NONE: Re: ONLINE-ADS>> Online marketing research dilemmas
Re: ONLINE-ADS>> Online marketing research dilemmas
TLee_at_webcmo.com
Fri, 19 Jun 1998 11:26:52 -0700
Richard and the list:
Since I started the discussion of this thread, I think I
need to say something more.
In my last post, my major points are: 1. Most marketing
research firms don't have sufficient analytical capability
to conduct sophisticated marketing researches; 2. Most
surveys conducted online should not be called marketing
research because they did not do any MARKETING RESEARCH
and help web marketers make better decisions; 3. (less
important) Self-selection is not the reason that we can
not conduct a good online marketing research.
To my first statement, please look around, except
something like 'how many online' or revenue or ad
expenditure projections, what else have you seen? Except
those cross tabulations and frequencies, what other
analytical techniques have you seen used in the analyses?
To those who could not master sophisticated analytical
techniques, these statistical techniques are very
important because if you don't know them, how could you
design a good marketing research study? This is the first
tragedy of online marketing research. However, once you
are able to master these statistical techniques, they
become less important or even unimportant because they are
only tools helping you achieve your marketing goal. Some
research firms (few, because only few companies are
capable of doing this.) sell their techniques (models,
such as loyalty model) instead of their marketing
research. This is the second tragedy because marketing
research is not about statistic models but marketing
solutions. There are others who do not master these
techniques but use them anyway. This is the third tragedy.
One good example of this third tragedy is: multiple
regression is widely used in customer satisfaction study.
However, it is the most useless technique in CS study.
Why? Because people usually use it for two purposes:
determine the relative importance of each attribute and
make predictions. For the first purpose, we know that
there is almost always sever multicollinearity in CS
survey data and it makes the multiple regression results
unstable and unreliable. More important, there are better
ways to analyze how important each attribute is, such as
squared correlation, squared partial correlation and , if
you really want to show your analytical muscle, average
squared partial correlation developed by a Statistic
professor at University of Chicago. How about prediction?
Heck! It is just another statistic game. (We will have an
article "Analysis of CS study analytical techniques (II)
discussing the above issue in detail. It will be in our
"Review of Customer Satisfaction Study" column.)
I would like to make another statement here: if you folks
find a marketing research firm using simple correlation
(instead of squared correlation) and multiple regression
in its customer satisfaction study or site evaluation
study, you have another evidence to distrust its
analytical capability.
Last time I said that if you don't see a research firm
using Causal Path Analysis in its customer satisfaction or
site evaluation study, you got to pay attention to its
analytical capability. Why? Because this Causal Path chart
shows all about such a study (Lets ignore whether such
study is useful.). If you are able to do it but haven't
done it, please do it ASAP. (Try to ask your clients
whether they like it to be included in your analysis.) If
you are unable to do it, how can you conduct a GOOD CS or
site evaluation study?
The second and third statements are related. A marketing
research is about a research for marketing. It is not a
survey, it is not about statistics. (Of cause, you have to
conduct a survey and have to master statistical
techniques.) However, when people talk about a research
design, they spend MOST of their efforts on sampling
issue. It seems that once the sample is random, you should
have a perfect result. It is simply incorrect!
For example, a research firm conducts a customer
satisfaction survey using perfect sampling method
(supposedly) with 1000 respondents. The purpose is to
understand how to retain customers. It uses all the
techniques necessary (including Causal Path analysis).
Perfect sampling and perfect analysis. Can you design
strategy to improve your customer retention based on such
survey information? NO! (Please read our article:
"Questionable CS study questionnaire' in about two weeks.)
On the opposite, another firm interviews 10 dissatisfied
customers and try to understand why they were
dissatisfied. The sample is absolutely biased and the
analytical method? Absolutely non-technical! But it might
get more information from it for customer retention
strategy decisions.
If you ask the wrong question, you will never get the
correct answers. (That's why I said that more than 95% of
the error in a survey comes from asking the wrong
questions.)
First you see a tree, then a forest, then a tree in a
forest. As a marketing researcher, first you have to
master the analytical techniques, then you have to forget
them before you use them.
Pop-up survey method is very popular currently. But it
cannot solve the self-selection problem, it cannot improve
the sample randomness: your visitors come to your site
randomly. Pick up the Nth sample from a random sample
won't make the sample more random. More important, if a
pop-up window let the visitors know that there is a survey
waiting for them, they will pay much different attention
to your site. It leads much more sever bias in a survey.
We do include a pop-up window in our free web marketing
research software. But it is used to increase the exposure
of the survey.
Please point out if we have anything wrong in the post. We
are merely some marketing researchers who cannot suffer
from the current situation of online marketing research.
We are not Internet Survey Experts. (It is interesting to
see that 'Internet Survey Experts' can be used as a
trademark.)
Can someone help us register 'Online Advertising Masters'
as our trademark?
Cheers!
Tim Lee
Director of Research
http://www.WebCMO.com
A site dedicated to web marketing research.
SiteSurv----free web survey software to improve your web site continuously.
========================================================================
------------------------------------------------------------------------
This week's Online Advertising Discussion List sponsors:
ICONOCAST, Larry Chase and ChannelSeven.com
"This continues to be the 'coolest' and most informative e-mail
newsletter I get." Randy Haldeman, Director of Marketing, TEN-TV
To subscribe, visit http://www.iconocast.com
---
Free access for one year to Web Digest For Marketers archives
when you buy "Essential Business Tactics for the Net," by Larry Chase.
http://LarryChase.com
---
Subscribe to the NEW ChannelSeven Newsletter - News, Ad Case Studies,
Mrkt Research, Turbo-Ad Resources, Event Info & People-On-The-Move...
EMAIL: ch7-news-request_at_list.channelseven.com with "SUBSCRIBE" in subj.
------------------------------------------------------------------------
========================================================================
Online Advertising Discussion List To Unsubscribe send UNSUBSCRIBE
http://www.o-a.com/ to online-ads-request_at_o-a.com
