Let’s Talk Data

I sat down this morning, and before my first cup of Java even got cold, I happened upon an article which essentially says that market segmentation is too inefficient to be an effective tool. The author goes on to explain how other methodologies can be employed to pinpoint “buckets” (instead of segments (?)) of customers by targeting individuals and their behavior.

My reaction? “Uh, yeah…”

As if this were some great revelation.

I’m not trying to disparage the author. He makes valid points. And frankly, this isn’t the first negative article I’ve seen about segmentation. However — think about this. What’s the number one marketing issue facing most businesses today? If you want to grow, you need to add new customers. First, you determine who your best customers and then and you have to find more just like them and sell to them in order to increase profits. That’s called prospecting.

I have discovered an incredible amount of well-educated, well-intentioned marketers dismissing broad based neighborhood segmentation systems. And you know what? For companies that maintain lots of transactional data on their databases and can afford to append household level demographics — they are 100% correct. If you are looking to optimize your current customer database — then it is most efficient, most effective, to use modeling techniques to score individual customers with propensities to purchase different products/services the company provides.

But even the most data-rich companies still have to prospect like the rest of us. And you know what? Modeling is based on having similar data on both the customer file and the list of prospects that you wish to score. Simply put, you can maintain gadzooks of data on your own database but no other source is going to be able to put that exact same data onto any other database because it doesn’t exist.

Take this example. John Smith, your customer, spent $500 last year. He called the support line 10 times, he pays with an American Express card and you know all the products he bought. Now you are going out to buy a list of prospects. Meet Mary Jones — she’s on the list you purchased from expiring, InfoUSA, Acxiom — what do you know about her? You’ll get her name, address, demographics (some individual, some neighborhood (block group or zip+4) based, and usually some summarized credit information. You don’t know how many times she’s called the support line or what credit cards she pays with — because she’s a prospect — not a customer! You have no customer data.

Repeat this with me. The only way to tie customers to prospects through modeling is to find characteristics that are on both files and use those characteristics to predict potential behavior.

So, when I read broad stroke articles like the one below, I shake my head. After 15+ years in this business (working for the leaders in the industry) I can tell you that the majority of small to medium sized businesses — and truthfully, some of the bigger ones too, HAVE VERY LITTLE TO NO TRANSACTIONAL DATA worth modeling. And, budgets are limited, time lines are short (”Help! We need to mail in 3 weeks!”) so we can’t append all kinds of household level demographic and behavioral data for prospecting purposes. What’s that saying? You can have it fast, you can have it perfect, and you can have it cheap, but not all three at the same time.

So, what’s the solution? Neighborhood segmentation — with a twist.

No matter what flavor you use — Community Tapestry, MOSAIC, Personix, PRIZM — you still need to tweak a bit to get the maximum lift if you plan to do direct marketing (mailing) by clusters. I think one of the best ways is to isolate the top clusters that have the greatest propensity, then take that small group, have household-level universally available data appended, and then further refine your target.

Imagine if you will, this scenario. You sell (relatively) pricey clothing and furniture for young children. If you just target neighborhoods with children you’ll likely get a small response. If you target to people with a certain income (and children), you’ll get a slightly better lift. However, if you use the segments that have lifestyles which are typically status-conscious (upscale suburbia and urban elite) along with presence of children and income — you’ll get the most bang for your buck without the expense and time needed to do advanced custom analysis.

If you simply accept those articles that dismiss practices that have been used successfully for years — without addressing the positive reasons why these tools can still be used — then you’re not getting the complete picture.

I maintain that prospecting is a completely different ball game — and segmentation tools can help businesses of all sizes further narrow their prospect lists.

from: Segmentation: thin edge of the marketing wedge.

source: B&T Weekly, October 14, 2005.

via: HighBeam™ Research

COPYRIGHT 2005 Reed Business Information Ltd. All Rights Reserved.

I’m always on the lookout for new tools to help my clients characterize and segment their customers into easily marketable target groups.

For years, it seemed like Claritas was the only game in town. And when I worked for them, I liked it that way. I must admit, they did they have an incredible marketing engine. Plus, they hired top notch people, and let them apply their tools in novel applications. They were aggressive and knowledgeable — a killer combination.

When I first started selling for Claritas, I used to make this analogy. Remember high school — and all those cliques that formed? You had the jocks, the geeks, the cool kids, and even the invisibles. Despite the fact that “school spirit” is all about one for all and all for one — it’s human nature for like to attract like. Now not all the popular girls, jocks, etc. were exactly alike — however, they had certain things in common that made them stick together under certain circumstances.

Marketers know that its human nature for people to bond and gather with other people that share common characteristics. And they seize on that — using data and computers — to get messages out to prospects that are similar to their best customers. Now, there’s no way that everyone in a neighborhood, much less a street is exactly the same. However, if you look around your own town, you can easily see that there is a wealthy section, an ethnic section, a college dweller section, even the “poor side of the tracks”.

Back to Claritas. Sure, there were other companies, other systems out there (MicroVision, ClusterPlus, Acorn, etc.) but no other company seemed to focus on getting links to syndicated data souces like Claritas.

The crazy thing is that this is one of the EASIEST things in the world to do. If you are not familiar with commercially available marketing tools like PRIZM, Tapestry, MOSAIC, Cohorts, Personix — here’s how it works.

99% of all the commercially available systems (likely 100% but I’m not positive!) start with a “base” of data from the good old US Census. Through the magic of powerful computing strength under the watch of a team of demographers and statisticians, a data soup base is started. Each company adds its own data to the mix — usually list counts, behavioral data (from surveys, warranty cards), magazine subscription data, auto ownership — even aggregated credit data — for a unique flavor.

There are a variety of techniques that are used, but suffice it to say, at the end of the day, a number of groupings (aka clusters, types, segments) are formulated. Each group has a great deal of similarity on a number of characteristics (such as home value, presence of children, degree of urbanicity, etc.)

It’s easy to apply commercial segmentation systems because most of them create assignments for neighborhoods at a variety of levels — ZIP+4, Block Group, etc. and can assign types by geocode. (A geocode is series of numbers that tells you the State, County, Tract and block group an address or location is coordinated with. Latitude/Longitude codes let you put an address on a map. “Point coding” is the term most commonly used.)

So, anything that has an address on it can be geocoded and likely point coded, and therefore can be crossreferenced by the commercial segmentation system.

What does that mean? Well, all surveys that contain an address can be type coded, all transactional records can be coded, etc. etc. And that is pretty powerful stuff.

So, back to IRI and Acxiom. This is a brilliant move. IRI has great household level panel data, and now, their Personix tool will be able to link segments (types, clusters, “cliques”) by the household products they purchase — using REAL transactional data. How incredibly cool is that?
Here’s the announcement.

@import “http://www.highbeam.com/css/docLink.css”;

from: IRI and Acxiom(R) Introduce More Efficient and Actionable Approach To Consumer Segmentation and Targeted Marketing.

source: Business Wire, January 25, 2006.

via: HighBeam™ Research

COPYRIGHT 2006 Business Wire

@import “from:%20%20IRI and Acxiom(R) Introduce More Efficient and Actionable Approach To Consumer Segmentation and Targeted Marketing.
source: Business Wire, January 25, 2006.
via: http://www.highbeam.com?refid=blogger”> src=”http://www.highbeam.com/img/h-icon-small.gif” alt=”HighBeam Research Logo” border=”0″ align=”baseline”/> HighBeam™ Research
COPYRIGHT 2006 Business Wire