>I'm still testing HTML structure versus rankings. (takes a while to create pages,
>submit pages, track their performance...
Wish I had the will. However, having crunched a few direct marketing databases,
I was thinking:
1. Might be useful to load all the pages with a unique character string
so you can pull them all up at once for quick relative rankings,as
well as cascading through the search results for absolute rankings.
(Already done that, huh?)
2. Tests have a way of answering questions, but not of asking them.
For that reason I've always spent a bit of time on randomly searching
for patterns. Wonder what could be learned by:
Locating an industry that uses fairly simple pages with no frames and
Saving the code for the first 100.
Parsing them into a consistent field structure.
Importing them to a spreadsheet laid out as Title, Description,
Keywords, Hx, follow text, Hx, follow text, ...
Stripping the common elements such as "META"
Sorting the bejesus out of the records by every field.
(Yeh, a database is more versatile for sorting, but it's right
back to asking the questions I already have. Sorting spreadsheets
with an utterly blank mind always seems to work better.)