1 Introduction

We study E-Commerce job offers in a selected list of countries based on two data sets:

We limit our scope to Australia, Canada, China, Denmark, France, Germany, India, Italy, South Korea, Spain, United Kingdom, and United States.

The Glassdoor original data features only job titles containing the term “E-Commerce” or close variants, when we restrict it to our countries in scope we keep 7,770 observations.

The LinkedIn original data features only jobs offered in the countries in scope, when we restrict it to observations whose job title contains the term “E-Commerce” or close variants, and to locations that can be identified, we keep 17,551 observations.

Those two filtered data set were merged into a main dataset of 25,311 observations, to merge them we kept as many variables as possible, manually creating new variables for both datasets (GlassDoor: seniority and employment type; LinkedIn: sector) based on text matching of job titles and descriptions.

We display below a further breakdown of the data by country and language identified from job description.

country LinkedIn: English LinkedIn: Other GlassDoor: English GlassDoor: Other All: English All: Other
Australia 300 2 72 0 372 2
Canada 1468 95 352 72 1820 167
China 377 100 286 15 663 115
Denmark 61 26 15 22 76 48
France 54 2138 84 825 138 2963
Germany 252 3249 92 1011 344 4260
India 1231 6 473 1 1704 7
Italy 68 353 22 168 90 521
South Korea 28 18 11 1 39 19
Spain 160 357 26 56 186 413
United Kingdom 2880 2 1066 3 3946 5
United States 4293 3 3104 13 7397 16
All 11172 6349 5603 2187 16775 8536

The job descriptions are mainly in the English language, and the US have the most offers.

The two datasets, from Glassdoor and LinkedIn, don’t contain exactly the same data, we matched the data that could be matched, cleaned existing variables or reworked them into higher order categories, and extracted additional information from existing variables to build new ones, such as average salary (from the salary range), language (from the job description), or latitude and longitude (from the location)

The GlassDoor data contain information that are missing from the LinkedIn data such as estimated salary range, rating, employer, industry, and size (no. of employees), so some of the analysis will be restricted to the Glassdoor data.

Text mining will be restricted to English Speaking offers.

It is interesting to note that many offers are duplicated, where we define duplicated as featuring the same employer, job title and description. Some were posted the same day at the same location, some were posted on different days or/and different locations.

We decided to keep those in the analysis because two duplicated offers are still two offers in the market. The most duplicated ads were found in the LinkedIn dataset, where we found most notably 545 times a Walmart Canada ad for a job title simply described as “Ecommerce”, the second most duplicated ad was found 103 times. Overall 68% of ads are unique.

2 Company Info

2.1 Biggest advertisers

out of 25,311 global adds advertised by 9,217 distinct employers, we find that the biggest ones, as measured by the number of posted ads are Kroger and Walmart Canada, with respectively 1,192 and 994 ads each. These are ahead of other companies by a fair margin.

2.2 Size

The size of the company is often given in the Glassdoor data, when it is given, almost half of the ads are for jobs in companies of less than 200 employees.

2.3 Revenue

  • How much revenue do hiring companies make?

We counts unique companies per revenue class

  • How different are job descriptions from higher revenue companies to lower revenue companies

We compare companies with a revenue of $100M or more (1,555 ads) to companies with a revenue of $20M at most (1,569 ads). We then tokenize the descriptions and after removing stopwords, retrieve words that appear with (the most different frequencies, among words that constitute at least 0.2% of the corpus on each side.

On the chart below:

  • A value of 0 means that the word have the same representation on both sides.
  • A value of 1 (resp -1) means they’re twice more present in ads from higher (resp. lower) revenue company.
  • A value of 2 (resp -2) means they’re 4 times more present in ads from higher (resp. lower) revenue company

We see that higher revenue companies are much more preoccupied by issues of diversity, while lower revenue companies are much more pragmatic, discussing remote work, schedule, pay, hour etc.

2.4 Sector/Industry

  • What is the distribution of sector and industry among job offers ?

country sector n pct
Canada Retail 99 35.7%
Canada Business Services 53 19.1%
Canada Manufacturing 51 18.4%
Canada Information Technology 38 13.7%
China Business Services 71 80.7%
China Retail 4 4.5%
China Media 3 3.4%
China Arts, Entertainment & Recreation 2 2.3%
France Business Services 142 28.0%
France Retail 105 20.7%
France Information Technology 94 18.5%
France Manufacturing 80 15.8%
Germany Retail 151 32.1%
Germany Business Services 112 23.8%
Germany Manufacturing 81 17.2%
Germany Information Technology 75 15.9%
United Kingdom Retail 229 36.2%
United Kingdom Business Services 167 26.4%
United Kingdom Manufacturing 92 14.6%
United Kingdom Information Technology 66 10.4%
United States Retail 650 32.6%
United States Manufacturing 472 23.7%
United States Business Services 375 18.8%
United States Information Technology 168 8.4%