Johnson & Johnson uses Machine Learning to find and hire highly qualified candidates

Machine Learning - Human Ressources

Finding qualified employees is very hard...

Qualified candidates are very difficult to find. This is particularly true when it comes to specialised roles in engineering, research and other technical fields, where there are less candidates than open opportunities in the job market.
For Johnson & Johnson, their career website is their "front door" and a very important tool to attract the best talent. 


But for candidates, finding the right job is even harder !

This article was written inspired by the original story published here.
Suppose a highly qualified PhD in London is looking for a job in allergies research and by chance, he makes it to Johnson & Johnson career's website. He initially starts by typing key words like "allergies", "research" or "London", but each time the search engine only provides job results that match these specific keywords on the job title or the job description.

  • Johnson & Johnson is indeed the manufacturer of Zyrtec, an allergy medicine which could potentially interest the candidate, but on the keyword "Allergies", the search engine also returned hundreds of non relevant sales and marketing jobs related to this product.
  • On the keyword "Research", the search engine returned thousands of non relevant jobs, simply because all of them contained a generic text like "Johnson & Johnson is the world leader in research".
  • On the keyword "London", all non relevant jobs in London were returned.

After a couple of minutes of frustrated search, the person loses his patience and left the website. Johnson & Johnson just lost a potentially highly qualified employee. 
The sad part is that there was indeed a great job opportunity labeled as "Zyrtec science liaison EMEA" that the candidate simply failed to see.


​The problem with keyword search and internal jargon

The right candidate failed to see an existing opportunity that matched his skill set. Why did this happen?
Keyword search
Relying on search by keywords is a very unreliable method to build a search engine. Besides, a search for "allergy" will not match a text with the word "allergies". Technically, both words are different and without Machine Learning, there is no way for a computer to know that both words are the same.

Internal Jargon
What may be obvious to someone, may not be to someone else. Company folks know that "EMEA" is the acronyme of "Europe, Middle-East and Africa" and that by definition, this includes UK and all of London. They are also familiar with the term "Science liaison", as this is the way they call their researchers in a specific field. However, candidates are not necessarily familiar with any of these terms and acronyms and they may simply fail to see the opportunity in front of them.

Gladly, we live in the future and there is an incredible Machine Learning model "as a service" developed by Google, that tackles this specific issue and radically improve job search engines... Introducing Google Cloud Jobs Discovery Api !




Convert more candidates !


Google Cloud Jobs Discovery is a plug and play web solution (api) that allowed Johnson & Johnson to train and implement a machine learning model that better understands the company jargon and classifies job opportunities in their right categories. By using this api, the job also returns more relevant results based on:

  • Seniority alignment (it it very different to search for "manager" and to search for "assistant to the manager")
  • Query broadening (automatically expands results to similarly suitable positions)
  • Spelling correction (in case there is a spelling error on the query)
  • Concept recognition (to separate "server" results, as in "waiter", from "Linux server admin")
  • etc

This is an incredibly potent ML model that allowed Johnson and Johnson to increase high candidate conversion by 41%.

MLab, the Machine Learning specialists at your service!


If Machine Learning inspires you and you think you would like to implement a use case in your organisation, please contact us. We are independant and we will recommend and integrate the technology that adapts the best to your needs. If available technologies don't satisfy your needs, we can always develop a custom model tailored to your project.

Best of all, you would be surprised to learn the implementation cost of such a solution :-)


DisclaimerMLab was not directly involved in the development of this project. We simply publish this case study as a source of inspiration on what Machine Learning can achieve.