A number of years ago, an interesting debate raged between a senior manager and an agency analyst in a Boston ad agency.
“I’m not buying it,” said the senior manager. “You mean to tell me our client, a large telecommunications company, is wasting millions of dollars because of the way they conduct their in-market tests?”
The analyst making the suggestion wasn’t sure exactly what to say. It was true. The client was spending more money on testing than was necessary to obtain the answers it wanted. Delivering that news was difficult, as the agency itself had designed and conducted the tests for years.
The concept of properly designed multivariate tests has been around since at least the early 1900s. The first practical applications were in the agricultural sector. Farmers wanted to increase crop yields, so they’d test how different combinations of variables (fertilizer, water, etc.) affected plants.
It wasn’t feasible for them to test one variable at a time (known as an A/B testing). They didn’t want to treat half their fields with one amount of fertilizer, the other half with another amount, then wait for harvest time for the results. And then have to wait again until the following year to find out what affect water had on the yield.
Instead, the farmers use “factorial” or “fractional factorial” tests. This is a carefully managed methodology of treating different parts of the fields with different combinations of factors (fertilizer level, water level, etc.). Results were carefully analyzed to reveal the optimal levels of all tested factors.
The telecommunications company mentioned above (I was the analyst, by the way) was in a similar situation. It was testing multiple factors, including offer levels, promises, feature combinations, and so on, and wanted to know the optimal combination of all the factors. Unlike the farmers, the client was using very carefully crafted A/B testing rather than the more useful fractional factorial tests.
Not a New Idea
Applying advanced test design to marketing isn’t new. It’s mentioned briefly by David Ogilvy in his classic “Ogilvy on Advertising.” The practicality of designing and conducting such tests in marketing is greatly increased by the technology advancements of the last decades. Computer power and software are readily available to help marketers (and others) take advantage of the designs.
The benefits of using multivariate testing include reduced sample sizes, faster results, and potentially better results than A/B testing.
Not Trivial, But Doable
As evidenced by that Boston discussion a few years back, marketers must study testing techniques before they understand how the tests work. They can be quite beneficial. Gordon Bell wrote a good article explaining the concept of advanced test design in the marketing world.
Did I Mention Software?
Marketers and analysts can chose a number of software routes to help with the design and analysis of advanced tests. SAS offers two: a helpful interface into the relevant SAS procedures, and one from a company formerly known as JMP.
The updated version of Charles Hicks’ “Fundamental Concepts in the Design of Experiments” discusses how to use SPSS and Minitab to analyze test results.
Today and Yesterday
I’m happy to report that today, the Boston agency uses these testing techniques to successfully design, execute, and analyze tests and experiments that examine many variables. The tests are conducted both in direct mail and Web site design applications.
How do I know about the history of these tools? Many years ago, I spent some time in the agricultural world.
Marketers create personas to better understand their target audience and what it looks like. If marketers can understand potential buyer behaviors, and where they spend their time online, then content can be targeted more effectively.
What’s behind a successful data-driven marketing strategy?
One of the major challenges in the martech industry is getting the attention of prospects in a world where they are bombarded by content and emails on all sides.
Facebook is addressing one of the biggest missing pieces of its chatbot offering: analytics.