Bringing businesses closer to job applicants
I assisted Zenjob customers in improving communication with users by simplifying the process.
How does the business work?
The problem
Zenjob serves customers from various fields and industries. Certain customers have their own specific needs. Previously, account managers handled these needs manually, which limited the number of staffing orders that could be processed without intervention, leading to higher operating costs for the company.
“I need specific flashy titles for my job openings to get quality talents”
- Event manager for Gigs
“I have unique door codes that change every day, so I ask my Account manager to add it to the Job posting.”
- Quality manager for Clothing brand
“I ask my Account manager to get me physically stronger talents for my jobs”
- Talent lead for Packaging company
“My warehouse forbids phone usage so talents need to know how to use lockers to keep their devices”
- Floor Lead for Warehouses
Solution in short
Customers can share their feedback, ideas, instructions, and custom titles directly on the job postings when they order from the B2B web app.
This cut out the need for Account managers to manually process these orders with custom requirements.
Impact
93%
Drop in comments that
were previously being tackled
manually for the ordered jobs
45,000
Ordered Jobs saved from manual work by account management.
The Design decisions made the biggest impact in hitting the OKR of reaching 99% automation of customer orders. It reduced significant time spent by Account management previously thereby reducing costs to the company.
Additionally, we were able to bring our customers closer to our talents.
Discovery & research
Based on the outcomes of the research & workshop, I out the following together a user journey to understand & visualise the problem areas.
Stakeholder Workshop & User research
I started the discovery process by facilitating a workshop with the PM and Account managers who were processing these orders manually. The workshop helped us better understand the problem and nature of the manual requests.
Problem 1:
A need for personalised job titles for certain customers.
Problem 2:
A need for customers to share job-specific comments with users.
I collected all the manual requests from the top 20 customers with the highest orders with the help of Account managers and ranked the requests based on frequency.
2. Validating problems with data
3. Benchmarking Multi user marketplaces
Based on on the benchmarking several marketplace products where two user groups communicate, I concluded that text based input fields are the most common interaction pattern for similar problems.
Ideation and First iteration
Sketched out wireframes based on the above insights to identify the journey point to introduce a solution.
Collaborated with the Product manager & Lead developer to align on the ideas & run a feasibility check
Analysing the Mental model of the users
Problem 2:
Input field within a Modal for users to add their custom job titles.
Modals were the interaction pattern of choice of the users to complete any action on the platform.
2. Low fidelity ideation & feasibility check
3. First iteration
Problem 1:
Input field for users to add their custom job titles.
✅ Upon user testing the solution showed positive results on the UX metrics & I agreed to release it to affected users to track our success metrics.
Overall solution:
Users could share the specific job related information by clicking a CTA which would let them add the info on the text input box.
This information will then be shown on the job post for the applicants to see.
Rationale: Users can see how their requests will be shown to the talents so that they are mindful with their comments.
Idea 1: Show customer manual requests within a comment directly on a real time talent App within a pop up modal:
Verdict: Aligned with the team that this idea has a Higher engineering cost & delivery time.
Idea 2: Show customers some guideline text and an input box to collect their manual requests.
Rationale: Faster to build & deliver the solution and easier to iterate further based on user feedback.
As a team we decided to learn by releasing the above idea instead of Prototype testing it with users to align with the delivery timelines.
Mixed results from First iteration release
Problem 1: SUCCESS ✅
Customers could now add their own titles, this lead to auto processing of a large chunk of orders previously processed manually by Account managers.
Problem 2: FAILURE ☹️
Customers were confused and didnt read the instructions on the pop up modal. They ended up sending more requests to Account managers to service their orders. The initial solution failed to solve the problem & instead escalated the problem.
Second iteration for Problem 2
Analysing the feedback from the release
1. Users don’t know who the comment is intended to at the first glance.
2. Users closed the modal as the guidance text had a high cognitive load on the users.
3. Users don’t know what type of requests can be shared with the talents for the particular job.
2. Refining the previous idea to design the next iteration
3. Visual Design based on Design system
Improving the previously rejected idea based on the above
feedback and constraints from the developers.
Initial success but with more problems
The second iteration hit the metric that I had set in the beginning but it led to more user generated problems that I had anticipated as its a free text field.
The customers started adding discriminatory comments that are not simply acceptable to be shown as part of Zenjob’s job offerings.
Further, customers also disclosed personal information of applicants by adding them as a comment in the free text input field.
Solving anticipated problems with error states
To tackle gender-biased comments, developers used an AI model to prevent these terms when users type. I adjusted the design to display an error and block the primary action.
Problems with discriminatory content
We faced more challenges: customers sharing personal info in comments about talents. Other applicants saw it on their app.
Solution: implemented similar approach as before, and developers updated ML model to recognise applicant's personal info.
Problems with sharing personal info
Final design
Key learnings
Any feature can always be made better, but it won't fix every possible problem.
Most problems are complex. It's best to test early and iterate to achieve a good result.
Learning by doing limited releases helps get faster user feedback on design decisions.
What would I have done differently?
After the release, I realised that letting users type anything they want in the text fields can cause problems. If we decide to improve this feature later on, I'd prefer to move away from unrestricted text and instead help customers customise their orders by sorting their comments into categories.
These categories could then be turned into selectable options when placing an order.
Team setup
Product designer
Product manager
2 x Front end devs
UX researcher
Content writer
Process
The solution was shaped in an iterative approach. Ideation, alignment and testing were the key steps.
Designs were validated through simple exercises like sketches and wireframes.
Metrics
Percentage Reduction in comments processed manually on ordered jobs.
OKR: Reduce staffing cost to serve by automating 99% of our orders.
