Julien 'Jubijub' Chappuis

Jubiblog

A blog about IT stuff, Corporate life, Management, Data science, and the occasional rant.

Why the hiring process can seem unfair

Reflections on why and how companies hire, and why this can seem unfair from the candidate's point of view.

Julien Chappuis

15 minutes read

unsplash-logoMarina Zanotti via iStockPhoto

Introduction

The idea for this post came from discussions with people on Hardware.fr forum about the unfairness of hiring. I also drew from my own experience as a candidate, and as a recruiter getting feedback from candidates.

The general consensus about recruitement is that the best person should get the job. Yet somehow the recruitment process can appear unfair in many ways. There is the obvious gender / origin bias at hiring, whether people are on the discriminated side, or consider themselves “victims” of the positive discrimination schemes. The unfairness perception can also come from hiring processes filtering for attributes that don’t seem immediately relevant for the position. Or simply, it can come from not getting a job for which we thought we were the ideal candidate.

Obviously these are perceptions, that may or may not be backed by facts. But judging from the number of comments I have seen online about this, I thought this was a topic worth exploring.

Having been on both sides of the fence, this post will try to explain why such unfairness happens, and why everything may not be so unfair after all.

Why do companies hire?

Companies hire to fill a need. A good hiring process is thus a process that allows the company to fill that need, in the fastest and cheapest way that guarantees to find the most suitable and willing candidate(s) for the position(s) in an ethical way. Sheesh, that’s a mouthful.

Let’s take an example, assuming the company is looking for 5 Software engineers:

  • fill a need : the company just needs to find 5 suitable and willing candidates to fill the positions : no more, no less.

  • fastest : hiring takes time from the people involved in the hiring process (which disrupts their own job), and having a position open may slow down or prevent critical activities to start. Companies thus want to minimize the time it takes to fill those 5 positions. It also leads to trying to find the right candidates within the smallest possible sample of the population.

  • cheapest : there are costs to hire for a new position (time of people involved, delayed benefits, hiring fees, …). Also, due to the risks posed by a bad hire, companies need to design hiring processes that mitigate these risks. As with any risk management, the cost to mitigate the risk must be adapted to the risk impact : it’s about finding the right trade-off. The culture of the company / hiring team, and the hiring style it drives, also has an impact on the cost. Some companies like long and thorough processes, some are more “instinctive”. As a result, it’s not about the cheapest process in absolute, but cheapest process that guarantees all the other elements.

  • suitable : a candidate that matches all the demands of the job. In a future post, I’ll write on how we hired Business Analysts and Project Managers at Nespresso, and how we defined and tested for suitability. This is mostly a trade-off about what to expect from the candidate from the start versus what the company is confident they will learn on the job. The company must also be realistic with what they are trying to hire, and understand their attractiveness on the job market for this type of profiles (Nespresso was a paradise to hire marketing people due to its brand image, but hiring for IT was pretty difficult as the company has no notoriety in that field).

  • willing : not all candidates the company likes accept the position in the end. This can be due to being overpowered by a more competitive offer. The candidate may also have disliked the position / manager / team after finding out more during the interview process (remember, candidates interview the company too).

  • ethical way : ensuring the process leads to no discrimination, treat the candidates respectfully, etc.

Companies can influence these parameters by optimizing their job descriptions and hiring processes, and are fine tuning the parameters as the process moves forward depending on the flow of suitable candidates.

How do companies hire: the hiring process

The most reliable way to hire would be to take all the candidates, and test them on the job for 6 months. It would allow to test them in real situation, and really see what they can or cannot do, assess their behaviours, etc.

It’s impractical for obvious reasons :

  • On most jobs companies can receive from a dozen to a few thousands of resumés. Testing candidates in parallel would be too disruptive (there wouldn’t be enough job for all candidates), and testing them one by one would take centuries.

  • The receiving team would incur a long term productivity loss due to the constant training of newcomers

  • The company would incur a performance loss every time an unsuitable candidate would be tested.

  • It would be costly, as each position would cost [number of candidates-1]×[monthly salary]×6 to test all candidates.

As a result, companies need to find a way to test candidates to find the right one in a cost effective way. It’s also about finding a process that minimizes bias when hiring (more on that below). This is what the hiring process is about.

Any hiring process is thus a binary classification test in the sense that the outcome is binary : it tests all candidates, and classifies them in two categories : Should be hired, and Should not be hired. The performance of this test can be measured through the notions of accuracy, specificity and sensitivity.

Those notions are often explained using medical tests performance as an example. Patients expect a medical test for disease X to be accurate. The test should accurately tests as positive patient that have the disease X, and as negative patients that don’t have the disease X. But a good medical test should also have a high sensitivity (it should avoid false negative, i.e. telling patients who have the disease that they don’t have it), and have a high specificity (it should avoid false positive, i.e. telling patients who don have the disease that they have it). These expectations are absolutely reasonable, as the consequences of a wrong medical test can be disastrous.

Somehow most people expect any hiring process to behave the same way. It is expected that the test should be positive for all good candidates, and negative for all the unsuitable ones.

In reality this is different. Hiring processes are optimized for specificity, absolutely not for sensitivity1.

Optimizing for specificity : avoiding hiring unsuitable candidates

Hiring an unsuitable candidate is the worst possible outcome of a hiring process. An unsuitable candidate will have a negative impact on the company throughout their employment. Worse, the company will have to incur again the time and cost penalty of finding a better candidate, and even the cost of laying off the unsuitable candidate if the mismatch is detected after the probation period. As a hiring manager, this is THE thing you are absolutely trying to avoid.

As a result, hiring processes usually take good care to be as specific as possible. This is why any hiring process usually involves different people, testing different things, to maximize the chances to spot a problem and reject unsuitable candidates.

Sensitivity : not much, actually

Sensitivity (considering good candidates as unsuitable) is of no concern for a hiring process. This may sound shocking, or even insensitive (pun intended): What is this process that doesn’t recognize good candidates?

To explain this, we have to come back at the root of why companies hire: to staff x open positions. If you need 5 Software Engineers, it doesn’t matter whether your hiring process identifies 20 good ones, or just 5. Beyond a safety margin, selecting more candidate once you’ve found enough suitable and willing ones is a waste of time and money.

That does’t mean companies disrespect candidates. I will take a daring metaphor to explain this: Imagine you happen to live next to a kick-ass Italian shop. Imagine they stock the best scamorza you’ve ever tasted. You are about to bake a pizza, and all you need is one2. It doesn’t really matter is the shop has 20 in store, equally as good, you would still only need one, right? Look at the cover picture of this article : assuming all pieces are equally as good because they likely all come from the same producer, and all you need is one, it doesn’t really matter which one you pick, does it ?

The case of for positive discrimination

I have frequently read that positive discrimination (for instance, trying to ensure gender balance, or a greater mix of origins) is unfair 3. The reasoning is that this is a breach of the principle that the best candidate should win. This line of thinking assumes that the recruiter has to lower the bar to hire women, people from diverse origins, etc.

This has largely been debunked, and it’s beyond the scope of this article to explore this in details. To continue on the topic of fairness, I will challenge the idea that gender balanced or diverse hiring is unfair because it requires to lower the bar. I have hired more than 35 employees over the last 7 years, and probably selected around 30 consultants (I’ve always chosen my consultants using the same criteria as for employees, for obvious reasons). I have never had to lower the bar to strive for gender balance or diversity. What is true is that this requires more work during the hiring process. The hiring manager has to strive to remove any implicit notions of gender / origin in the job descriptions. I learned recently how “recommended” skills can also be harmful as men are more confident to apply without having all the skills, which creates an unwanted filter for women to apply. But it’s doable without much hassle, and like everything I got faster at it through practice.

During CV processing, my strategy for men and women was different. Due to the sheer volume of men CVs (95% pre-selection), if I had any doubts, I would discard the CV. For women I would keep the CV unless there was big blockers. Given the relatively low number of woman applying, this was a suitable trade-off to accept higher uncertainty in the early phase of the filtering. I don’t mention origins / ethnicity because I read the CV header last. This simple yet effective trick removes any bias during CV reading as I don’t know anything about the gender, nor the ethnicity or any other criteria some discriminate against. I check this only after having read the CV.

I knew it, right there, this is blattant unfairness !

Seeing as there are less than 5% more men that women of age to work on Earth, and since sending a CV versus not sending it is not a proof of skill, I really don’t think so. There are no reasons why woman should be less qualified than man for any IT job (unless you believe any of this crap), and history has shown women were in fact pretty good at it.

This approach created a slight over representation of women at the phone screening step versus their original representation on the original CV pool (I reached around 80-20… however, this never overcame the lack of woman CVs due to the pipeline issue. I work in IT and it’s a real problem4), and this was the whole point.

From this point onwards however, I would do the phone screening and interviews exactly the same way for men and women, asking the same questions and expecting the same answers. Candidates were tested the exact same way.

Unsurprisingly, by applying the same testing standards, I got very consistent performance out of all my employees. I have never found any evidence that gender or origins were a predictor of individual performance, so I am confident that a hiring process can both strive for diversity, and yield good candidates at the same time.

So if you ever lose a job to a non white male candidate, simply assume that person was a better fit than you for the job. It’s also the most logical explanation from a statistical point of view : by tapping into more pools of talents, the likelihood of surfacing a better candidate increases.

Hiring is testing for proxy indicators

Performance is a complex topic. There are many dimensions to what makes a “good” employee, some of which are not even related to raw performance. In fact, there are probably too many dimensions to test all of them thoroughly during a hiring process. So how does one decide what makes a “good” employee, and thus what to look for in candidates?

Even with the right set of criteria, there can be failures too. I especially like that video from Moishe Lettvin sharing his learning after 250 interviews at Google. At 9’ he gives this amazing anectote when the HR tricked the hiring committee by submitting their own interview packets. The Hiring Committee members wouldn’t have hired themselves 😂 . That video gives extremely good insights of Google (and other tech companies like Microsoft) approach to hiring, and I recommend watching it.

The HR of Google, Laszlo Bock, infamously said the company has found no single person in Google that had a strong history of hiring, showing interviews are not necessarily good at finding future performing candidates.

Know your existing employees

To hire good new employees, it’s a good idea to find out what makes your existing employees good. What could be possible common traits?

For the role of Technical Product Manager, GAFAM and other SV companies have those traits well documented : business acumen, product design, product strategy, agility with estimations, decent Software engineering understanding.

For the role of Business Analyst, we defined those golden traits in the context of Nespresso as someone who could act as a business / IT translator (traditional business analyst scope), a system analyst (hands on person that dive into the system config), with decent IT skills (can analyze data or read some code), with autonomy (can find their own path to achieve an objective).

In both case, that doesn’t mean that having these traits is sufficient on its own to be good. But experience has shown that these traits are excellent predictors of future performance.

As a result, keeping efficiency in mind in processing hundreds of CV to interview dozens of candidates, focusing on testing those traits is a sensible approach. It increases the yield of phone screening and face to face interviews by targeting the sweet spot of the bell curve.

In fact, doing so probably removes a lot of good candidates. But that’s OK because it also removes even more bad candidates, and reduces recruitment risks. From the company perspective, it is an acceptable trade-off to reduce the test sensitivity in favor of reducing the hiring risk, and speed up recruitment.

But it’s unfair: Role x sure doesn’t mandate skill y!

Back to the topic of fairness, any selection based on these golden traits can appear as unfair. A blog post I read last year, or this tweet give a fairly good depiction of the perceived unfairness. All testimonials follow a common pattern :

  1. I do job X with some success
  2. I did an interview for job X in another company
  3. They tested me on something Y I couldn’t do well
  4. I didn’t get hired
  5. Profit Frustration, because if Y was mandatory to do my job, I’d know by now.

As with any sampling approach, you will find outliers. You can probably find amazing Product Managers out there that are not so good with Software engineering. Or, as with the examples above, excellent developers that don’t know how to solve a particular algorithmic problem with a decent complexity.

This is precisely where companies are trying to reduce the risk of hiring. If your internal studies show that all your best performers are good at Y, then you will focus your hiring efforts to find candidates that are also good at Y. In other words, if you found that skill Y is a good predictor of performance, there are few reasons to take the risk to not look for Y skill in your candidates.

Here my recommendation to candidates would be to search for the typical “predictors” that companies in your field will look for in a specific role, and make sure you master them. If you are genuinely good it shouldn’t be too much problem. Or you can simply avoid companies hiring on predictors you disagree with. Sites like Glassdoor offer good insights in what companies are looking for, especially for the big ones.

But I am the best for this position!

You don’t know, you haven’t read all the CVs.

If you are Usain Bolt breaking the 100m world record, assessing whether you are the best is easy. If you are in the final round, this means you are among the 8 fastest people that took part in the competition. Therefore if you see no one in front of you, you are the best. Hiring is different as you have no access to competition. This is closer to a school exam. You cannot know your ranking until all exam papers have been scored, and the ranking published.

So that’s it, this is another case of unbalance favoring the employer / companies ?

Having been on both sides of the fence, I really don’t think it is. The reason is that it ain’t over until there are two signatures on the contract. I have lost quite a few good candidates for a lot of different reasons : I failed to sell my company / role good enough, my offer wasn’t competitive enough, another company had a more attractive position, the candidate used my offer to present a credible threat to their current employer to get a better job or a raise, their significant other decided against moving to my city5, etc.

In the end, it all boils down to whether your type of role is in a buyer or seller market, and hiring processes have to adapt to that. There are many fishes in the pond, and if you managed to keep a good employability, you will find a job elsewhere.


  1. that Coursera Data Science specialization surely has paid off, hasn’t it ? [return]
  2. I swear, the fact this happened this morning had nothing to do with my choice of example. I simply took bribes from the Italian cheese lobby. [return]
  3. I heard it from white males, as you would have guessed 😀 [return]
  4. I know the pipeline issue is not the only issue, cf this article. I am not hiding behind this. [return]
  5. You’d think people would check this before applying, but apparently not 😂 [return]

Recent posts

See more

Categories

About

I am Julien Chappuis, formerly known as Julien Bidault a.k.a. Jubijub, or Jubi for shorts. It's complicated.