Machine Learning for Enterprise Recruiting
Machine Learning for Enterprise Recruiting
The advent of machine learning (ML) and artificial intelligence (AI) have ushered a whole new way of doing business in the talent recruitment industry.
Join the DZone community and get the full member experience.Join For Free
Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.
Today's post is the sixth of our blog post series authored by the speakers at the upcoming 2ML event. Each post presents a summary of the main topic that each speaker will be showcasing at 2ML on May 8-9. David J. Marcus, Sr. VP of Special Projects at PandoLogic shares below the role of Machine Learning (ML) in PandoLogic's proprietary campaign algorithms. For more details, we invite you to attend his session, "Machine Learning and AI-Enabled Job Advertising Platform for Enterprise Recruiting."
The advent of machine learning (ML) has ushered a whole new way of doing business in the talent recruitment industry. Traditionally, recruitment was done via two primary methods:
- Using a professional recruitment firm ("headhunters").
- Advertising the positions in local/national newspapers.
The cost of hiring via a recruitment firm was prohibitive (typical fees: 25% of the first-year salary!) — clearly not an option except for select high-caliber positions. The alternative, advertising in newspapers, was typically priced by the number of words in the ad or, more likely, by linear vertical print space (measured in inches in the U.S.). This encouraged short ads and lots of abbreviations to the point that some ads appeared almost inscrutable to the first-time job searchers and were certainly devoid of real content. On the plus side, an employer could advertise for blue-collar positions.
The explosive growth of the Internet brought a paradigm shift. Job sites like Monster.com, CareerBuilder.com, and others flourished by offering a very simple model: the employer paid a fixed fee (with volume discounts, of course) for each ad of unlimited length and content (text or text and graphics).
These job sites provided a rudimentary search capability (essentially a search for keywords) and then displayed a list of job postings that contained these keywords. The displayed list would consist of two to three summary lines for each posting. The order of display was paginated (15-25 job ads per page) with jobs appearing first. If a potential candidate liked the posting summary, they would click on the ad to see the full content. If they liked the job description, they would click on the 'Apply' button. That's it! Not much in the way of sophistication, but good enough to make Monster a multi-billion dollar company. For sure, it was a great alternative to job ads in print (newspapers). Now, candidates could search and apply for jobs, in full privacy.
A big downside to these simple job sites is the fact that new job postings were shown on the first pages of the results set. This meant that a job that was just several days old would begin to appear too far down on the paginated search result sets. Ask yourself: How many users ever get to page 17, where their perfect job might have shown up?
The next phenomenon to impact this industry was Google and its keyword bidding algorithm. In Google's ecosystem, advertisers compete for prominence in the search results. For job sites, this was a call to action since employers had learned very quickly that being on page 17 was virtually worthless. Also, job sites realized that bidding for prominence is a great way to increase revenue.
The modern job sites now charge by the click (Cost Per Click, abbreviated as CPC), with Indeed.com being a recognized leader. Now, when a person clicks on a job summary in the search results, the employer is charged a CPC amount. The bidding directly affects the positioning in the result set. Employers now have to bid a higher CPC to ensure their jobs get visibility.
At PandoLogic.com, we have developed an ML-driven algorithm that uses our vast historical data to predict (the word) the level of CPC required for a job posting to gain visibility on any vendor site that we work with. Using this prediction we can manage the full lifecycle of a basket of employer's jobs. This capability is transformative in the world of enterprise recruitment. At PandoLogic, we can now optimize an employer's budget to programmatically (dynamically) adjust the CPC we bid on for any given employer's job postings at each publisher's site on a basis. This significantly increases the budget efficiency, and significantly increases the job postings' visibility resulting in a higher number of job applicants, which is really the employers' main goal in the first place.
To be exact, our ML algorithms budget and optimize recruitment campaign spending in real-time by utilizing over 10 years' worth of historical job performance data containing nearly 200 billion data attributes. The models work by establishing predictive-performance benchmarks that drive when, where, and how each employer's job is dynamically campaigned online, in real-time. This automated process significantly improves efficiencies and maximizes the ROI attainable with the employers' budgets. We'll looking forward to share more details about our workflows on stage at 2ML.
Published at DZone with permission of Maria Jesus Alonso , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.