Eyes Glazing Over? Call in an Expert

My ClickZ colleague Mark Sakalosky’s recent article “Test Versus Control, Part 3” took me back to the days of my college statistics classes. I remember hoping as I sat in class that I’d never need to use anything I learned there. I hated it that much. And here I am today, a marketer in a world where data multiplies exponentially every time I blink.

But even the statistically challenged can appreciate the need to apply advanced methodologies to interpret the complex data we have at our disposal. Software has improved over time to make this easier for us, and as today’s kids get smarter and smarter you’ll eventually just be able to ask any 10 year old to help you with a complicated regression analysis.

Thankfully, some statisticians out there work on a project basis. A statistician can be particularly helpful in several situations, such as when there’s a lot you need to know in a short amount of time.

Not Enough Time

One of my clients acquired a new business unit right before marketing plans for the next year were due. How do you develop a marketing plan for a business unit you’ve just it a fast-track introduction to customer traits and behaviors.

In just a few weeks (rather than a few months or even years), the marketers learned some critical things about their soon-to-be-integrated new customer base. The product was a service sold by the month. As you’d expect, customers were most likely to cancel in the first 30 days, and low service usage was the primary indicator of cancellation risk.

But they discovered another significant period of increased cancellations: in the 13th month. Low service usage had little to no correlation to likelihood of cancellation in this case. The jury is still out, but the marketers were instantly suspicious of a “happy anniversary” email that went out after a year and might be causing customers to reevaluate the product necessity. The statistician’s findings gave great insight so the marketers weren’t forced to develop the plan blind.

Too Much Data

Another situation where a statistician might be particularly helpful is when you have data that contains so many variables it exceeds your capacity to evaluate. Demographic data, usage data, dependent variables, independent variables — where does one start?

A former employer collected extensive self-reported demographic and interest information from its subscribers. The demographics and interests alone amounted to thousands of possible combinations. Then add in the online and email usage trends. How can one overworked analyst make heads or tails of all that data to understand who these subscribers are?

Enter the statistician, who was able to sift through the data and highlight correlations we might not have thought of. Perhaps we would have checked the likelihood that football fans were also baseball and golf fans. Or maybe we would have checked to see if golf fans were also cigar smokers. But the statistician took us to a deeper level because he was able to consider more simultaneous variables than our feeble minds could fathom.

We suspected our subscribers’ interests did not fit neatly into the categories on our registration page; they weren’t simply sports fans or online gamers or travelers. The cluster analysis confirmed this. It crossed our predetermined interest categories to help us categorize based on a totally different set of attributes. For example, one cluster’s categories indicate they are primarily time-challenged people looking for efficiency in their lives, while another cluster appears to be made up of thrill-seekers looking for the next adrenaline rush.

Confusing Data

Statisticians might also be helpful when you just don’t understand your data. You know the answer is there, but you can’t find it.

This story is one I heard secondhand, but it’s interesting nonetheless and has long been one of my favorites for inspiring “outside the box” thinking. A site that sold plants, gardening supplies, and related items to an initial customer base of hard-core gardeners was having trouble understanding its customers’ purchasing patterns. The avid gardeners were easy to pick out and develop offers for. For example, offer fertilizer or potting soil along with the appropriate plant.

But a growing number of other customers were buying plants here and there, in combinations that didn’t always make sense, and showing no interest in the additional supplies that go along with gardening.

The group that didn’t fall into the avid gardener category was large and hard to categorize. The marketers were having trouble developing offers because the group’s purchases were so varied, such as people in the north purchasing plants that only grow in tropical regions and other confounding trends. Were they gifts? Were they for indoor use?

The statistician presented clusters of customers from this rogue group linked by purchase types and behavior. Suddenly the marketers had new customer categories with predictable purchase patterns. With a little market research, they were able to confirm their new hypotheses. For example, they had a segment of outdoor entertainers who were using plants to round out their decorating schemes without regard to whether the plants would continue to grow after the main event. An offer for outdoor candles was therefore more appropriate than one for gardening gloves. A stainless steel planter could be offered as a creative approach to cooling beer.

I don’t profess to understand how to choose the best statistician, and, as you’d expect, their rates can be all over the map. A good place to start for a referral is the statistics department of the nearest college. Your choices range from a college intern to a full-blown specialty firm, and your results can range from a confirmation of what you already know to mind-blowing revelations, with a few head-scratchers thrown in for good measure. Whatever the results, in my experience they’ve been useful, thought-provoking, and a great help in developing effective marketing programs.