200+ pages, +34% conversion.

The challenge

I owned more than 200 landing pages at Chegg, built on a headless CMS in Contentful with a media pipeline running through Cloudinary. The system was large, the traffic was real, and the stakes were commercial. But the pages shipped on instinct. Someone had a strong feeling about a headline, a hero image, or a button, and the page went live. When conversion moved, up or down, nobody could point to the change that caused it. We were flying without instruments on a property that mattered to the business.

That left every debate stuck as an opinion. Design wanted one layout, marketing wanted another, and there was no evidence to settle it. Decisions came down to who argued hardest or who had the most seniority in the room, not to what the pages actually did for the people visiting them. Across 200 pages, that guesswork compounded into a lot of wasted effort and conversion we were quietly leaving on the table. The scale that should have been an advantage was working against us, because a bad instinct copied 200 times is a very expensive habit. I wanted a way to replace conviction with proof, and to do it at the pace the business needed.

What I did

• Designed and ran a full A/B testing program on Optimizely, so changes to the landing system got measured against a control instead of shipped on faith.
• Started every change as a written hypothesis. Before anyone touched a page, we said what we believed, why, and what result would prove or disprove it.
• Tested UX, content, and layout in tight iterations, keeping the loop short so we learned fast and moved to the next question without stalling.
• Produced the designs and copy directly when the design team was stretched, so a promising test never sat in a queue waiting on someone else's bandwidth.
• Rolled winning changes into shared components and templates, so a win on one page propagated across the system instead of staying a one-off.
• Worked with engineering to take product ownership of the experimentation and martech layer, turning the program into durable infrastructure rather than a side project.

The outcome

The program lifted conversion 34% across the full 200-plus page system. Because winning changes flowed into shared components and templates, each proven improvement kept paying off on every page that used them, so the gains stacked instead of staying isolated. A test that started as a single question about one page ended up raising the floor for dozens. Just as important, the arguments stopped being about opinions. When someone had a view, the answer was to test it, and the data decided. That changed the culture as much as the numbers, because it gave everyone a fair and fast way to be right.

The work earned me a promotion to Senior PM for the martech stack in September 2022. I took product ownership of the experimentation and martech layer alongside engineering, so the testing program stopped being something I ran and became something the organization owned. I carried the same habit into everything that followed: write the hypothesis first, measure the result, roll the winners into the system so they compound, and ship with my own hands when that is faster than waiting on someone else.