Methodology
In order to produce a multivariable test, Optimost uses a proprietary methodology that was developed in consultation with leaders in the field of experimental design from institutions including Harvard Business School, Columbia Business School and the Yale School of Management. Better than Taguchi designs, Optimal Design was built to help you get the best web page conversion rate in the shortest amount of time. See a description of test design methodologies in our Design of Experiments section.
The methodology that Optimost has developed allows you to evaluate hundreds, thousands, even millions of web page design possibilities in the same amount of time it would take to conduct one A/B test. Using Experimental Design, Optimost is able to iteratively test a much smaller subset of the permutations. As a result, we can obtain highly predictable data as to how much each individual value and variable is contributing to performance. In this way, Optimost identifies the web page that will maximize your online conversion rate, based on a universe of thousands or millions of possibilities.
Optimost is dedicated to ensuring the data from optimization is valid and actionable. Even with 95% confidence in any given experiment, there is a 5% chance that the data is giving you the wrong conclusion! With multiple variables being tested, this likelihood grows. That’s why Optimost recommends conducting tests in a series of mini-tests, or "waves," which allows you to narrow down winning creatives and variables and validate the results of previous waves. All of this ensures that you have indeed optimized your site!
Multivariable Optimization is:
Holistic
- Far more robust than just "offer testing"
- Covers much more than just landing page optimization
- Recognizes interrelationships with the page (it is not a "silo" approach)
Efficient
- Requires minimal traffic to get statistically reliable data
Iterative
- Reduces the likelihood of "false positives"
Consistent
- Can validate results by persona
- Allows decisions to be based on reliable representative samples
Transparent
- Does not have any "black box" elements; you understand why things worked
