SimilarTours Is Live: Why We Built Yet Another Travel Site

Somewhere out there, right now, a perfectly reasonable person is trying to book a Colosseum tour. They typed six letters into a booking site and got back 1,232 different ways to pay for the same building - we counted, in our price study. They now have nine tabs open, three of which look identical, two of which they suspect are identical at different prices, and it is 11:40pm the night before the trip.

We built SimilarTours for that person, mostly because we kept being that person.

As of today, SimilarTours is officially live: one search across more than half a million tours, tickets, and activities from four major booking sites, lined up side by side so you can compare them the way you'd compare flights. This post is the story of how it got here - including the parts that went wrong, because launch posts that skip those are no fun to read.

It started as a weekend experiment

The first commit landed in late November 2025. The plan, if you can call it that, was to spend a weekend seeing whether the idea held up: take the big tour booking sites, pull the same experiences from each, and put them next to each other on one screen.

The weekend got out of hand. By the end of November there were 671 commits, a working prototype, and a color scheme we can only describe as “aggressively teal.” It looked like an internal analytics dashboard, because that's essentially what it was - a tool for staring at tour data, built for us, not a site a traveler would ever want to plan a trip on.

That version taught us the two things that mattered. First, the core hunch was right: the same experience really does show up on multiple sites, sometimes at different prices, and nobody was putting them side by side. Second, the thing we had built was not the thing anyone needed. Useful lesson. Slightly expensive way to learn it.

Then we did the healthiest thing a side project can do: nothing

December managed 42 commits, and then the repository went quiet for four months. No dramatic pivot story, no burnout memoir - life happened, and the teal dashboard sat there, judging us.

The silence turned out to be the best product decision we made. When we came back in May 2026, we had enough distance to admit that the prototype was a proof of concept, not a foundation, and we rebuilt from scratch: new stack, new data model, and a rule that every feature had to work across every partner site from day one rather than being bolted to one and retrofitted later. By mid-May the first booking-site partnership was live with real inventory. The other three followed over the next six weeks.

Since that first commit, the project has crossed 1,200 commits. A respectable number of them are titled some variation of “fix”.

What SimilarTours actually is

The simplest way to say it: SimilarTours is a comparison site for experiences. You search for a place or an activity - “Colosseum”, “Barcelona food tour”, “day trips from Tokyo” - and we show you the matching options from the top booking sites in one list, with their prices, durations, ratings, and review counts. When you find the one you want, you click through and book it on the booking site itself, at its normal price. We add no fees and we never touch your payment; we're the step before the booking, not the booking.

A few numbers, verified this week, to make it concrete:

• 504,796 bookable experiences in the catalog, across four major booking sites
• Thousands of destinations worldwide - 5,883 of them at the last full count, from Rome and Paris to towns you'd need a good reason to name
• 14 languages, including one that reads right to left, which required flipping the entire interface like a pancake

The feature we're proudest of is the one the site is named after. Every experience has a Show Similar button: press it, and we compare that listing against the rest of the catalog and pull up its closest matches across all partner sites - same landmark, same style of tour, different operators and prices. In our price study, the median listing for a major landmark cost roughly 2 to 3.5 times the cheapest tenth of options for that same landmark. Show Similar exists to close that gap in one click instead of one evening.

Things that went wrong, a partial list

Every launch post says the journey was hard. Few say how, which is a shame, because the specifics are where the entertainment lives. A selection:

The $300 image bill. Half a million tours means half a million-plus product photos. Our hosting platform offers automatic image optimization, which is lovely at normal scale and ruinous at catalog scale - it cheerfully re-processed everything, and our first real bill arrived at around $300 for what was, at the time, a website with almost no visitors. We now serve partner images directly from the partners' own servers, optimized by exactly nobody, and the bill went back to pocket change.

The alphabet ate Rome. Country pages show a country's top destinations, and early on the list was capped at 30 cities - sorted alphabetically. Which meant the Italy page proudly listed Ancona, Assisi, and Bari while silently dropping a little-known Italian city called Rome. Japan lost Tokyo. France lost Paris. Nobody noticed for longer than we'd like to admit, because the pages looked perfectly plausible - full of real Italian cities, just not the ones anyone was searching for. Destinations are now ranked by how much travelers actually visit them, and there's an automated check that yells at us if a major city ever falls off a page again.

We took ourselves down with our own blog. This blog publishes in all 14 languages, which multiplies into nearly two thousand pages. During site builds, every one of those pages politely asked our backend for live tour data - all at once. Our own blog was, in effect, stress-testing our own infrastructure into failure several times a week. The blog now reads from a snapshot, and the backend has stopped flinching whenever we publish an article.

None of these made it into the brochure version of the story. All of them made the product better, mostly by forcing us to design for half a million products instead of the dozen we tested with.

How we make money, the honest version

Comparison sites live or die on trust, so here is the model with no varnish: when you click from SimilarTours to a booking site and complete a booking there, that site may pay us a commission. The commission comes out of their marketing budget - the price you pay is the same as if you'd typed their address yourself.

Two things follow from that, and we consider both features. First, using SimilarTours costs you nothing extra, ever - no fees, no markups, no premium tier. Second, we earn the same way from every partner, which means we have no reason to nudge you toward any particular one. Our incentive is embarrassingly aligned with yours: help you find the right experience at the right price, so you come back for the next trip.

What we will never do is inflate a price, hide a better option, or rank listings by who pays us more. The day a comparison site does that, it stops being a comparison site and becomes a billboard.

What's next

The catalog keeps growing - the four partner sites add and retire products daily, and we reconcile the whole catalog every night so you don't land on a tour that quietly stopped existing. Beyond that, the roadmap is shaped by one question: what else can we compare for you that you're currently comparing by hand at 11:40pm?

More partners, smarter matching between near-identical listings, and better answers to “is this actually a good price?” are all on the list. If you try the site and something annoys you, we genuinely want to hear it - the alphabet-ate-Rome bug was the kind of thing a stranger's email would have caught in a day.

Common questions