Case study · EdTech · B2B platform

Taking the talent team out of the equation.

The goal was to simplify job posting for recruiters, enable real-time candidate tracking, and reduce their dependency on the GrowthX talent team. I had 10 days to design it, end to end.

Role Product Designer Company GrowthX Team Solo Duration 10 days
What I owned User interviewsService mappingCard sortingIAInteraction designVisual designPrototypingUsability testing
−62%Form completion timePrototype test, 4 testers
7 → 4Steps in the posting formCard sort with 3 recruiters
0Talent-team touchpointsFlow map, before vs after
10 daysResearch to prototypeSolo, end to end
01Project overview

About GrowthX, and the process I walked into.

GrowthX is an eight-week cohort course helping professionals become better growth leaders. Companies come to hire from that alumni pool, and to do it they fill in a job posting form. On paper it was self-serve. In practice, a small internal talent team sat in the middle of every single posting.

A recruiter would fill the form, get something wrong or leave it vague, and the talent team would email them to clarify. Then the talent team would manually read through candidates, pick the ones who fit, and email those names back. Every job post consumed hours of a team that should have been free for strategic work. As cohorts grew, that model was going to break.

Service map of the original flow: recruiter fills a Tally form, post appears in Notion, talent team manually curates and connects candidates

Understanding the current flow · the talent team appears in four of six steps

02Research · user interviews

User interviews: two sides of the same problem.

I conducted user interviews with the GrowthX talent team, who oversee every aspect of recruitment, and with representatives from three recruiting companies, who write the postings. I deliberately spoke to both sides before drawing the flow, because the two sets of complaints only make sense together: what confused recruiters became the talent team's workload.

What recruiters said

Ambiguity in form filling
Recruiters were unsure what to write in specific sections, so they filled in whatever seemed plausible. The job requirement came out unclear, which is exactly the thing the form exists to capture.
Overwhelm and complexity
The extensive form asked for everything at once, so people rushed it and skipped the attributes that actually attract good candidates.
Transparency gap
Once a post went live it disappeared. No view of applicants, and no way to halt hiring when priorities changed inside their own company.

What the talent team said

Clarification queries
A steady stream of "what goes in this field?" questions, each one interrupting real work.
Engaging with partner companies
When a form came back wrong, someone had to reach out to the partner company and collect the missing information by hand.
Manual candidate selection
Every recommendation was a person reading candidates and writing an email. Cohorts were growing, and scalability was becoming a real concern.

"The real fix was not a better form. It was taking the talent team out of the equation, so they could focus on strategic work."

The reframe that set the direction
03Design ideation · job posting

The posting form: from seven steps to four sections.

Since most recruiters fill the form on a laptop, in the middle of a working day, we decided to optimize for the desktop flow first, for both the posting form and the dashboard. My first attempt at the structure was wrong, and it taught me something.

Where I was wrong

In the first version, my objective was to give hiring managers complete transparency by revealing all seven steps from the get-go. I soon recognized this was overwhelming them instead: seven visible steps made a routine task feel like an onboarding flow, and it read as heavier than the form it replaced.

I aimed to divide the form into a few digestible sections, aligned to what a hiring manager already has in their mind. Rather than guess at the grouping, I initiated an open card sort with three members of the GrowthX recruitment team and let them cluster the questions themselves. Seven steps collapsed into four sections: basic details, budget and ESOPs, interview and role expectations, then everything else. This activity acted like a guiding light for the rest of the design process.

Card sorting output grouping form questions into basic details, budget and ESOPs, interview and JTBD, and other details

Open card sort · 3 participants from the recruitment team

Three additions that came from listening

Save draft
Hiring decisions are not made in one sittingA chat with my product manager highlighted the need for Save Draft: budget and headcount get argued over for days inside a company.A form that cannot be paused forces recruiters to either guess or abandon. Save Draft lets the form match the pace of the decision behind it.
Examples
Tackling the uncertainty about what to writeWe noticed recruiters leaving the form to generate wording in ChatGPT, then coming back.So we introduced example answers inline. Importantly, this did not just improve the accuracy of what got written. It gave people better raw material for the tool they were going to use anyway.
Share draft
A job post has more than one authorDuring the prototyping phase, the talent team shared a fresh insight: recruiters wanted to collaborate on the form with their team leads before publishing.Share Draft turned a solo form into something a hiring team completes together, which is how the decision actually gets made.
V2 of the job posting form with four tabbed sections, sample post link and save as drafts

V2 · four sections, save as draft, sample post always within reach

04Design ideation · dashboard

The dashboard: real-time tracking, deeper candidate insight.

The other pain point we had to address was the transparency gap: boosting recruiters' visibility with real-time job post tracking and deeper candidate insights. I mapped the states a job post moves through and gave recruiters three they can act on: active, inactive, and drafts. Inactive matters more than it sounds. It is how a recruiter halts hiring when priorities change inside their own company, without deleting the post or emailing anyone.

Dashboard user flow showing job statuses, applicant views, shortlisting and interview stages

Dashboard flow · active, inactive and draft states with applicant tracking

One of the key aspects I wanted to understand was what a recruiter would like to see upfront. The job card went through several rounds, and through feedback from the product manager and the recruitment team we arrived at the optimal card: budget, function, and applicant count, because those three answer "is this working?" at a glance.

Iterations of the dashboard job card converging on budget, function and applicant count

Job card iterations · converging on the three signals recruiters check first

In addition, we made the process more efficient by automating candidate recommendations, doing away with the manual emails entirely. That single change is what took the talent team out of the equation, and it removed the ceiling on how many cohorts the platform can serve.

Updated service map showing recruiters and candidates operating independently with no talent team intervention

The updated flow · no intervention from the talent team required

05Usability testing

Insights from usability testing.

Keeping the time constraint in mind, I invited four team members to test the prototype and asked for their feedback on two tasks: post a job for a Head of Growth role at an early-stage startup, then read the resulting post back and share what you expected to see, and why.

−62%Average time to complete the form, roughly 8 minutes saved per posting.
Fewer questionsTesters on the new form asked noticeably fewer questions than testers on the old one, which was the whole point: questions to us were the talent team's workload.
Mixed signalThe 30-60-90 day plan section divided people. Some found it valuable, others found it heavy. I left it in and flagged it for real-user validation.
Where this number comes from, and what it is worth

Four internal team members, not real recruiters, tested the prototype, and most of what we collected is qualitative. That means the 62% is a directional prototype signal, not production data: the structure was clearly faster, but the exact magnitude may not hold. We anticipated validating it with task timings from real recruiters over a full posting cycle. I am stating the limitation because a number is only as good as the method behind it.

06Reflection and learnings

What this project taught me.

I truly enjoyed my time on this project, delving into the genuine challenges recruiters and the talent team face daily. Three learnings I carry forward.

Look for the people holding the system together.The brief was a form redesign. The actual problem was a team quietly absorbing the product's gaps. Drawing the current flow end to end is what turned a UI cleanup into a scalability fix.
Be tactical about research when time is short, and honest about it.External recruiter interviews were hard to arrange in ten days, so I used the company's own recruitment team for the card sort and colleagues for testing, and labeled the confidence level rather than dressing it up.
Design for the tools people already reach for.Recruiters were leaving the form to use ChatGPT. Adding inline examples worked better than trying to make the form so good they would not need it.

Future scope

Based on recruiter feedback, two directions are planned for the current product: AI-powered form personalization, using a recruiter's past behavior to pre-fill sections like the 30-60-90 day plan for easy, personalized adjustment, and email-integrated candidate updates, so recommendations reach recruiters where they already work.