The owner was paying 18% commissions to OTAs while their own website converted at 0.6%. The math told us the brand wasn't a hospitality problem; it was a distribution problem solved by software. Every guest the site converted was a guest the owner kept margin on.

We bet that the path to direct bookings ran through three things, in order: brand presence that matched the property, a checkout flow that didn't ask the guest twice, and content tuned for each guest segment. Photography would do most of the work. Software would carry the rest.

We started with a custom brand system tied to the property's physical materials — limestone, brass, cypress. Photography was reshoot from scratch, large-format and unhurried. The brand showed up on the site within four weeks.

The booking engine connects directly to Stripe with idempotent webhooks and a small inventory layer that syncs to the existing channel manager. We chose Sanity for the CMS so the owner could publish a new room or seasonal special without filing tickets. Hosting on Vercel with Cloudflare in front gives us sub-second page loads on global cold cache.

Marketing pages were segmented per audience — couples, families, business retreats — each with its own lede, photography, and offer. We A/B tested the booking CTA on every segment for the first sixty days and locked the winners.

The original plan didn't include lifecycle email. We added it in week six after seeing that 43% of inquiries didn't book the first time. A simple three-email sequence (offer, social proof, urgency) brought 18% of those back to direct booking within thirty days.

We also rolled back our second checkout iteration. We had built a multi-step booking flow that felt safer for the owner; guests preferred the single-page checkout we shipped on day one. Lesson: the simpler flow won, even on a high-trust transaction.

Ninety days post-launch, direct bookings doubled and effective OTA fees dropped from 18% to 6% of total revenue. The owner now has a working channel they actually own.