Automated job application questions

Background

Over 40% eligible applications were not marked evaluated by recruiters due to their heavy workload. The screening questions which are aimed to test an applicant’s ability to meet basic qualifications are not working effectively.

The problems include: (1) the content of the existing job-specific questions library is flawed, e.g. 3 different recruiters have the same question in 3 different ways. (2) it’s a manual intensive process and a low return on investment effort for recruiters to use those questions or to submit new questions. (3) there is no scalable way to grow the questions library.

Due to the sensitivity of the project, the details are not available on public domain.

Objectives

This product is a reimagining of the qualification creation process during job creation.

  • To provide a flexible and compliant qualification template framework for defining basic and preferred qualifications. This removes the need to manually type in qualifications from scratch.
  • To speed up the job creation process by recommendating qualifications to job creators based on the role specified.
  • To speed up the screening question creation, by automatically creating corresponding screening questions for applicants.

thumbnailimg

↑ Defining the MVP scope

My Role

Primary designer collaborating with a direct team of 8 people to deliver MLP. I was involved from the start, contributing to the product definition, rollout milestones plan and the scaling strategy (how can we transition from rule-based recommendation engine to machine learning implementation).

My learnings

Key design challenges include: (1) work within the constraints and known challenges of an existing system; (2) navigate through uncertainties and complexities of the recommendation engine; (3) adapt users’ mental model in a helpful way for the long-term benefits; (4) proactively find ways to remove blockers and keep moving forward the design work.

One thing I learned was that by diving deep into the interconnectedness either between the content and the recommendation engine, or between closely-related yet different systems, the insights we are able to gather can give us real footing for making tradeoff decisions.