• by Peter Moloney , Op-Ed Contributor, March 3, 2015

Twenty years ago, if you wanted a marketing piece to look good, you had to go to a professional graphics service to get it typeset, and then a printer to get it printed. Now everybody can do that on a PC.

But then again, you can now use purpose-built templates to make your own just-about-anything: really good-looking marketing pieces, websites, emails, you name it. In other words, you can go a long way for a lot less time and money these days.

For the most part, if you’re a marketer seeking to improve marketing effectiveness by taking advantage of the data science of predictive analytics, it’s like 20 years ago in typesetting and printing. You may know that advances in data science make it possible to predict customer behavior with surprising accuracy, and when those techniques are used to drive marketing campaigns, they have proven to deliver extraordinary results.

If you’ve looked closer, however, you also may have found it requires data scientists to mine customer data and build models used to “score” or segment customers around your marketing objective (for example, which customers are most likely to buy product A?). It’s a lot of work by people who have deep math and analytic skills, thorough knowledge of your company, data, and marketing objectives, and specialized expertise in analysis tools. So it’s expensive and not exactly nimble when you’ve got deadlines.

Unfortunately, that’s not all. Even before the analysis work can begin, you’ve got to collect all the data and integrate it into a form suitable to your analysis tools. That’s no small challenge.

Practically every tool or platform for predictive customer analytics on the market today is designed to find correlations between customers you are looking for (e.g., people who buy product A) and attributes that seem to be associated with those customers, which may be predictors of their interest in buying product A (e.g., people who are over 40, own a dog, and clicked through a certain landing page seem most likely to buy product A.) To discover that in your analysis, you’d first have to create a data store that included every customer’s age, pet ownership status, and click-stream history.

Since you don’t know ahead of time which of your customer’s attributes will turn out to be the most predictive, you want to load up your data store with as many customer attributes as possible. Since this data likely comes from many different systems, some person (or persons) has a massive data integration project to do before you can start the analysis. You don’t hear about it from most analytics platform vendors, but this integration project easily consumes 90% of the time for many analytics projects. And if you can’t get all the data you need for every customer, well, you’ll just have to make some compromises and figure a way around that in your analysis.

The good news? As problems are better understood, everything moves in the direction of do-it-yourself. For marketers, predictive analytics is finally on the move, too. For specific purposes, such as amplifying response rates from a cross-sell campaign, the predictive data science that creates that magic can be highly automated, Web-based, and easily used by marketers executing campaigns. Predictive customer analytics is becoming practical for marketers to exploit on their own, even for making personalized 1:1 best offers to each individual customer.

There will always be a place for the specialists with exceptional skills, but everyday marketing by the rest of us is about to get a whole lot better.