Behavioral versus Demographic Data
# 67: 5 / 2006
Drilling Down - Turning Customer
Data into Profits with a Spreadsheet
Customer Valuation, Retention, Loyalty, Defection
Get the Drilling Down Book!
In This Issue:
# Topics Overview
# Behavioral versus Demographic Data
Hi again folks, Jim Novo here.
Most businesses want their visitors or customers to "do
something" - to take an action of some kind. Trying to
drive action, businesses engage in marketing / advertising to reach
"audiences" with their message.
These audiences can be quantified in a number of ways using
Demographics, Sociographics, and Psychographics for the purpose of
"targeting" the campaign. The idea is to make the
campaigns more efficient by focusing resources on the types of people
thought to be more interested in the product or service.
This is fine. But from psychology and actual practice, we
know behavior predicts behavior and demographics do not. So
given you want people to engage in a behavior, why would you not use
behavior to target campaigns? This question we explore in the
Also, I didn't find anything particularly new or relevant in the trades this
month, so we're going to skip the article links for May.
OK? Let's do some Drillin'!
Questions from Fellow Drillers
Behavioral versus Demographic Data
Q: Just finished my print out version of the latest
Drilling Down newsletter, and came across what is probably your best
quote ever: "You should be really most interested in what people
do and why, rather than who they are, because behavior predicts
behavior, and demographics do not".
A: "Print out" version? Are you
implying my newsletter is too long? You're not alone... :0
Q: Man !... I'm having the design department make a big
banner and hang it next to the web analytics team cubicles...
A: My favorite story on this issue: for years we
thought the "best buyer demo" at Home Shopping Network was
affluent women 50+. I mean, you hear their voices on TV, you see
their letters, you just know, right? Then we did an enhancement
of the database with what was then the most comprehensive and powerful
demo package available. And it didn't look right, there were
"too many young people". So we rejected it.
Then we started doing "best buyer focus groups" and had a
"best buyer advisory panel". Guess what? Only
about half were affluent women 50+. So we finally got the
message. It ends up the behavioral trait all best buyers had was
**they watch a lot of TV**. It so happens as a coincidence
that a lot of these people are affluent women 50+, but that's a
LifeStage thing, it's not "behavior". And it is not
*predictive* of being a best buyer, only *suggestive* of being a
best buyer. This means for acquisition purposes, we would be
better off to go with heavy TV viewers than affluent women 50+.
Q: This has profound impact; I mean, I know you know
that thousands of marketers will see this as heretical and just insane
A: It's funny, this idea is finally gaining some
traction. It caused a huge uproar in a discussion group I am in
composed of mostly agencies and media branding / exposure /
impressions types. They almost killed me. But I explained
to them this "behavior versus demographics" idea doesn't mean what they do has no
meaning or is not legitimate, it is just what it is, as far as they can take
it, without any kind of behavioral data.
These people are in the "reach" business, as in "200,000 men
25-54 saw the ad at least 3x". Fine. Nothing wrong
with that. But that's where it ends, it doesn't mean any of
these people "did anything" as a result of that reach,
regardless if they are the target demo or not. That
"branding" business is about nameless, faceless
impressions. Generally, and specifically offline, you have no
idea which individual people received these impressions.
Behavior is about individuals with specifically known behavioral
characteristics and prior actions. The person most likely to make a purchase
online is the person who has made a prior purchase online - regardless
of their demo. Behavior predicts behavior.
Q: I am presently in the middle of such a debate with
one of my clients; I'm trying to make him understand that we would be
way better off segmenting his site visitors by usage patterns than
trying to connect arcane psycho-socio-demographics characteristics to
their web site usage.
A: Ask the site owner what he wants as an outcome or
goal for the site. Does he want to "sell impressions",
or does he want people to take action, engage in "behavior"?
Also, the site owner should keep in mind psycho-demo stuff
self-reported online is likely poor quality.
Not that the psycho-socio-demographics thing is worthless -
especially when you can't get to behavior, there is no way you can get
your hands on a list of people with a known desired behavior.
Then you have no choice, and something is probably better than
nothing, as long as it's not too expensive.
Q: People just don't seem to understand this concept
because marketing has always been focused on demos. How do you
convince people that the demo data don't really matter?
A: It's not that the psycho-socio-demo data doesn't
matter; it's a question of is there anything better and more suited to
the goals of the marketing program in question; this goes just as much
for online as offline.
Think of it this way. On the web, there are two macro bodies of
knowledge, "who visitors are", and "what visitors
do". "Who visitors are" I'm sure is fascinating
and that's all about the panels and surveys and all the stuff many
marketers, especially offline, care about.
Branding and impression campaign managers are usually more
interested in "Who visitors are", goal-oriented campaign
managers are more interested in "What visitors do".
Some brand-oriented people may care about both data sets, but I don't
think many goal-oriented folks care about "Who visitors are"
in a demographic sense, because that is irrelevant.
Here is why.
Visitors enter the site from specific sources, start the visit with
specific pages, follow specific paths, and either accomplish or don't
accomplish specific goals. The source of the visitor (including
content / placement of the ad, search engine / search phrase used
etc.) and the content of the entry page largely *predict* if the
visitor will accomplish site goals or not. So for behavioral
purposes, "who visitors are" is really defined by source
analysis - they are from Google or from MSN, and they clicked on this
ad with this copy - not age / income / likes / dislikes etc. It's
segmentation by behavior, as opposed to demos. If what you want
in the end is behavior, you segment by behavior. It's that
That may seem like precious little information to brand / media
marketers, but it is predictive of likelihood to accomplish goal, and
that is what matters if you are trying to optimize a web site's
contribution to a business model.
You often find, for example, that visitors from web sites that are
clearly *not* "in the demo" are the highest converting
visitors, and vice versa. Demos may be *suggestive* (many of our
best customers are women over 50) but they are not *predictive*
(people who watch a lot of TV are our best customers, and women over
50 in general happen to watch a lot of TV). Behavior predicts
behavior, demos don't.
For example, many people read Hot Rod magazine and never buy
anything advertised in it, but some people do. This as opposed
to a list of known "hot rod" merchandise buyers, regardless
of whether the product has ever been advertised in Hot Rod
magazine. So here again, you get to a "suggestive"
versus "predictive" kind of thing.
You can think of these data sets as a "pyramid" with
increasing power to target. Demographics are generally at the
bottom; least powerful but largest "reach" because the
definition is so generic. Above that, you have socioeconomic,
which is really (to me, anyway) a "finer cut" on
demographics, the same data used in a different way. Then you
have psychographic, which is usually created by modeling data from the
previous two plus additional sources with "Life Style",
"Values", or "Attitude" implications. From
the bottom up, each is more powerful in "suggestive"
capability because you are using more data.
And each of them is useful if you don't have or can't get anything
else to target with. If you can get behavior, which is usually
in the form of a list of people who are "known to have taken
action X", you get to prediction of likelihood to "take
action X again", because there are direct correlations between
current behavior and past behavior.
To be clear, "Reads Hot Rod magazine" can be
predictive - it predicts they are likely to read similar magazines,
because reading magazines is a behavior. But this behavior does
not translate directly to other behaviors, it does not mean you are
more likely to buy related gear or go to the track when compared with
someone who has actually bought gear or gone to the track before.
And of course, the more Recently
you have engaged in a behavior, the more likely you are to engage in
it again for most behaviors.
It is true that you will "reach" fewer people using
behavioral targeting, but the question you should be asking yourself
is this: for your product, how important is it to "reach"
masses of people as opposed to communicate a relevant message to the
people likely to buy?
Q: How do you reconcile this "behaviorist"
view with the current interest in "personas" for designing
web sites to increase conversion? Aren't personas based on
A: Not as data providers define it, and I would argue that
using typical psychographic data to design personas is a faulty way to go
There are two issues in this area:
1. In the demo / psycho versus behavior issue above, we're
talking about audience definition and source - "I want to show my ad to Men
25 - 54"
versus "I want to show my ad to people Searching for hot rod
parts". This has little really to do with personas;
personas come into play after the audience has been selected.
Personas are about message, not audience. In other words, either
audience, "Men" or "Searchers", certainly contains
multiple personas. You could argue that ad copy may be written to be
reflective of a persona, but I think that approach would be suboptimal
- unless all your customers have the same persona, which is unlikely.
2. Much of psychographic data is implied by things like what
magazine subscriptions you have, what clubs you belong to, stores you
shop in, sometimes combined with zip code in the case of models like
PRIZM - audience segmentations like "Shotguns and Pickup
Trucks" or "Furs and Station Wagons".
Personas (at least the ones I think make sense to use) are a
psychological profile of a person based on a personality model like
Myers-Briggs - it's about how people think and buy. I don't
believe you can buy a list of people with a certain Myers-Briggs
profile or any other true psychological parameters; that would be a
massive violation of privacy. That's what I mean when I say
psychographic data is "implied": based on what kind of car
you drive, where you live, where you shop, etc, certain conclusions
are drawn about your "Attitude".
True personality models are
predictive of the approach someone takes to interacting with the
or "Who you are" data is fundamentally different from "How
you think" as a personality, and in the case of personas, the way
you make decisions.
In other words, engineers that read Hot Rod magazine probably make
buying decisions very differently than marketing execs who read the
same magazine, even for the same product!
If you are a consultant, agency, or software developer with clients
needing action-oriented customer intelligence or High ROI Customer
Marketing program designs, click
That's it for this month's edition of the Drilling Down newsletter.
If you like the newsletter, please forward it to a friend! Subscription instructions are top and bottom of this page.
Any comments on the newsletter (it's too long, too short, topic
suggestions, etc.) please send them right along to me, along with any
other questions on customer Valuation, Retention, Loyalty, and
'Til next time, keep Drilling Down!
- Jim Novo
Copyright 2006, The Drilling Down Project by Jim Novo. All
rights reserved. You are free to use material from this
newsletter in whole or in part as long as you include complete
credits, including live web site link and e-mail link. Please
tell me where the material will appear.
What would you like to
the book with Free customer scoring software at:
Out Specifically What is in the Book
Marketing Models and Metrics (site article