← All posts

Don't Let Your Website Promise More Than You Deliver

By Build Log

TL;DR: Don't put anything on your website that your team can't deliver. At my tour company, that's standard operating procedure. Today a bot found a side-project website that promises more than its booking bot can do. A test bot also caught a privacy bug before any real user saw it.

What got built

  • Yesterday's Build Log passed SEO review. I gave my go, and it went live tonight.
  • kayhonig.com was health-checked after the hosting move. It's fine.
  • The inbox bot archived a large backlog of old mail. Nothing important was deleted.
  • Routine market checks ran: competitors, reviews, ports, weather and sargassum. Only minor changes.
  • The infra bot reviewed email-authentication reports. Legit mail passes. Unrecognized senders were flagged.

Why this question came up

On a side project, the builder bot found a gap. The website promises bookings and payment links. The bot behind it can't create them yet. Staff confirm those by hand. The fix is parked for now.

Probably plenty of travel companies promise more online than they deliver. I don't recall specific cases.

My tour company doesn't do it, and that's cultural. It starts with sales. We under-promise and over-deliver. It's standard operating procedure not to promise anything the operations team or the guides can't deliver. You can see it on our tour pages. It's part of why travelers don't trust tour operators and what we do about it. More on who we are in Cancún.

What changed technically

  • The booking-system bot fought session expiries all day. I signed it in this morning. It dropped mid-review and was logged out again by midday. Its own lesson: never open records in a new browser tab, because that likely kills the session. That's the bot's guess, not proven. A card is now on our developer's board to fix how the system handles the bot's sessions. I'm hoping it's done tonight.
  • I widened the own-visit analytics filter on all 4 properties myself. It's a manual console step, with no API. I'll check them against tomorrow's numbers.

Side projects

  • A split onboarding flow (a short form before a call, the full form after) went through 3 QA rounds in about 30 minutes. Round 1 failed on a real privacy issue: the full-form link still worked after submission and could leak through the browser. Round 2 fixed it. Round 3 confirmed the links die on submit. I'm glad a test bot caught it before any real user saw it. I didn't even set it up. Another bot did. Running quality checks is now standard when agents start websites. It saves time. If you don't think of it, you'll do it yourself, or your clients will find the bugs.
  • A chatbot answer-quality update went live in the morning after my OK. It fixed a wrong setup fact and a wrong promo line, and the demo bot no longer offers a follow-up it can't deliver. Smoke tests passed. A pricing page update also shipped.
  • A role-play test client finished intake end to end through website chat. Confirmations went out automatically.
  • About 3,680 requests turned into about 2 real humans after filtering. A test has 0 real conversations so far. It doesn't change what I measure. It changes my expectations. Thousands of page requests on a site that was indexed a day or two ago are not realistic. Building with AI still needs human oversight and human experience.
  • A messaging business verification was rejected a third time, with no reason given. Next step: ask support why.
  • On a printables side project, day 4 of one product per weekday is in review. One redesign is in progress with the design bot.

What I'd do differently

Ship at 80%. New websites are hardly visited, so 80% is good enough to go live. We correct the last 20% while it's live. (Related: 35 minutes of downtime on a new site.)

Still open: an old-hosting cleanup review, a production OK and a privacy-notice review for the onboarding change, the verification support case, and the parked booking and payment-link gap.

FAQ

How do I keep my website from over-promising? Make it a rule not to promise anything your operations team can't deliver. At my tour company, that starts in sales.

Is 80% good enough to launch a website? For a new site with almost no visitors, yes. Fix the last 20% while it's live.

Do I need automated QA for a small website? I'd say yes. If you don't run it, you'll do the checks yourself, or your clients will find the bugs.