---

MSN Commercial Intent

Just click to go to the bottom of this page if you are here just to see the next video of the Micro Niche Finder tutorial.

Commercial Intention

It doesn't matter how many people you entice to your pages if they are not interested in what you are offering. This is particularly true if  it is costing you money to bring them.

Microsoft has (or at least, had) a page to help you determine the online commercial intent" of the people searching with a particular keyword phrase. ("Had?" It has not been available for over a month as of May 20, 2011. It may be gone for good. Anyway, to assume it's coming back...) You can go there and access the information for free; albeit, you have to copy and paste the phrase into their page and copy and paste the result back into your spreadsheet. You can also ask about the intention of people visiting a particular web page.

The source for this information is the Microsoft Online Commercial Intention page:

What does online commercial intention indicate? An intention of the person using the query immediately to buy, sell, or "commit to some commercial service." The person who is gathering information is not showing commercial intent, even if it will later lead to a commercial decision.

How does the fraction you get on the Microsoft Online Commercial Intention page indicate the commercial intention? I'll give you the basic idea. If you wish, you can read the journal paper by Kathy Dai (and others) , "Detecting Online Commercial Intention," IW3C2, WWW 2006, May 23-26, Edinburgh, Scotland, ACM 1-59593-323-9/06/0005 and a brief report on another test of the algorithm at  Algorithmic Commercial Intent Detection of Search Queries.

Kathy Dai devised an algorithm to classify web pages and search queries as generally commercial or noncommercial in intent.  Her intention was to produce a strictly binary classification, one or the other, but the algorithm produces a fraction between zero and one, zero meaning non-commercial and one meaning commercial. 

The fraction does not indicate the fraction of people searching with commercial intent. It does not indicate a probability. It simply indicates a non-statistical confidence that the classification is correct. The closer the fraction is to 1.0, the more confident you can be that the keyword in a query indicates commercial intent; the closer to zero, the more confident the query is only for information. The closer to 0.5, the less confidence you can have one way or the other.

She used a learning algorithm and trained it against a data set independently classified by several people.

Where the algorithm and the majority of humans agreed the query should be classified as showing commercial intent, the average value returned  by the algorithm was 0.83. Where they agreed that the query showed no commercial intent, the average score was 0.20.

The algorithm for classifying keywords depends on that algorithm for classifying web pages. The algorithm queries a search engine with the keyword and (depending on the version of the algorithm), combines the classifications of the web pages returned on page one of the results, or just classifies the commercial intent of page one of the results directly. (Since the results page contains snippets out of the web ages that the search engine judges to be most relevant, page one of the results is pretty representative of the overall collection of pages.)

These are the basic lessons:

1. DO NOT THINK OF THE FRACTION AS A PERCENTAGE OR PROBABILITY. It does not make any sense to multiply the number of searches by the fraction to estimate the number of searchers who want to buy.
2. Think of the fraction just as a non-statistical level of confidence in an indication of commercial intent: the closer to 1.0 or 0.0, the more you can feel confident the keyword is or is not used with commercial intent. 
3. Do not expect values much less than 0.17 or more than 0.85.
4. Divide up the range, picking your own boundaries if you wish:
• Above 0.75, it's pretty likely to have commercial intent.
• Below 0.25, it's pretty likely to have non-commercial intent.
• Close to 0.5, it's a toss-up.

There are some conceptual problems with this tool:

• It assumes a binary classification, rather than a probability. A probability is more realistic, and more valuable. (But how would you ever test that?)
• The algorithm was trained to reproduce the classifications made by humans, and the humans used do not constitute a good statistical sample, nor is there evidence they are experts in the domain.
• The algorithm for classifying keywords examines the first page returned by a search engine. It could be just feeding us the biases and assumptions of the search engine, not the people searching.

Nevertheless, this is probably the best thing we have to go on.

Micro Niche Software Review Part3 - Internet Market Research

This video shows how to access Microsoft's Online Commercial Intent as one of Micro Niche Finder's integrated Internet marketing research tools. The higher the fraction, the better the keyword will be for article marketing. 

gigglypaws
November 24, 2008http://www.reviews.digitalmindfood.com/micronichetool.htm  (last I looked, this link was broken)


The next video shows another of Micro Niche Finder's integrated Internet marketing research tools, a keyword competition tool, which computes a Strength of Competition score.