and proceed to ask me, “How did you end up in online marketing?”
The unsexy answer to that question is, “I just changed my mind.” But rather than stonewall the conversation with such a boring answer, I often like to explain how my degree has benefited me in online (specifically performance) marketing by teaching me how to use the scientific method.
In the event that you actually studied in a field that’s relevant to online marketing, here’s a very simplified outline of the scientific method:
1. Construct a hypothesis (educated guess)
2. Test your hypothesis through an experiment
3. Analyze your data
4. Draw a conclusion that answers your hypothesis
Another practice that’s critical to employing the scientific method properly is making sure the experiment is a “fair test,” when only one factor (variable) is changed while keeping all other conditions the same. More often than not, unfortunately, I’ve seen online marketing tests that neglect this practice and end up either wasting a lot of time or more detrimentally lead to decisions that end up doing way more harm than good.
Here’s a simple example of how I employed the scientific method when optimizing a landing page for an email sale-alert shopping property owned by POPSUGAR called MyPerfectSale.
A simpler, cleaner, and more modern landing page will result in higher conversion rates and lower cost-per-acquisition (CPA).
1. Build a simpler, cleaner, and more modern landing page for one designer (True Religion in this case):
2. Set up a brand new ad group for True Religion and split traffic evenly to two identical ads with one going to the old landing page and one going to the new. (In the spirit of the “fair test” principle, we made sure that ad history did not taint our results.)
3. Launch the new ad group and allow at least one full week and 1,000 clicks on each ad to accommodate for day-to-day fluctuations and statistical significance.
Old landing page:
– Conversion rate: 13.6%
– CPA: $2.46
New landing page:
– Conversion rate: 25.1%
– CPA: $1.71
The new simpler, cleaner, and more modern landing page generated significantly higher conversion rates (85% improvement) and lower CPA (30% drop).
Seeing the results of this test in a structured and disciplined format gave us the confidence to make the bold move to change all the other landing pages immediately (of course it didn’t hurt that the results were so dramatically favorable). Here is the actual graph from the account once we made the switch to most of our campaigns:
An additional benefit of conducting tests using the scientific method is that they feed into the next tests in a logical fashion. Soon after we finished rolling out the new landing pages, my team was eager to propose the hypothesis for the next test in our landing page optimization efforts.
It could just be my attempt to justify sitting through over 300 hours of organic chemistry labs many years ago, but I still to this day find using the scientific method extremely helpful, not only in performance marketing but in achieving overall operational efficiency.
Have you had similar marketing successes using the scientific method? Leave a comment!
– Dave Yoo, COO