One start page now serves several communities on a secure network. Leadership picked the design themselves, round by round, and they were excited when it shipped. Here's how that happened, and why the hardest part was not the code.
The problem: too many front doors
Users on our network work with half a dozen tools: a virtual desktop, source control, tickets, documentation, file sharing, and the help desk. Each tool has its own web address. New users didn't know where to start, and long-time users kept a pile of bookmarks.
The fix sounds simple: one page with a button for each tool. But each community on the network wants that page to feel like theirs. And the people who approve it each have their own taste.
Design is its own discipline
I built the first page for one community, and it was well received. Soon after, a second page was built by someone else for another group. Comparing the two taught everyone something: good design takes more work than the code and the infrastructure. Hosting a page is easy. Making people want to use it is not.
My manager then asked me to build pages for two more organizations.
The real challenge: subjective preferences
Once several leaders are giving feedback, design becomes a conversation about preferences. One person wants it bolder, another wants it calmer. Feedback on one page quickly turns into feedback on the organization's brand as a whole. Nobody is wrong, because it's all taste.
Trying to win that argument with one "right" design is a losing game. So I stopped arguing and started offering choices.
The approach: rounds of options
Two wireframe layout options: a card grid and a sidebar with search
Every round started from the same small UI framework. The page always has the same parts: the consent banner, a header with the community's logo, and one button per tool. Only the layout and the look change. Option A puts the tools in a grid of large cards. Option B puts them in a sidebar with a search box. Because the parts stay the same, each new option is a change of layout and colors, not a rebuild, and leadership can compare options side by side.
I worked with leadership in a group chat, in rounds:
- Ask first. Before designing anything, I asked what they liked and didn't like about what they'd already seen, including colors.
- Show directions, not a final answer. The first round had several very different layouts and color palettes, from bold and animated to plain and quiet.
- Narrow and vary. They reacted, and I took the favorite and posted a handful of variations on it.
- Repeat until one wins. Each round got smaller and more specific, until one design was the clear favorite.
It turned into almost a game. People enjoyed seeing their ideas show up in the next round. By the end, the design wasn't "mine" anymore. It was the vision they'd been describing, made visible. That's buy-in you can't get from a slide.
The build: simple on purpose
The creative side only works if the technical side never gets in the way. So I kept the build as simple as I could.
One self-contained HTML file per design. Styles, scripts, and images all live inside the file. There are no outside libraries, fonts, or calls to the internet. A reviewer can open it by double-clicking. Security can read the whole thing in one sitting. It works the same on a laptop, a file share, or a web host.
Static hosting on Amazon S3. There are no servers to patch and no runtime to harden. Deploying means uploading a folder. A new community gets the same page with a different logo and color palette, and the rest is reused.
Reference architecture: users reach an internet-facing Application Load Balancer in public subnets, which forwards through an S3 interface endpoint in private subnets to a private S3 bucket
Where the page lives. The diagram shows the placement:
- Users reach the page through an internet gateway.
- The gateway sends them to an internet-facing Application Load Balancer in public subnets. It is the only public entry point, and it serves HTTPS only.
- The load balancer forwards requests to an Amazon S3 interface VPC endpoint in private subnets.
- The page itself sits in a private S3 bucket. The bucket is never public. The only way in is through the load balancer and the endpoint.
The page only links to the tools. Each tool stays in its own enclave with its own access controls. This is a common AWS pattern for static sites behind a load balancer, and it kept the security review short.
Security built in, not added later. Government systems require a notice and consent banner that users must accept before they reach anything. The page shows it as a blocking screen on load. It fails secure: if JavaScript is turned off, there is no way past the banner, so no access is granted.
Accessibility from the start. The layout scales to desktop, tablet, and phone. If a user's device asks for reduced motion, the animations stop. Motion is fun until it bothers someone, and respecting that setting costs one CSS media query.
Write down the how-to. Each package ships with a plain-language README: what the page is for, what each button does, and how to deploy it. Before, only a few people knew how to do this. Now anyone on the team can.
Getting it through change control
On a secure network, a page full of links still goes through change control. Enterprise, project, and leadership groups all need to agree. Two things made that faster:
- Leadership had already chosen it. The design question was settled before the change request, so the review was about security and hosting, not taste.
- There was little to review. One static file with no dependencies, on a hosting service the network already approves, is an easy change to say yes to.
Results
- Approved and shipped, with the design leadership chose.
- Leadership was excited, because the page reflected what they'd asked for.
- The pattern spread to three or more communities. Each one gets its own branding on the same base.
Retrospective
Turn opinions into choices. When feedback is subjective, more argument doesn't help. Give people options and let them choose. Rounds keep each decision small.
Ask what people don't like first. It tells you more than "what do you want?" And it lets people be heard before they see a single design.
Make it fun. People engage more when they can see their ideas in the next round. Buy-in follows engagement.
Keep the tech boring. One file, static hosting, and security built in meant the conversation stayed on design, where it belonged.
What I'd do differently: set up a shared color-and-logo kit for each community at the start, so a new brand is a ten-minute swap, not a redesign.
How do you get buy-in on design when everyone has an opinion? Tell me in the comments.
Next in the series: from vague answers to a diagram everyone signs.



