LinkedIn Design Challenge

Humanizing machine learning for the recruitment process.

LinkedIn Design Challenge

Dec 2018

 
Cover.png
 
 

introduction | linkedin

As part of the hiring process at LinkedIn, I had the opportunity to take on an interesting design challenge. LinkedIn is popularly known as the leading platform for professionals and companies to network, recruit, and more with over 575 million users, 20 million companies, and 15 million active job listings (TechCrunch, 2018). With such established popularity and the growing trend of automation, the company has increasingly started to build a number of other business-intelligent related tools and services.

 
 
 

brief | 1 week

Design an improved job posting experience for recruiters and hiring managers, which would help them attract and recruit relevant candidates for positions.

 
 
 

process | google searches, surveys, deep burning questions

Understanding recruiters and seekers

 
 

After reading the brief, I set out to understand as much as I possibly could about the world of recruiters, such as their current business market, process of recruitment, and frustrations. On the flip side, it was also important for me to understand job seekers, their process of job seeking, and frustrations with job postings.

To achieve these goals:

  • I sent out a job seeking survey with 15 respondents

  • Conducted three interviews with job seekers

  • One interview with a senior employee at a large company

  • One interview with a small business owner

  • One in-person interview with an entry-level HR assistant

 
 
 

Competitive Analysis

Analysing the playing field

I was also curious about the experience of other platforms besides LinkedIn. In order to see what worked and what didn’t, I did a competitive analysis and audit of Craigslist, Indeed, Facebook Job Posts, and general job postings to see what I could leverage.

 
Competitive Analysis.png
 

Competitive Analysis

Takeaways

01. Prompts and suggested keywords can improve the posting experience
Through my research and interviews, I found that people saw value in prompts in body copy that help frame the details that they are entering when posting. Keywords also helped people gauge what type of position they’re looking for.

02. Posting previews are valuable when contextual
Craigslist and Indeed provide posting previews, albeit simply being a text box to review their job posting. On the other hand, Facebook had an interesting aspect where they included previews for both desktop and mobile. This allows posters to alter any details as needed. However, it’s important to note that Facebook’s previews don’t include what the job post actually looks like, but rather how it appears on desktop and mobile newsfeed.

03. Social connections can help job postings reach more people
Facebook’s prompt to share within your social circles was an interesting aspect I noticed with job posting. Interestingly enough, Indeed and LinkedIn lack that area of sharing a job post after one is created. With this in mind, I also found that most people post job links directly to their profiles already in order to reach more people through word of mouth and connections.

 
 

Affinity Map

I affinity mapped my research in order to visualise clusters of patterns.

 
 
Affinity-resize.png
 
 

Key Takeaways

01. Automation is a key tool in recruitment

I found through recruitment surveys and reports of 2016 and 2018 that there is a growing trend of machine learning and automation for recruiters and talent acquisition leaders. LinkedIn this year explained how Artificial Intelligence will change the way recruiters and talent teams seek candidates as the company pushes out services like LinkedIn Recruiter and Talent Insights.

Talent acquisition leaders who invest in analytics and AI will have more success scaling to meet their recruiting demands. Those that arm their teams with the intelligence and automation they need are the ones who will win the war for talent.
- Madeline Laurano, Chief Research Officer, Aptitude Research Partners, 2018

02. Job postings are often generic and impersonal

Through my conversations, respondents to my survey and interviewees all responded with how job postings are often generic and how they would spend more time considering tailored posts to the position that they were looking for.

03. Recruiters may cast wide nets when they’re unaware of who they’re looking for

Through my in-person interview with an HR assistant of a medium sized company, I recognised signs of frustrations as the interviewee spoke about uncertainties they have when crafting job posts for positions that they aren’t familiar with. Even more interesting, is how they mentioned that they’d resort to sharing the open position with colleagues in their personal social circles.

04. Small pools of qualified candidates

Less people are unemployed as the U.S. unemployment rate fell to 3.7% in November this year. With a shortage of great talent out there, the economy is raising candidate expectations and poses challenges for recruiters as this means they’ll face more competition to sway qualified candidates into the door. (Huntscanlon Media, 2018)

05. From “advertise and apply” to “find and engage”

I’ve learned through a 2018 report that the dynamic of posting jobs is shifting from a conventional method of recruiters posting and receiving applications to recruiters having to market themselves and the employer brand.

 

painpoints and frustrations

Establishing the challenges

Afterwards, I synthesised the frustrations and challenges of recruiters and job seekers.

I saw a pattern of how job seekers and recruiters were two sides to the same coin. Seekers disregard generic posts and spend more time with tailored postings to a position; whereas some recruiters may cast a wide net and post generically as they are not familiar with the job position.

 
 
Frustrations.png
 

Through analysing the challenges and pain points that seekers and recruiters have, I asked myself some general questions of how I could solve them. By asking ‘how might we’ questions, I was able to generate a couple of avenues I was interested in going down with the prototype.

HMW.png

PERSONA

To help narrow down my focus, I created a persona and their scenario in order to identify my target audience’s journey and friction points with creating a job post. This helped me guide the development of the prototype as I had a point of reference to who I am designing for.

Persona.png
 

Objectives, points of reference, and ethical design

The design goals

01. Help recruiters speak the same language as job seekers
I wanted to bridge the gap between recruiters and candidates.

02. Provide valuable prompts/informational popups
I saw value in an experience that helps guide recruiters as they create a job post in order to help them understand the candidates and their needs.

03. Utilise social connections to improve reach
It was interesting how job posting platforms (Indeed, Craigslist) lacked the social counterpart to job posts. Similar to Facebook, I wanted to design an experience which highlighted the social sharing and power of connections of LinkedIn.

04. Keep machine learning and automation human
Albeit a powerful tool, I wanted to keep myself accountable of the harm that machine learning and automation can cause during the recruitment process, such as alienation of demographics, candidates, and biases in design.

Begin by sketching

I put my thoughts to paper and quickly ideated some ideas of what I could make.

 
Sketches Annotated.png
 

Turning machine learning into bite-sized learning moments

Sketching my ideas helped me think about how machine learning and automation would look like during the recruitment process. I looked into the integration of machine learning into the already existing info tips that LinkedIn has on the right side of the job posting experience. However, I found these info tips to be too passive, which was also validated through my interviewees when I took them through LinkedIn’s job posting experience.

 
 
 

I wanted to experiment how info tips could provide more value besides 2-3 lines of advice in body copy. I explored card designs where I could incorporate machine learning into the application of automated matched profiles and skills to support the creation of postings. This would help promote more tailored posts as recruiters would have a point of reference of someone’s relevant experience and skills of the position at a glance.

LinkedIn already has business-intelligent tools and services that help match profiles to job postings, such as LinkedIn Recruiter. However, I noticed that these services and tools are implemented after a job posting was created. I found an opportunity to potentially integrate that technology during the creation process.

Matched Profiles.png
 
 
Relevant skills.png

Empowering recruiters through machine learning

Through my initial research, machine learning and automation poses a threat for workforces although an extremely useful tool in the world of online recruitment. With this in mind, I wanted to create an experience that helped automate relevant profiles, while empowering recruiters and build trust with intelligent technologies.

I recognised that machine learning can only go so far. In order to provide human control and choice (and not rely solely on machine learning), I took the stance where machine learning could help inform and fill in the gaps of fields that recruiters may not be too familiar with.

I originally wanted to pre-import, even import entirely, the skills relevant profiles showed when recruiters were adding skill keywords. However, through feedback and user testing, that could potentially further the generality of job postings. Instead, I opted to give choice to the recruiter and showed statistics. Through machine learning and data, skill comparisons in relation to the people of the job position can promote the understanding of recruiters regarding the field.

 
 

workflow screens

 
 

Proposed workflow

In this workflow, I proposed changes to the job posting experience.

Firstly, a recruiter can reference relevant profiles matched with people who also have the same job position that the recruiter is posting for on LinkedIn. This provides a point of reference for recruiters as they are able to read a profile’s experience and see their relevant skills. Machine learning is incorporated to provide data which may help inform recruiters to write the tasks and responsibilities of posts with the goal of creating tailored postings.

Secondly, through user testing and feedback, it was evident that showing a contextual preview of the job posting was valuable as this allowed recruiters to gauge what needs to be added or taken out of the job posting before it goes live.

Thirdly, I proposed to integrate the recruiter’s network and personal connections in order for them to share the job posting and have it reach more people.

 

small businesses, machine learning, and ethics.

I really enjoyed this project as it took me out of my comfort zone and familiarity. Through this experience, I was able to understand some frustrations that recruiters, job seekers, and small business owners have in the world of recruiting and hiring. As such, I designed an experience for those who needed guidance as they create job openings at their businesses.

I wanted to create an experience for those who are daunted by unfamiliar terminology of the positions they are hiring for. Through my industry research, interviews, and user feedback I was able to understand how frustrated or ashamedly confused people are with languages that they personally don’t speak in growing industries and fields.

It was a humbling experience and my designs are imbued with a sense of understanding and modesty for people who want to grow their businesses. As machine learning and automation grows, I was able to understand the worries people may have with automation potentially replacing workers. As such, I wanted to build trust and reliability through these new, emerging, intelligent technologies in this project.