Careerjet is a job search engine that began as a bet by its founders, Thomas Busch and Jean-Benoit Andrieu.

It appears they have won - it now aggregates content from over 59,000 sites daily, and currently displays upwards of 21m vacancies.

We talked to Busch about the site’s development and the semantic technology behind its search functions.


What was your background before you started Careerjet?

Jean-Benoit and I come from strong engineering backgrounds: my background is in aerospace engineering, and Jean-Benoit’s is in physics and chemistry. I was working as an engineer for Yahoo! when we started Careerjet, and Jean-Benoit was working as a Quality Assurance Manager in plastics.


How did the idea come about?

In all honesty, it started off as a bet between three young professionals. At the time, around ’98, the online recruitment industry in France was just barely emerging, and so Jean-Benoit and I were coming up with all these ideas, one of which was a job site.

We were brainstorming with our friend, who maintained that we were all talk and no action, and so we bet him that we actually would follow through this time, as a fun project on the side. But then all of a sudden came the dotcom boom, and there were major job sites popping up everywhere.

We didn’t want to be competing with the crowd, so we figured we would use this rapidly saturating market to the best of our advantage by aggregating them all on one site — a job search engine.

When we started Careerjet, there was only one other site doing what we
did at the time, which was Flipdog, a US job search engine that no longer exists.


Where does CareerJet get its job listings from?

There are two facets to the site; crawling information from other sites, and then indexing the information into our system (and updating this information daily).

The crawling part of the system is completely automatic, as we use semantic analysis to extract job offer content (such as job titles and descriptions) from other sites, which is updated daily to reflect changes in the job offer or new ones added to a particular site.


How does Careerjet differentiate itself from other jobs/career websites?

Careerjet is a job search engine, and as such does not do the same business as job/career websites. Job sites provide services for recruiters to find candidates, whereas search engines like ours provide visibility to other job sites in order to improve their candidate base. We function as traffic-traders.


How big is the team behind the site?

We have a growing team of international web developers and translators
that is currently 15-strong.


How is Careerjet funded? Is it profitable yet?

Careerjet is completely self-funded. At the beginning, it was just two guys using whatever minimal leftover funds they had to start this side project. In Careerjet’s earliest days, we used the cheapest hosting system we could find: a shared server from the US for just $15/month.

However, after just a few months, we were overloading the server and they told us we had to start paying for our own. Nine years later, I’m happy to say that Careerjet’s profits continue to increase, and despite being approached by several venture capitalists, at the moment we are working on defining our future goals and maintaining its privately-funded status.


How many users does the site have worldwide? How many in the UK?

We currently have around 15m visits worldwide per month, and around 1m visits per month in the UK — but these numbers are still (happily) growing and growing.


Can you tell me about the semantic technology behind the classification of job listings?

We use semantic technology to perform the automatic crawling function of the site. Its role is to find information from websites that contain job offers and then extract exactly what we want from it: job titles, job location, and job descriptions.

We also use semantic analysis technology to improve the search functionality — for example, the German language is notorious for melding together several
words to make one long, complex one.

For complicated linguistic matters like these, we use semantic analysis technology to recognise the information, no matter what the language or construction, and form a coherent job listing out of it.

So basically, we use semantic analysis technology to standardise both the user (job seeker) queries and the job offer information placed on websites around the world so that our system is compatible with the user needs.


How accurate is it?

Using semantic analysis technology is like teaching language skills to children: at first they can’t speak very well, and then their language skills increase and their linguistic errors decrease.

So for now, our different sites are at different accuracy levels — Chinese is one of our newest sites, so its semantic analysis technology hasn’t yet adapted to the level of, say, our French and English sites which are at much more advanced and capable linguistic levels.

We’re currently operating in 25 languages, so the accuracy depends on how long it has been since the site launch for the technology to adapt and improve.


How do you promote the site?

We promote the site partially by SEO and partially by our new affiliate programme. We practice clean, non-abusive SEO, and our affiliate programmes include web publishers and freelance sales agents. Apart from our affiliate program, we don’t really use paid search to promote Careerjet yet.


Can you tell me about your affiliate program?

The affiliate programme shares our revenue with our partners (affiliates). We use a double affiliate programme: on one hand web publishers embed their sites with our content, and on the other hand freelance sales agents market our services to potential clients in other countries. 

Since it’s not feasible to be in as many countries as we have websites right now, we are relying on our worldwide sales agents to represent Careerjet on a global scale.


Do you have many publishers on board?

The publisher programme is one half of our affiliate programme, in which website administrators embed Careerjet content onto their sites, such as logos, banners, or JobBoxes. We currently have around 200 web publishers on board, but as it’s a pretty recent service the numbers are increasing.


You have recently expanded into Turkey and Russia, where next?

We are currently working on Arabic versions of our Middle East websites and an Israeli site in Hebrew.


Related posts:

Q&A: Prof. David Crystal and Sacha Carton on semantic targeting

Q&A: Indeed CEO Paul Forster