Bayesian Analytics Upends Direct Mail Testing

What’s old is new again. I’ve recently been introduced to Bayesian Analytics and you’re going to want to know about it, too. It’s been around for a long time (since the 1700s!), but its current usage is new and spreading into many fields because of more powerful computers. Bayesian Analytics is being used in astrophysics, weather forecasting, insurance risk management, and health care policy. And, now, a few cutting edge mailers have successfully used this analytics approach, too.

Over the years, direct mail marketers have relied on either A/B testing or multivariate testing to evaluate winning campaigns. But a confluence of technology and Bayesian Analytics now enables direct mailers to pre-test and predict future response before mailing. 

I think Bayesian Analytics will upend how direct mailers test to identify the highest profit-producing control faster. 

Usually, direct mail marketers test four categories of variables. For example:

1.    Price
2.    Headlines
3.    Imagery
4.    Formats

Within each of those variables, direct marketers often want to test even more options. For example, you might want to test $5 off, versus $10 off, versus 10 percent off, or versus 15 percent off. And you want to test multiple headlines. And images. And formats.

Let’s say you want to test four different pricing offers, four headlines, four imagery graphics and four direct mail formats. Multiplying 4x4x4x4, there are a possible 256 test combinations. 

It’s impractical and costly to test 256 combinations. Even if your response rate dictated you only needed to mail 5,000 volume per test for statistical reliability, you’d still have to mail over 1.2 million pieces of mail. If each piece costs $0.50, the total cost is $600,000.

Knowing there must be a better way to affordably test multiple variables, a new pre-test methodology offers a cost-efficient breakthrough for direct mailers that’s based on Bayesian Analytics (refer to Bayesian Statistics for Dummies for an easy-to-grasp essay).

Bayesian Analysis works with a fraction of the data required to power today’s machine learning and predictive analytics approaches. It delivers the same or better results in a fraction of the time. Applying a Bayesian Analysis methodology, direct mailers can make big and statistically reliable conclusions but from less data. 

The International Society for Bayesian Analysis says the “Bayesian inference remained extremely difficult to implement until the late 1980s and early 1990s when powerful computers became widely accessible and new computational methods were developed. The subsequent explosion of interest in Bayesian statistics has led not only to extensive research in Bayesian methodology but also to the use of Bayesian methods to address pressing questions in diverse application areas such as astrophysics, weather forecasting, health care policy, and criminal justice.”

Bayesian Analysis frequently produces results that are in stark contrast to our intuitive understanding. How many times have you used your intuition to test a specific combination of variables thinking it would result in a successful direct mail test, only to be disappointed in the results?

Bayesian Analytics methodology takes the guess-work out of what to test in a live mailing scenario. Instead of testing and guessing you can now pre-test those 256 combinations of variables before the expense of a live test mailing. The pre-test reveals which combination of variables will produce the highest response rate in the live test, a substantial test savings.

There’s another benefit: you can learn what mix of variables will produce the best results for any tested demographic or psychographic group. It’s possible to learn that a certain set of variables work more successfully for people who are, for example, age 60+, versus those age 40-59. This means you may be able to open up new prospecting list selections that previously didn’t work for you.

As stated earlier, a handful of mailers have already pre-tested this new Bayesian Analysis methodology. It has accurately predicted the results in live testing at a 95 percent level of confidence. Now that beta testing has been completed and the methodology is proven to be reliable, look to hear more about it in the future.

There’s more about this methodology than can be shared in a single blog post. To learn more, download my report accessible from my Home Page