The Hall of Fame
18 years of rankings, one page that tells the story
Client
Glassdoor
Role
Product design and art direction
Team
PM, 6 engineers, SEO, content design, brand, Claude (AI design partner)
Timeline
2023–2026 · Hall of Fame in under 4 weeks
Live page
Overview
Glassdoor’s Best is Glassdoor’s awards program: Best Places to Work, Best CEOs and Best-Led Companies, all chosen from anonymous employee reviews. I’ve designed its landing pages, winner cards and campaign assets since 2023. In fall 2026, the PM brought me an idea and a rough prototype: one page for 18 years of Best Places to Work history. Working with Claude as a design partner, I had concepts ready for review within days, and the page launched less than four weeks later.
26 days
from first concepts to launch
18 years
of rankings on one page
<1 week
from wireframe to working prototype
Source: Glassdoor Best Places to Work Hall of Fame, launched October 26, 2026.
01
Problem
Glassdoor invented Best Places to Work, but 18 years of winners sat on scattered pages, so search engines and AI tools credited other sites for Glassdoor’s own award.
Each year’s list lived on its own page, and AI answers cited the award through Investopedia and CNBC instead of Glassdoor. As the PM’s brief put it: “The problem isn’t interest. It’s discovery.”
One canonical page. Replace 18 scattered pages with one evergreen URL that search engines, AI tools and the press would cite.
A chart that grows every year. Each January adds a column, and most visitors are on phones.
No bad news for winners. Employers are customers, so a gap in a streak can’t read as a decline.
02
Making 18 years of rankings readable
The PM’s prototype proved the idea. My job was to make it work for 18 years of data, 100 companies a year and a phone screen. Seven decisions shaped it:
Let the chart grow without scrolling. Each January adds a year, so I added a “Last 10 years” switch. It’s the default on mobile, where it always fits.
Never make a winner look bad. Employers are Glassdoor’s customers, so a missing year shows as a neutral gap, never a “broken streak.” Our general manager cut an editorial module about ended streaks, and the chart’s wording follows the same rule: “Last on the list in 2022,” not “dropped off.”
Prototype on real data. Building with all 18 years of lists exposed a data error before launch: the Top 5 changed.
Make #1 read as a win. A plain yellow dot read as a data point, so I added a trophy for when ranks are hidden. I tested three sizes and weights for the rank dots and chose 20px with semibold numbers. The trophy dots are 20px too, so turning ranks on or off never shifts the layout.
Choose contrast over tidiness. Joining the gray “no longer on the list” bar to the green looked cleaner, but the two measured 1.02:1, nearly the same lightness and hard to tell apart for colorblind users. I tested four fixes, and the team then dropped the gray state entirely. The legend now has just three states: on the list, ranked #1 and not on the list.
Give everyone the same words. An anatomy panel named every part of the page and chart, which sped up reviews with engineering, SEO and content.
Spec the motion. Ribbons draw in, #1 dots pop and totals count up, with timing, easing and a reduced-motion version written down for engineering.
How Claude and I built it in under four weeks
The PM brought me a rough, wireframe-level prototype and a short brief. Less than four weeks after my first concept review, the page was live, and much of that speed came from working with Claude as a design partner at every stage. I set the direction, judged the work and made every call. Claude did a lot of the heavy lifting:
Wireframe to working prototype. Claude Code turned the wireframe into a mockup, then into a responsive prototype running on all 18 years of real data, in under a week. That prototype became engineering’s reference. I set the rules and caught where it drifted, like keeping it on our Jewel design system.
Researching the chart. Claude researched and compared six ways to show 18 years of rankings: streak ribbons (a Gantt-style chart), a bar chart race, a bump chart, a rank-tier heat map, story cards and a simple rank table. Weighing them side by side led to the streak table we shipped.
Documenting the decisions. Claude drafted the pros-and-cons write-ups behind each decision, the interaction statements, and the anatomy that named every part of the table. That shared vocabulary sped up every review with engineering, SEO and content.
Handing off to engineering. Together we wrote the engineering spec and the Jira ticket descriptions, down to the technical detail developers needed to build the streak table: motion timing, easing and reduced-motion behavior. That level of front-end detail usually takes an extra round with engineering; this time it was in the first handoff. Engineering built the page in about two and a half weeks.
Building the program
Hall of Fame sits on three years of earlier launches I designed: Best CEOs, Best Places to Work and Best-Led Companies (with winner descriptors Claude wrote under the direction of our product marketing director), all under the Glassdoor’s Best identity from 2024.
Six ways to show 18 years of rankings. Streak Ribbons won because it states the streak instead of implying it.
Ribbons draw in, #1 trophies pop and totals count up, once per visit, with a reduced-motion version that shows the finished chart.
Selecting a company opens its record in place: key stats, its rank every year and a summary written to be quoted.
03
Outcome
Launched October 26, 2026, less than four weeks after my first concept review. Engineering built it in about two and a half weeks.
One page for 18 years. Every Best Places to Work list since 2009 now lives at a single URL built to be cited. See it live on glassdoor.com.
What I’m measuring next. First-month traffic, rows opened and View Jobs clicks in Amplitude, plus whether search engines and AI tools start citing this page instead of other sites. The first month closes in late November 2026, and I’ll add the numbers once they’re approved.
04
What I learned
A prototype on real data settles debates. Static mocks hid problems that real rankings exposed in minutes, from a wrong Top 5 to dots that collide on a phone.
Pick one source of truth. Moving fast, Figma and the prototype drifted apart, and engineering noticed first. Next time I’d name the prototype as the source from day one.
AI gathers and builds; people decide. Claude built prototypes, found photos and drafted descriptors. Each time, a person set the bar and made the call, and once that meant setting the AI’s work aside.
