pendo.io homepage

Key decisions

Move the decision into the page

Play

One day into the build, leadership asked for a way to flip through the copy variants themselves. I built a review toolbar into the working page, modeled on the Vercel toolbar, with switches for each version, each section, and the hero. Every choice was written to the URL, so a link opened that exact page for anyone. My manager’s review said it “helped the team align earlier, and reduced the rework that often results when content decisions are made outside their design context.”

Let the copy set the layout

Enlarge image

The use-case section changed shape four times because the copy did. Short taglines got a card grid, and long copy got full-width rows. Our director of brand design asked for a bento grid. The day before launch the story settled on a four-stage lifecycle, so it went back to four rows. Every product shot frame got a fixed aspect ratio, so each crop lands on the same part of the UI at every width.

Give the motion a job

Enlarge image

A designer gave me a wide waves graphic with no direction, so I built a lateral loop, a parallax, and a scroll-linked version and compared them on the page. Scroll won, and then came out on launch day because leadership felt the hero had too much motion. The word-by-word headline entrance stayed, since it pulls the eye to the headline when you land.

Borrow the interaction, then make it ours

Play

For customer quotes I borrowed Stripe’s logo tabs, then tuned them until they felt like ours. I sized the logos by hand so they looked equal, evened out the type-in speeds so a two-line quote and a five-line quote felt the same, and fixed the section’s height so it didn’t jump between quotes.

End every use case in a demo form

Enlarge image

Each use case opens a large product modal with an embedded demo form next to the opening pitch. Forms are where pages usually look broken, so the form holds its exact final height while Marketo loads, tells sales which use case the visitor came from, and confirms in place.

First output vs. what shipped

Headless Pendo

Play

Headless Pendo means using Pendo’s data from inside AI tools like Claude, ChatGPT, or Cursor. We started with a screen recording of one tool, and I proposed recreating each tool’s interface in code with short animated conversations.

What we had	What shipped	Why it changed
The agent’s first build: one chat window re-skinned per tool, with branded bubbles	AI replies as free-flowing text, the way each tool shows them	The bubbles didn’t resemble how these tools actually look
Claude’s “allow Pendo” panel copied into every tool	Each tool’s real MCP connection step, matched from screenshots	It was wrong for ChatGPT and Cursor
Fin as a support chat	Fin stepping in on a travel booking site as a visitor clicks around	Fin’s value is acting proactively
Each conversation ending on an answer	A follow-up action, like Cursor turning off a broken guide through the Pendo API	An action shows Pendo actually doing something

Outcome

The homepage launched on August 4, 2026, in six languages. Within a day, sales was asking for demo videos in the Headless Pendo style. Since launch, about 59% of the interactions tracked in the page’s sections happened in Headless Pendo, more than the other sections combined. Traffic dipped after launch, but demo form submissions held steady on about 5% fewer visitors. A month later the same system carried the Pendo for Agents pages.

Process

The page started as a sandbox route separate from the live homepage. I wrote more than 900 prompts across about 60 Cursor chats, and bigger changes started as a written plan I reviewed first. Trying an idea only took minutes, so I tried a lot of them and threw most away, including a shimmering text mask, decrypting stats, and a rotating wheel for the use-case tabs. The last two weeks went to production work: gradients moved into code, a code review, six languages, and memory fixes after launch.

Reflection

I’d agree on the limits earlier. Next time I’d put the motion in front of decision makers early, at full strength, and set a memory and load budget on day one. The idea I most want to push further is one that didn’t ship: a concept-car homepage that changes for returning visitors, customers, and Free users.

Enlarge image