Archive for July, 2012


Zaltman Metaphor Elicitation Technique

July 22, 2012

If you recall, Señor Lopez Lomong, last year we had some back and forth about curation.  Courtesy of a curator I’ve been following recently, I came across a recent article about curation especially in the Tumblr and Pinterest context.

“Someone on Pinterest once posted a slide that read: ‘Pinterest: Where women go to plan imaginary weddings, dress children that don’t exist and decorate homes we can’t afford.’ But to focus on the ‘aspirational’ aspect is to miss the point. People don’t post stuff because they wish they owned it, but because they think they are it, and they long to be understood, which is different.”

You talked about surveys in your motivational misalignment post, but conflated them with observational data.  In my view, surveys are very different from observational data.  In my reckoning, surveys are explicitly carried out with explicit question-asking and elicitation, whereas observational studies try to implicitly elicit things just by watching things unfold as they normally do, with their inherent sample selection biases which should then be corrected.  (In fact, the Anderson and Oliver (1987) paper I mentioned is empirically validated through surveys by Cravens, Ingram, LaForge, and Young (1993), whereas in this day of social media data and enterprise data collection, a validation based on that observational data could be attempted.)

The Zaltman metaphor elicitation technique is a survey method of sorts that tries to elicit consumer insight not from words, but from images. To understand the thoughts and feelings of customers about products or brands, Zaltman says that image-based elicitation is more effective than text-based elicitation. Rather than filling out a question-answering survey, a customer is tasked with taking several photographs representing their feelings on a product; researchers then interpret those photographs metaphorically in conjunction with conversations with the customer to draw conclusions on the voice of the customer.  This is quite a costly and labor-intensive process.

Instead of the explicit photograph-taking to represent feelings on products, wouldn’t it be great if there were a ready source for such images that could be analyzed as observational data?  Oh right, there now is: Pinterest.  How could a machine analyze the images instead of a researcher?  Couldn’t Torralba, Fergus, and Freeman’s 80 Million Tiny Images approach be used to find similar images and collect all the text surrounding and associated with the similar images, with that text information then being further analyzed by a machine good at natural language and evidence-based learning?

As commented upon by Wolff,

“At the heart of the Internet business is one of the great business fallacies of our time: that the Web, with all its targeting abilities, can be a more efficient, and hence more profitable, advertising medium than traditional media. Facebook, with its 900 million users, its valuation of around $60 billion (as of early June), and a business derived primarily from fairly traditional online advertising, is now at the heart of the heart of this fallacy.

“As Facebook gluts an already glutted market, the fallacy of the Web as a profitable ad medium will become hard to ignore. The crash will come. And Facebook—that putative transformer of worlds, which is, in reality, only an ad-driven site—will fall with everybody else.

“Facebook has the scale, the platform, and the brand to be the new Google. It lacks only the big idea. Right now, it doesn’t actually know how to embed its usefulness into world commerce (or even, really, what its usefulness is).”

He further states that the big idea will have to be something around the knowledge about people found inside the observational data that Facebook possesses.  I couldn’t agree more.  Along the same lines, the big idea for Pinterest should be image-based elicitation for consumer insight.


Salesforce Control Systems

July 17, 2012

Varay chiva Señor Aboutrika?  I hope you have a fulfilling journey to the roof of the world. I’m glad that some discussion about causal inference will be of interest to you, but before that, I wanted to reply to your comments on motivational misalignment, specifically from the perspective of some material I’ve been reading recently on salesforce control systems.  (By the way, I will probably stay away from Pearl’s perspective on causal inference and stick to Rubin’s; there seems to be conflict between the two perspectives.)

One passage from the paper “Perspectives on Behavior-Based Versus Outcome-Based Salesforce Control Systems” by Anderson and Oliver (1987) on cognitive evaluation theory reads,

“Rewards are thought to mediate judgements of self-determination and competence by virtue of the informational or controlling nature of the feedback they provide.  If feedback is interpreted as providing information relevant to improving one’s performance and competence, internal locus ascriptions are enhanced and intrinsic motivation increases.  Alternatively, if feedback is functionally similar to a control system, intrinsic motivation is decreased.”

Another passage from the same paper discussing agency theory is as follows:

“Agency theory is an analytic, normative microeconomics/accounting approach to the question of how principals can control the activities of the agents to whom they delegate decision-making authority.  A central premise of this theory is that principals and agents have divergent goals.  Agency theory is concerned with the design of control systems that realign the incentives of both principal and agent so that both parties desire the same outcome (‘incentive compatibility’).”

The overall discussion in the paper is about two diametric views of setting up systems to manage (control) salespeople: behavior-based and outcome-based.  In behavior-based systems, it is about measuring and rewarding the process of selling including things like quality of presentations, services performed, number of calls made, knowledge of products, and overall pleasantness.  (Purely such a system has salary-based compensation only with a subjective performance rating methodology and there is quite a bit of coaching.)  In outcome-based systems, it is about measuring and rewarding the end result of selling such as total revenue or total profit.  (In a purely outcome-based system, compensation is fully on commission and the salespeople are left to fend for themselves.) 

There are a few factors that go into which system is preferred.  One factor is the relationship between sales effort and revenue: is it almost deterministic or wildly uncertain?  Another factor is how costly it is to measure behaviors and outcomes; measuring outcomes may be costly if, for example, team-based selling is practiced widely.  A third factor is how risk-averse and intrinsically-motivated the sellers are.  A behavior-based system is called for if there is great uncertainty between effort and performance, if measuring outcomes is costlier than measuring behaviors, and if the salespeople are risk-averse and intrinsically-motivated.  An outcome-based system should be chosen for the opposite case, and also for very small salesforces.  Mixtures are also possible.

This paper was written in the 1980s, well before social media could even be imagined.  These days, selling is becoming much more based on social media, and I think that makes measuring behaviors and providing informative rather than controlling feedback much easier.

Anderson and Oliver wrap up by saying that “the more a control system is behavior-based than outcome-based,” the more professionally competent, loyal, committed, cooperative, intrinsically motivated, risk-averse, and thoughtful the salespeople will be and have a more planned and low pressure style.  Sellers in behavior-based systems will put the firm first, customers and principals second, and the self last.  However, they will also have worse short-term individual-level performance measured in traditional ways like revenue.

J. G. Arora writes in Bhagvad Gita for Supreme Success: “For Gita, real worship consists of doing one’s duty with perfection without being distracted by thoughts of outcome of our action. It liberates us from all bondages, doubts, self-imposed limitations, anxieties and fears, and enables us to lead happier, fuller, contented, peaceful, blissful and supremely successful life.”

I’ve been thinking about religion being a control system for several months, where the best or ‘optimal’ religion depends on the characteristics of the society in which it is to be used.  In the framework discussed above, it should be possible to characterize different religions and philosophies as more or less outcome-based.  I would imagine that it is not costly at all for gods and other such heavenly beings to measure people’s behavior.  The level of uncertainty mediating actions and outcomes is different in different societies.  But let me leave this line of thinking to wiser folks like Nanaji.  Please say Ram Hari to him for me now that you’ve landed in North India.