Increase number of candidate applications accepted and improve candidate recommendations engine
Duration: 16 weeks
Research: User feedback, user interviews, analytics
Prototyping: Sketch mockups + Invision prototype
Testing: In-person usability testing, iteration w/ clients
Strategy: UX + product strategy
Visuals: Company branding
This project is a redesign of an already existing feature that was built for the company's old business model. As the company shifted to a subscription model, this feature needed to be redesigned to move away from the old business model which limited how recruiters were able to view candidate profiles and move to the new business model were recruiters were encouraged to talk to candidates they are interested in.
The redesign focused on two things:
- Showcasing good candidate profiles and enabling recruiters to talk these candidates
- Creating a feedback mechanism so that the platform recognizes mismatched candidates and improve on recommendations
Insights showed that recruiters who maximized the Kalibrr messaging platform had a higher chance of successfully hiring a candidate. As such, encouraging the use of the messaging platform became a focus of the product strategy. With this redesign, messaging options are more visible and prominent.
Kalibrr is powered by artificial intelligence. One of our value propositions is we make it easy for recruiters to find great candidates by recommending those candidates to them. To make recommendations more accurate, a feedback mechanism is needed. This feedback mechanism empowers recruiters to tag recommended candidates they think are not good recommendations. Doing this gives the algorithm a better sense of what the recruiter like or does not like in a candidate. This feedback mechanism was a key factor in the redesign of this feature.
Although not a key focus, this redesign also tried to improve the efficiency of how recruiters processed applications. We learned that, if there were a lot of applicants for a certain job, recruiters are not able to respond to them. This leads to jobseekers being unhappy with how their applications were handled and recruiters who were stressed with the amount of applications they have to process. In trying to solve for this, the design and product team decided to emphasize an already existing but under-utilized feature: filters.
What the filters did in this feature was group the candidates into the following categories:
- Candidates who have a complete application and passed the filters
- Candidates who have complete applications but did not pass the filters
- Canddiate whose applications are still in progress
However, we found out by doing ethnographic research, was that recruiters had a difficult time navigating through these groups, especially when each group contains a lot of candidates. There were a lot of usability issues as well. To fix this, we redesigned the interface and made it easier to navigate especially if there were a lot of candidates. Using AI, we were able to sort candidates according to their likeness to the recruiter's desired profile.
Although also not a focus but an added feature to make recruiter's work more efficient, we added a section where a recruiter can see how much the candidate profile matches their profile requirements for work, education, job level, skills, and location. We found through research that these five were the primary information the recruiter considered in each profile.
Below is a mockup of the new Candidate List interface when viewed from the Applications stage. The emphasized calls-to-action here are to "Shortlist" and "Reject candidate applications". Also notice the sectioned candidate applications.
Below is a mockup of the new Candidate List interface when viewed from the Leads stage. The emphasized calls-to-action here are "Invite to apply" and "Not a match". In this interface, there are no sections because there is no filter. The filter is removed because the number of candidates recommended in this stage are usually in the hundreds making it difficult to filter. These, however, are sorted by the algorithm according to the likeness to the desired candidate profile as deduced from the job post.
Below is a mockup of the a candidate profile in the Applications stage. The actions that can be performed on this candidate stick to the right hand-side of the profile and, thus, are always accessible to the recruiter. The emphasized calls-to-action "Shortlist" and "Reject" are listed at the top.
Below is a mockup of the a candidate profile in the Leads stage. The actions that can be performed on this candidate stick to the right hand-side of the profile and, thus, are always accessible to the recruiter. The emphasized calls-to-action "Shortlist" and "Reject" are listed at the top.