Thursday, January 28, 2016

Can Algorithms Make Job-Hunting Suck Less?

Dave Slater got his tech start-up job the old-fashioned way. Slater was at the Burning Man music festival. As he directed the assembly of an “art car” from cast-off materials, a woman approached him. She’d admired his management of the group effort, she said, and wondered what he did for a living.

Fast-forward a few months and the questioner is now Slater’s boss at Sourcery, a tech start-up she founded.

Out with the old

For generations, the saying “it’s who you know, not what you know,” has applied to job searching. And the adage isn’t wrong: labor economists say that at least one third of all jobs are found through family and friends.

But some Silicon Valley companies claim they can take the subjectivity out of the process, and improve outcomes for both individuals and hiring organizations with new platforms and algorithms.

The claimants are bold. One of the largest new firms, claims in some marketing material to match candidates with quality job offers in minutes. Another service, Anthology (formerly called Poachable), describes their mission with “We believe the world would be a better place if everyone were in their best fitting job all the time.”

For individuals the allure is simple– plug in a résumé, answer a few simple questions, and these sites will quickly pair you with potential employers interested in your skills even as you avoid time-consuming (and anxiety-inducing) networking and waiting.

And for jobseekers with well-defined in-demand skills it seems to work. Aaron Hsu, a mobile software engineer, signed up for Hired and matched with several prospective employers within the first week. He found a new job within a month. He describes the appeal as “They do most of the searching for you [when you] don’t want to spend too much time.” (Both Hired and Anthology advertise heavily to tech professionals in the Bay Area.)

Companies, too, would seem to benefit from these new services. Any stream supplying well-matched, on-the-market candidates beats the standard approach of hiring a dedicated recruiter to trawl LinkedIn. On its “For Employers” page, boasts of having helped 3,300 companies issue more than $17 billion in job offers.

So new services have demonstrated an ability to match job seekers with tech jobs. But can they change the job search game entirely?

Algorithm in a haystack

Prior internet companies have focused on serving jobseekers and recruiters. LinkedIn, and before that, have aspired to play roles similar to a global hiring bulletin boards. But newer players have an ambition to automate and improve the human judgment that goes into finding, and offering, jobs.

Fundamentally, the job market is a matching problem. The question is how to best sort individuals from the pool of job seekers and pair them to specific opportunities within the array of jobs on offer at any given time. And Silicon Valley’s algorithms have proven their ability in the past to build companies by matching up two sides of a market. Uber, for instance, reached a valuation of $60 billion by pairing those seeking rides with those offering them.

Job market services want to capitalize on this success. Anthology’s process“start[s] with advanced machine learning algorithms to predict potential matches.” Smaller player, meanwhile, applies “a combination of human intelligence and technology.”

In the abstract there’s a clear opportunity to better match job-seekers with openings. The monthly count of unfilled job openings in US from the Bureau of Labor Statistics stood at 5.4 million in October of 2015. And the time it takes to fill a given job vacancy has actually increased in recent years which indicates a decline in the efficiency of our economy’s current process for matching individuals with opportunities.

A human touch

In conversation, founders of job search-focused start-ups tend to recount their own unsuccessful encounters with the status quo. The common story is of looking to hire for roles in a prior start-up, and being forced to screen through a lot of milky candidates for very little cream.

But the same founders — as well a close examination of publicity materials — also reveal the ways in which new services fail to break with the status quo. Despite the talk of fancy algorithmic approaches, these seem to play a limited role.

Asked about the balance of screening candidates through human judgment versus algorithms, Underdog founder Josh Muir says the service’s current process “is mostly human,” and says the effort to manually sort through and grade the site’s tens of thousands of applicants is taxing the team. Of the bigger players, Anthology notes that every match it sends out will be screened by a human, and Hired touts the large number of candidates who are referred into the platform the old-fashioned way, by friends.

The degree to which new services rely on human screening and referral networks raises questions of how much these companies can truly distinguish themselves from traditional headhunters.

Hsu, the software engineer and success story, says he found his most recent job not via Hired, but through a venture capital firm. It’s common practice for VC’s to recruit on behalf of their portfolio companies, and to Hsu the experience was essentially comparable to using an online matching service. Just like Hired, the VC firm knew a number of companies with openings and could offer advice and referrals for getting into them.

A Tech Industry Thing?

Another question for new applications and services is whether approaches and lessons from hiring in the technology sector — where most have their initial foothold — can scale to other industries. Aline Lerner, a former start-up recruiter who blogs about technical recruiting, distinguishes between two problems within job search: “sourcing” is the challenge of getting candidates in the door, while “evaluation” is the problem of scoring and filtering candidates once they’re there.

In Lerner’s words, “Engineering hiring isn’t a filtering problem. It’s a sourcing problem.” She’s implying that good software engineers will remain in extremely short supply, regardless of how good your matching algorithm is. In this respect, the tech industry these start-ups are conquering may differ greatly from other domains, where firms may be overrun with seemingly comparable applicants face a real filtering problem.

There is software to aid companies with this filtering problem, but it’s relatively crude. Word-matching tools help recruiters make a first cut on their résumé piles. The tools surface those job-seekers whose submissions match the most words within a job post (so if you’re applying for a “tutor” job, your “teacher” résumé might not make the cut).

The coolest consumer-facing tool may be one that reverses the power of these algorithmic hatchet men in your favor. Give a copy of the post you’re applying to and it will help you choose words that optimize your résumé to beat the computer on the other side. In this case at least, the algorithms are here, and they’re in for a fight.

I reported and wrote this as part of a journalism workshop at UC Berkeley Extension.
Original date: December 13, 2015 / Updated: February 3, 2016

No comments: