Project Manager needed. Must have 5+ years of experience, be Six Sigma certified, have advanced deep learning knowledge, and be able to perform surgery on occasion.

Who really has all that? Turns out, basically no one. You’re as likely to get a job interview meeting 50% of job requirements as meeting 90% of them.

requirements_required
So requirement is a bit of a flexible word in this context, then…

We were curious about how many job requirements are actually required, so we analyzed job postings and resumes for 6,000+ applications across 118 industries from our database of users. We found that while matching requirements is important, you don’t necessarily need to match all of them.

  • Your chances of getting an interview start to go up once you meet about 40% of job requirements.
  • You’re not any more likely to get an interview matching 90% of job requirements compared to matching just 50%.
  • For women, these numbers are about 10% lower i.e. women’s interview chances go up once they meet 30% of job requirements, and matching 40% of job requirements is as good as matching 90% for women.

You only need 50% of job requirements

You’re just as likely to get an interview matching 50% of requirements as matching 90%. We saw a clear upward trend in interview rates based on matching requirements, but with an upper bound. When users applied to jobs where they matched 40 – 50% of job requirements, they were 85% more likely to get an interview than when they matched less, and applying to jobs where they matched 50 – 60% of requirements made them an extra 192% more likely to get an interview over the 40 – 50% matches.

But after that point, you’re in diminishing returns. Applying to jobs where they matched 60% or more of job requirements didn’t provide any additional boost in interview rate.

Job Search Tip #1: Apply for jobs once you match 50% of job requirements.

For women, the % of requirements required is lower

You may have seen stories before about how women in particular don’t apply for jobs unless they’re 100% qualified. We wondered if they were on to something – maybe there’s gender discrimination at play and hiring managers look for women to meet more of the requirements. Turns out, our findings apply just as much to women as to men, and actually, for women, the chances of getting an interview start increasing as soon as you meet 30% of requirements.

requirements_required_by_gender
Women get interviews at higher rates with fewer matched requirements – if only they applied to those jobs in the first place.

As you can see in the graph above, we see the same general trend for women as for men, but for women, you’re as likely to get an interview matching 40% of the job requirements as matching 90%. Note also that, as we’ve seen in previous analysis, women in general have higher interview rates than men.

Yet, despite this, among our users, we’ve observed the same trend that has been studied elsewhere. Women are more likely to turn down jobs where they match some but not all of the qualifications – over the last 8 weeks, 64% of our female users rejected at least one job where they matched 50 – 60% of the requirements, while only 37% of male users did.

requirements_required_feedback_rates

So, yes, women, you too should be applying to jobs where you don’t meet all the requirements.

Job Search Tip #2: Stop second guessing yourself – you DO deserve that job.

You’re not guaranteed to get an interview, even when you match 90% of job requirements

Base case scenario, you’re looking at about a 15% chance of getting an interview. Applying for jobs is still fundamentally a numbers game – the more applications you put in, the more likely you are to get an interview, and the more interviews you have, the more likely you are to get a job offer.

Put another way, if you want to get a job offer, the number of jobs you need to apply to is a function of your interview rate (what % of applications do you get interviews for) and your job offer rate (what % of interviews do you get job offers for), specifically: # of applications needed to get n job offers = n / interview rate / job offer rate

Interview Rate Job Offer Rate # of Applications Needed to Get 1 Job Offer
5% 5% 400
10% 10% 100
15% 15% 45

Clearly, improving your interview rate and job offer rate pay off, but what if you can’t find 45 jobs that are perfect matches for you? It never hurts to broaden your search to jobs that feel like more of a stretch. Sure, your interview rate will be lower, but that’s balanced by applying to more jobs.

Job Search Tip #3: Apply to as many jobs as possible to increase your chances of an interview.

No time to fill in all those applications? We can help with that.

appbot
ApplicationAssistant will fill out all those applications for you (and submit them at the best days and times too).

Summary

When you’re out looking for the perfect job, don’t be intimidated by a long list of requirements!

  • Even if you only match 50% of the requirements, you should feel confident hitting “apply.”
  • This applies just as much to women as it does to men (actually, even more so!)
  • Cast an even broader net to improve your chances of getting an interview.

Remember, getting an interview is your big break – it’s your opportunity to prove that you can do the job even if you don’t meet all the “requirements.”

Methodology

First, we randomly sampled 6,348 applications for 668 different users from TalentWorks. Then we extracted the qualifications from the original job postings and the users’ submitted resumes using proprietary algorithms. Finally, we grouped the results based on qualification match and regressed the interview rate using a Bagging ensemble of Random Forest regressors. All analysis and graphing was done using python with pandas, sklearn, scipy, and bokeh.

Why Are We Doing This?

With ApplicationAssistant right now, we can boost the average job-seeker’s hireability by ~5.8x. But, what makes ApplicationAssistant work has been an internal company secret until now. We’re fundamentally a mission-driven company and we believe we can help more people by sharing our learnings. So, that’s exactly what we’re doing.

Creative Commons

We’re not only sharing this but also sharing all of it under the Creative Commons Attribution-NonCommercial-ShareAlike license. In other words, as long as you follow a few license terms, this means you can:

  • Share: Copy, redistribute the material in any medium or format.
  • Adapt: Remix, transform, and build upon the material.

14 comments

  1. This is an interesting analysis, but I wonder about the accuracy of the random forest. Are the blue dots the datapoints that are being regressed over? If so, it seems that a linear regression would not only be more accurate, but also tell the opposite story. Just going by the datapoints, it looks like the interview rate increases substantially as the match percent increases. Am I reading the plot wrong?

    1. Yes, the blue dots are the datapoints that are being regressed over. A few things to note – sample size isn’t depicted in the graph, but was taken into account for the fitted line, and there’s a datapoint around 75% match down at 0% interview rate as well as another datapoint that’s cut off to the right of the graph (in the 90%+ qualification match) that had a lower interview rate and was included in the fitting as well. If you take a look at the raw datapoints, it’s fairly clear that there’s a big jump right around 50 – 60% and then some up and down oscillation (likely related to sample size). A linear regression misses this sort of step-wise increase (there’s also another around 40%). Hope that helps clarify!

      1. Sorry, it doesn’t really clarify. What does “sample size isn’t depicted in the graph” mean? A scatter plot shows one point for each data point in the sample, so implicitly it indicates the sample size because you can count the points. Why is there such a small number of data points shown in the scatter? Is it a weird subsample of the full data set? Or is that all of the data? If that’s all of the data, the regression shown is not appropriate at all.

  2. It reads like you’re using this data to counsel jobseekers to make recruiting harder by increasing the signal-to-noise ratio. Why can’t we improve the quality and reliability of skills and job-listings? Isn’t that the real disconnect?

    1. You’re totally right that the job postings are the real issue here – but realistically, our goal is to help people who are looking for jobs right now, and we’re not going to be able to change the fundamental issues with job postings anytime soon (although I’m open to suggestions on how to, if you have any – other than striving for honest job postings ourselves, which we do). To be honest, I care a lot more about making things easier on job seekers than on recruiters – sorry, recruiters!

  3. @Chris Bielanski That’s because recruiters are the one that produce most of the noise. If you’re ready to hire someone who has only 30% of the listed skills, what is the other 70%? Personally, I call it noise.

  4. This is a pretty small sample…like miniscule. What was the T value on your regression for percent of skills being the independent variable?

    My initial reaction as a data analyst specializing in this particular space is 1) I want to see what variables your regressed, and 2) show a strong T value on 50% skills being the independent variable, and 3) have you also show how you know the candidate took an interview (I assume it’s a survey or you have ATS data).

    Andrew Gadomski
    Aspen Analytics

  5. Some but not all job descriptions divide their requirements into Required and Preferred sections. What does that say about your analysis?

  6. My question is about ATS. If you don’t have at least a certain percentage of the traits/qualities employers want, will your resume get past ATS and will human eyes ever set upon it?

  7. Interpretation of the first figure seems suspect. Based on the line alone you’re more likely to get an interview at 41% than 50% and maximally likely at 60%. Given you’re discussing 50% as the headline it seems weird to choose 4% axis ticks on the X such that 50% isn’t listed. Since you acknowledge the skewing influence of the 90% (and indeed the 75) arguably it would be better to not discuss anything over, say, 70%, since after that point there’s no further influence of data. Given the widening error bars and the low sample size it would seem reasonable to gloss over some of the high frequency variance and discuss the general trends which show a better match = more interviews. Simply going by eye I’d interpret this as a logistic curve with widest error bars at the midpoint.

Leave a Reply