Statistical Analysis of Search Results.

Because anecdotal records are not good enough.

[IITian]

For those of us who have not been turned off by the keyword 'statistical analysis' ;) :

I am interested in two major attributes:

1. Consistency: How does the result vary from day to day. Does my (and my competitors' ) sites fall on and off the search results, occupying #1 one day, and moving to #9999 the next day, or they remain fairly stable?

2. Relevance: Difficult to measure but we can compare acros major search engines. Does the same site occupy #1, #3, #1 on three of the major seach engines but #25 on fourth for the same keyword search? It will mean fourth is perhaps "off" either in a good way or bad.

Methodology

Obviously larger the sample size, better the conclusions. If some of us could keep daily accounts of our searches and combine them either after a few days, or interactively in a web-based manner, some conclusions could be drawn about the search engines/portals. (Note: I said daily, not 5 minutes-interval that many of us could be checking our rankings. ;) )

Data

Each day, we could keep record of the following:
For each of the keyword combination we are interested in most
- keyword
- for each relevant page (yours or competitors') its rank on different searches out of how many results, any comment regarding changes in page content, Google-banned etc.

and
- additional data like your experience as a SEO, size and number of websites, country of residence.
- comments - information that could be useful to others.

My thought process is still in the preliminary stage. If somebody has already thought along similar lines, or has some experience in this, I would like to hear from them. There is no reason why we couldn't implement something similar. It is going to help a lot of people too.

[amazed]

I doubt anything like this can be shared in really competitive areas.

I suppose every professional is doing it for their websites.

I myself would add to the list

refering sites
browsers used

Problem with the statistical approach is that you see what you have got, and how to continue on what already exists.

It doesn't tell you what you might get, being creative.

[IITian]

Good point there about professionals not wanting to share this info. Looks like some other information has to be used if one wants to do this study.

What if the information is annonymous and we can only look at the statistics but not the individual data components? Won't that make us more cooperative?

With stat analysis, at least we won't have so many speculative threads about SEs - there will be some data to back up the claims.