Technologists and Customers are Partners

For some reason, I've been intrigued by some posts made on a blog called vcastprofiles. I don't really recall how I stumbled upon this blog, but I manged to digest enough of the content (the text reminds me of the phrase 'eschew obfuscation') to understand that the author was commenting on an industry that I've been intimately involved with  for the past 5 years - mining social media for brand and product information.

As the author intends to give an impression of authority, I wondered why I hadn't visited here before. Digging a little deeper, I found that Alan Wilensky blogs at a number of places (vcastprofiles, Alan Wilensky's Weblog, and BlogWhine).

I get the impression from his write up on sentiment analysis that he has evaluated some solutions for a client and not been impressed. However, I'm trying to figure out which systems he has evaluated::

Good words, bad words, neutral words - blah…..all of the HSD’s used by the current ‘leaders’ (the term is used advisedly), are non-standard. Even test query sets run against the same data show non-repeatable results…the text scoring engines are not mature. Surprisingly, the Open Source tools are very robust (lingPipe for example.)

Alan has, I believe, a misunderstanding of the relationship between an emerging market and the early adopters that help sustain and evolve development of technologies that serve those markets. His writing suggests that he wants a fully baked solution to appear with no interaction or iterations with customers. This is simply not how technologies in spaces like this are matured. The technologists and customers are in partnership for mutual benefit. The technologists have to give some value and work with clients to understand where to spend effort. The customer has to find value in offerings that are imperfect and be a partner on the journey.

Comments

[Alan Wilensky]

I would never presume to refute THE leading light of the text mining industry, Matt Hurst, and I agree that all technologies that break new ground go through an adoption / refinement cycle.

That said, the six month study I conducted for France Telecom showed certain weaknesses that many end-user clients of text mining and brand monitoring services found noteworthy enough to mention to me, in the course of my survey. These folks have budgets that are able to encompass standard brand equity consulting from the industry giants, and the new-age text mining and brand monitoring services now emerging.

But more important, the trilogy of articles on my blog point to a more important and vital strategic issue beyond the current weaknesses of the state of the art - this is the fact that the practitioners of record for brand monitoring have missed the fat part of the market, the brand intermediaries, the mid-market, due to the fact that serving them is a more subtle and nuanced undertaking.

There is more to my observations than pointing out that sentiment analysis is weak, provides little fodder for creating decision support matrices, and misses the most subtle and important conversational topics that are the spring from which redress seeking and declarations of the consumer's perceptions of product performance and interaction outcomes gush forth from. ebay is pursuing this work now.

I invite the readers to skim the other articles where I spell this out and propose an outer architecture for detecting these dynamics.

And, I am sure that great scientists, like Dr. Hurst, will evolve the state of the art and make better tools. Certainly, that's what I and Dr. Charles Martin are pitching.

[Matthew Hurst]

Alan,

Can you share with us the vendor's that you reviewed? Was the study for english language data or french? Was it a brand based study or product/service?

Oh, and thanks for the sarcasm: refreshing to hear.

[Alan Wilensky]

Sarcasm, moi ? False modestly, et tu? You, Matt, are the current expert of record in the field. The market most in need is not being served.

You can probably guess that my NDA precludes verbatim elements of the study, but Karthik came to the lab. I have written about Cymfony, and how their acquisition by TNS is a repudiation of targeting text mining in order to mimic the dominant brand monitoring by the big boys.

And, I think that, yes, the strategic alignment of the products/services of the vendors I covered will go through a maturation and revelation cycle.

It was not meant to be a product vendor book, ala Gartner or Forrester. It was an (English Language)survey of the perceived applicability of such services:

1) Are the dashboards really all that sexy?

2) study of the latency of the setup of the CSA campaings (a real sticking point, btw).

3) Are the output of such surveys useful for steering decisions, or just interventions?

4) as I mentioned before, the market is missing the real opportunity. This article:

http://bizcast.typepad.com/clients/2007/06/a_colloquial_tr.html

and this paper:

http://bizcast.typepad.com/clients/2007/05/a_new_consumer_.html

and this:

http://bizcast.typepad.com/clients/2007/03/mining_the_blog.html

explain what we are getting at.

Again, it's not that Sentiment Scoring is bad, it is that the market most in need is not being served.