Example: at the beginning of the time period there was 1 Drupal posting, and now there are 3751. Each increase of 1 represents an additional 100% growth over the original, so you have 3750 * 100% = 375,000% growth.

A poor way to measure rapid growth, but that seems to be their system.

I think you're probably correct, but "poor" is much too generous a description of such a methodology. It rates at least an "awful", and "garbage" is not unwarranted.

Consider that it is only an accident of fate that there was merely 1 Drupal posting at time 0. If there had been two, I guess the entire trendline would be half as big.

Or consider this even wackier example:

http://www.indeed.com/jobtrends?q=ruby%2C+mongodb&l=&...

This goes to show that the algorithm is at least smart enough not to explode when dividing by zero. (There were, I'm pretty sure, zero mentions of "MongoDB" in job postings in 2005.) But, then, what does this graph measure? What are we dividing by for MongoDB? The number of postings at some arbitrary later time?

What's the proper way to do this?

How are fluctuations calculated if the reference is the start point? Could these be 'deleted' posts?

Exactly. It would be useful if you could set the starting point of the period you're interested in.

and why does Django look massive here:

http://www.indeed.com/jobtrends?q=ruby%2C+java%2C+python%2C+...

but not here:

http://www.indeed.com/jobtrends?q=ruby%2C+django&l=

?

One is absolute and one is relative.

doh, thanks

:) well spotted