Careers Experience Refresh

The jobs were on a different website.

Company

Workhuman

Year

2024 – 2025

Role

Designer, Information Architecture

Team

Web, with Marketing, Recruiting and HR

Screens, copy and reading order in this case study reflect the October 2025 launch. Workhuman’s careers site has changed since.

Context

Workhuman’s careers experience was a thin layer over Workday. Candidates landed, clicked a job title, and left immediately for the application system. Seven in ten sessions ended right there. This left Marketing with no journey data and candidates with no context.

The content to fix it already existed. It just wasn’t reachable during a job search.

3

teams with distinct cultures

4

regional benefit structures

2

offices and remote markets

Cross-functional: Marketing (3), Recruiting (2), HR (2). Competitive analysis across 10 careers sites — Canva, Figma, Stripe, Notion, Shopify, Atlassian and others.

Problem

In September 2025, 2,110 people reached the careers page. Five hundred and eighty-two went anywhere else on workhuman.com. The other 72.4% left, most over to Workday, where the jobs actually lived.

The jobs section barely registered because it wasn’t there: /company/careers/list/ recorded zero pageviews in late August and 269 in September. Everything a candidate came for sat on another company’s domain, which meant no context on the way in and no journey data on the way out.

“Candidates treat us like a job board, not an employer brand experience.”

HEAP, September 2025 — 2,110 careers-homepage visitors, 582 continued on site, 72.4% exit rate.

That cost showed up in three places at once. Candidates evaluated Workhuman on job titles alone, with no culture context to self-select against. Recruiters absorbed the gap and the same questions surfaced again in first interviews. Marketing could not attribute a single application to a campaign, which meant it could not defend the recruiting budget.

“We have no idea what content drives applications.”

Marketing, stakeholder interviews

Careers homepage at launch, October 2025. The Teams grid was removed in a later content update.

HX Careers Homepage Exit Rate

Date range: September 1 – 30, 2025

HEAP exit-rate funnel, careers homepage, September 2025: 2,110 users, 582 continued, 27.6% completion.

HX Careers Homepage Exit Rate

Date range: October 1 – December 31, 2025

HEAP exit-rate funnel, careers homepage, October to December 2025: 6,040 users, 4,467 continued, 74.0% completion.
HEAP’s own note on this chart: the exit rate is the inverse of the percentage shown in the funnel. 72.4% before, 26.0% after.

Solution

Candidates need different information at different moments, and the old page failed because it delivered all of it flat, at once, in no particular order. The redesign worked backward from the drop-off data: What were people looking for when they arrived, and where did the page fail?

Sequencing the answer to three questions

Lead with what Workhuman actually is as a place to work, give candidates a reason to believe it, then make the right role easy to find. Three layers, each answering one question before handing off to the next:

“Why Workhuman?”

Careers Homepage:

Reading order is the design: Testimonial video → Who we are → What we value → Ways of working → Jobs. Benefits and team profiles moved down a level onto the job detail tabs, where they answer a question the candidate is actually asking at that point.

“What roles are available?”

Open Positions page:

Filters stay optional while the tags do the scanning. Ex.: Product roles are pink everywhere they appear.

“Is this specific role for me?”

Job Detail page:

Three recommendations, persistent across every tab. This is where sessions stopped ending at one page.

Five tabs, five candidate questions: Role Description, Meet the Team, Culture, Benefits, Location.

DECISION FORK 1

Make filtering the primary way in, or keep it optional?

Most of the ten careers sites we audited led with filters, Canva, Figma, Stripe, Notion, Shopify and Atlassian among them. Filter-first was the safe pattern and nobody would have questioned it but filters serve the candidate who already knows they want Product in Dublin, and that wasn’t who we were losing. Almost nobody reached a job listing at all, so the failure sat upstream of search. Filters stayed optional and colour-coded department tags carried the scan, then the targeted searcher scrolls further than they would elsewhere.

The filters are still the first thing on Open Positions but the difference is that nothing is gated behind them. The list renders in full on arrival, and the colour-coded department tags carry the scan.

Five tabs, five candidate questions

The job detail page splits into Role Description, Meet the Team, Culture, Benefits, and Location, each tab mapped to a question a candidate actually asks. What’s the job? Who will I work with? Will I fit in? What’s in it for me? Where and how do I work? Tabs also made content depth independent: the Benefits tab could carry video and four regional card variants without making the Role Description tab heavier.

Each tab is its own URL rather than client-side state. That made the content independently linkable, cacheable and editable by content ops without a deploy.

There’s a structural reason the job page has to carry all of this rather than point back up at it. Across October to December, 16,503 sessions entered the careers section and only 4,436 of them started at the careers homepage, or 27%. Most candidates arrive at Open Positions or at a single job from search or LinkedIn and never see the values, the video or the team content upstream. For these users, the job page isn’t a supplement to the employer brand surface — it IS the employer brand surface.

Benefits tab, four regional variants. Adding a region needs no design or engineering change.

Recommendations that removed the back button

A persistent right rail, Candidates Also Viewed, carries three related roles across every tab, matched on department first, then location, then seniority, with a recency fallback when a department has fewer than three open roles, which at our headcount was most of the time. The rail is honest about being a category browse, not a match. Pages per session in the jobs section went from 1.0 to three or four.

DECISION FORK 2

Ship the simple matching rule, or build the recommendation engine?

Candidates Also Viewed runs on three rules: same department, then same location, then similar seniority. A real model was the more interesting problem and the better answer, and it would have pushed launch by weeks. The crude version shipped on time, tripled pages per session, and produced the click data a real model would need anyway. It recommends by category, not by candidate, so browsing sideways gets you more of what you just looked at.

Impact

Launched 1 October 2025. Measured in HEAP against September 2025.

72.4% → 26.0%

careers homepage exit rate

2m 13s → 42s

time on page, careers homepage

1m 28s

average time on a job detail page, across 73 roles

5,973

session entries on Open Positions, against 4,436 on the careers homepage

2.0

pages per session on Open Positions

28.9%

bounce rate, Open Positions

Time on page fell with it, from 2m 13s to 42 seconds. People spent two minutes on the old careers page because they were hunting for jobs that weren’t on it, and then seven in ten of them left. In contrast, the new page does its work in forty seconds and hands off.

The old careers page was a doorway to Workday: a candidate clicked a job title and left the site to read it. The whole visit averaged under two minutes and produced no journey data at all. Candidates now spend nearly that long on a single job page, on workhuman.com, with the team, benefits and location content one tab away.

The audience didn’t change size. September drew 2,110 people to the careers homepage; October to December averaged 2,013 a month. Same traffic, two and a half times as many of them going somewhere next.

What these numbers don’t say. Open Positions didn’t exist before October and recorded zero pageviews in August. So this isn’t a page that got better but it’s a step in the journey that moved on-site, and traffic to it is new by definition. Part of September’s exit rate is that same structure: the only next step was Workday, and leaving for Workday is an exit. What I’d defend as the design outcome is that there is now an on-site next step, that Open Positions became the most-entered page on the whole surface, and that the candidates who open a job stay on it for close to a minute and a half. One note on method: HEAP reports this page’s exit rate two ways, and both captures are here: 72.4% to 26.0% user-based in the funnel, 69.2% to 28.4% session-based in the page table. I’ve quoted the funnel throughout; the gap between the two is the counting, not the result.

October carried launch promotion behind it and November carried a LinkedIn push, so neither month is a clean read alone. The number I don’t have is applications — they live in Workday, outside HEAP, and I never got the join. That’s the real gap in this project and the first thing I’d instrument on the next one.

HX Total Careers Page Performance

Date range: September 1 – 30, 2025

HEAP page performance table, all careers paths, September 2025.

HX Total Careers Page Performance

Date range: October 1 – December 31, 2025

HEAP page performance table, all careers paths, October to December 2025.
Open Positions went from 12 session entries to 5,973; the careers homepage from 2m 13s and a 69.2% exit rate to 42s and 28.4%.
workhuman.com/company/careers/list/

Learning

Content is a design constraint, not a production step. The architecture assumed team profiles, benefit copy and video that didn’t exist yet, and content strategy arrived after the structure was set. The structure was right; the schedule paid for it. On anything content-shaped I now bring content in at architecture, not production.

Role Description sits first among the five tabs because I assumed it mattered most. I went back to the tab analytics. Role first was right, across seventy-three roles it outdrew the other four tabs combined, four to one. What I got wrong is the order after it; culture shipped third and finished last, at thirty-seven seconds of dwell against fifty-eight for the tabs either side of it. Given that 73% of sessions never touched the careers homepage, I can’t tell myself candidates got the culture story upstream. They just didn’t want it there. Meet the Team and Benefits are what a candidate opens when they’re deciding, and that’s the order I’d ship now.

Instrument for the business question, not the design question. I built the measurement plan around what HEAP could already see, pageviews on paths I controlled, and ended up with numbers that describe my own architecture as much as they describe candidate behavior. The question Marketing actually asked was which content drives applications. I should have started there and worked backwards to the events, even though it meant a Workday join I’d have had to fight for.

Credits

I owned design end to end — research, architecture, and every screen — inside a group that made the calls together: Amanda on positioning, David on what candidates actually ask and what Workday would allow, Cameron on everything the architecture assumed would exist. Where this case study says ‘we,’ the decision was the group’s.