More of the right freelancers, before the client asks
Toptal promises clients a top freelancer, fast. When the network ran short of React developers, that promise broke and the deal walked. I was the first designer on Toptal's Growth Experience team, and I led the squad whose pages brought freelancers in.
A shortage is a lost client.
A client arrives with a demand, Toptal promises someone quickly, and then nobody in the network has the skill. The gaps were always the same ones: React, JavaScript, most front-end work, product design.
Our pages had one job: be clear and convincing enough that the right newcomers start signing up. The old freelancer homepage couldn't do it, for two reasons. It took real work to change, so it couldn't follow the market. And it spoke to everyone at once, so a React developer never saw a page about React, or a reason to think they could make Toptal's top 3%.
- TemplateOne landing page per skill, launched without a designer.
- SearchEvery skill page one link from the homepage.
- FilterA one-minute screen before the sign-up, so volume didn't drown quality.
- SystemA component library under all of it, so the next page was cheaper than the last.
One page, every skill.
I designed the template, its blocks and the rules it runs on. The skill is a variable: it lands in the headline, the pitch and the button.
The PO fed in the data, so any PO could launch a page for a new skill without waiting for a designer. Toptal made as many as the market needed. The POs took over the routine launches, since they were watching the numbers anyway, and I spent design time on the pages that needed thinking.
One block came from legal. Search brings people in on brand names like Shopify, so every page states that the job is with Toptal's clients, not with the brand.
Every skill, one link from the homepage.
The bottom of the homepage reads like an index, and that's the point. Each line is a skill page.
Designers tend to want this block gone. It's long, and nobody would call it pretty. But a freelancer searching "Kotlin developer jobs" finds the page they need through it. The SEO team worked alongside us on it: skills grouped by field, the in-demand ones marked.
Growth that doesn't fill the funnel with no.
Toptal takes the top 3%. Every extra applicant is mostly an extra rejection, and a rejection costs the applicant's time and Toptal's screeners'. So we put a filter in front of the funnel.
The 1-Minute Talent Assessment sits on the homepage, ahead of the big ask. It checks English and a few hard skills, and gives a straight answer either way.
Not ready yet
We said what Toptal looks for, so the answer read as a plan rather than a rejection.
Ready
Congratulated, and sent into sign-up with a couple of steps already done.
Show the work you won't have to do.
Freelancers know the grind: portfolio, prospecting, consultations, invoices. The homepage draws it.
Seven steps on the traditional side, three on Toptal's. Below it, the proof in the order the doubts arrive: who the clients are, what freelancers rate Toptal, what they say, then the questions. Beside the pages sat a freelancer calculator, with an hourly rate and what it adds up to as hourly, part-time or full-time work. Its job was to make "good money" a number you could see.
Three carousels, tested on ten freelancers.
The Ultimate Freelancing Guide was a lead magnet, built for people who wanted in and needed to know what it took. Its articles sat in a carousel, and anyone who missed the navigation never got past the first card.
We tested three versions with 10 prospective freelancers, suggested by people already in the network: 5 developers, 2 designers, 2 product managers and 1 QA engineer. The task was to find the fourth article. Participants scored each version on a seven-point ease scale. Our PO ran the sessions.
- Prototype 0148
- Prototype 0257
- Prototype 0362
I have to scroll down the page to find the navigationOn 01 · arrows below the fold
It was easier to find the navigation but the arrows mix with the content…On 02 · arrows over the cards
This was the easiest to find.On 03 · arrows on the hero, in view
A library, so the next page cost less.
A template only scales if its parts do. I started the Public Pages Library in Sketch, then the team rebuilt it in Figma together.
It ran from atoms (type, sizes, corners, colour, spacing) up to whole components: buttons, forms, cards, floating elements. We documented each part on its own page: what it's made of, its spacing, how it behaves. We drew the visual refresh, a deeper blue with diagonal cuts, with the same parts.
Zero shortages within six months.
After years of constant shortfall, the network had no freelancer shortages within six months of our team taking over the pages.
That came from our work: we redesigned the pages and rebuilt them for search.
Andre is a designer who deeply understands the business context. He is interested in knowing the motivation of the users, in how they are using our product, and in what happens to them when they are no longer within the scope of our product.
Eugene Aleksandrov · Product Manager · Toptal
What I'd do differently
- Own the numbers as well as the pages.Our PO ran the experiments and I got the results. I can show four years of design and almost none of what it moved. Next time I'd keep a record of every test I designed for: what we expected, what happened.
- Get the system into Figma sooner.The library started in Sketch, and moving it took the whole team. Every month in between, we had two sources of truth.