Analysing the Profiles of Freelancers from Nepal

This past week I ran an analysis on the freelancers from Nepal and what follows are the findings of the same.

For the purpose of the analysis, only one of the freelancing website was selected – Elance.com. I created a scraper to scrape the contents of the listings on Elance’s website and did some data cleansing to remove NaNs and empty values. There were about 4000 freelancers from Nepal listed on Elance, and the scraper which went on for about 8 hours with a 5 minute break in between each page listing, was successful in grabbing all of those data. The analysis was performed on python using pandas, numpy, seaborn, and NLTK library.

The fields on the dataset are Level, Rate, Rating, Number of jobs (in the last 12 months), and Skills. Higher levels indicate greater activity on the site. Rate stands for the amount the user expects to be paid hourly. Higher rating indicates greater satisfaction of the client.

Q. What level do most of the freelancers fall in?

Undoubtedly, most of the freelancers fall in the level 1. Out of ~4000 freelancers, only 576 of them are in a level greater than 1. This shows that most of the freelancers are starters (as everyone starts with a level of 1).

Q. What rating do most of the freelancers have?

You start in the site with a rating of zero. So, it’d seem that most of the freelancers have a 0 rating. But, that will result in a bias. The number of users who have joined but have never done any job (hence a rating of 0), will overshadow the facts about the ones that actually did the job. So, lets remove all the entries with zero ratings, and look at what’s the next popular rating.

It turns out that 60 (which is equivalent to 5 stars) is the most common rating. Here, the higher the number, the better is the rating. So, amongst those that have been rated, most have been rated 5/5.

We removed the entries with zero rating for this analysis. Let’s repeat our first analysis of the level of the freelancers by removing the entries with a zero rating (assuming of course that, freelancers with a rating of 0 might not have done any job to begin with.)

We now notice a longer tail in the diagram. So, there are people in the 13th level as well.

Q. What’s the hourly pay most of the freelancers seem to look for?

Most of the freelancers (with a non-zero rating)  seem to work for $11 per  hour. $5 per hour seems to be a close second. There are people accepting a pay as low as $3 per hour (46 of them) and as high as $110 (2 of them).

Now, let us see the group of freelancers that have a certain rate. The rates have also been binned into 10 groups.

Most of the freelancers with the rate in the range of $2.9 – $13.7 fall in the first level (No surprises there I guess, since we saw before that most of the freelancers fell in the first level anyway.) From level 7 however, more freelancers seem to have a rate of $13.7 – $24.4. Freelancers in the 13th level even have a rate in the range of $56.5 – $67.2.

Enough with the rate, rating and level. Now lets look into the skills of these freelancers. Remember I had mentioned that there’s also a Skills column in the dataset?

Q. What’s the most common skill among the freelancers?

Okay, now let’s look at the collocations (two or more words that occur together).

adobe photoshop; computer skills; microsoft excel; search engine; 
microsoft word; data entry; engine optimization; content writing; 
microsoft powerpoint; adobe illustrator; mysql administration; 
social media; web design; excel microsoft; creative writing; 
.net framework; sql server; online writing; adobe dreamweaver; customer service

These seem to be the common skills among the freelancers.

Q. Then, what are some uncommon skills?
Some of the uncommon skills are :

mentalray; altium; typescript; appcelerator; cryptography; absynth; matte-painting;
microprogramming; fortran; cobol; calligraphy; coldfusion; silverlight; zimbra;
mathematica; copyediting; openvz; capistrano; arcscene; whmcs; vb6; docker

We saw that there are freelancers accepting pay less than $5 per hour. Let’s see if there are any differences in the skills that they possess compared to that of the high earners.

Q. What are the skills that the ones with a rate of $5 or less possess?

WordPress seems to be the most common skill among the freelancers with a rate of $5 or below. Note that although earlier we analyzed among all the freelancers on the dataset (whether or not they had a rating), this analysis is only based on freelancers who actually have a rating, and are accepting a pay of $5 / hr.

And the collocations of the same are:

adobe photoshop; data entry; computer skills; microsoft excel; microsoft word;
search engine; engine optimization; internet research; microsoft powerpoint;
content writing; adobe illustrator; facebook marketing; .net framework; web design;
excel microsoft; mysql administration; administrative support; word microsoft;
adobe indesign; logo design

Q. What about people with a rate of $15 or more?

Again, analysing freelancers who have actually been rated, and have a rate of $15 and higher:

WordPress still seems to be the most common skill.

Let’s look at the collocations:

adobe photoshop; search engine; engine optimization; computer skills;
mysql administration; ecommerce consulting; microsoft excel; app
development; linux system; microsoft word; .net framework; sql server;
content writing; twitter bootstrap; adobe illustrator; excel
microsoft; iphone app; system administration; server programming;
active directory

There’s certainly a difference in the collocations (and hence in the skill set of the earners). Namely, ecommerce consulting, app development, linux system, sql server, twitter bootstrap, iphone app, system administration, server programming and active directory seem to be the skills that most of the freelancers with a rate of $15 possess (but those with a rate of $5 do not). Things like WordPress, SEO, Photoshop, and office package related skills seem to be common to both the group.

Thank you for reading! Hope you enjoyed it.