TicketLounge

People rarely search the way a database expects. Someone might type "rap concerts this weekend near me" or "seats behind home plate under fifty dollars," and the older approach would often come back empty or, worse, show the wrong team or the wrong night. Getting it to understand loose, everyday language and still land on the right handful of events was the hard part.

We built a concierge that reads the intent behind a request and searches several ways at once, then ranks what actually fits. When an exact match is thin, it loosens one thing at a time, say widening the dates, and tells the person what it changed instead of quietly showing something off target. If it genuinely has nothing good, it says so rather than padding the list with filler.

Ticket prices and what's still available change constantly, and the catalog is large enough that you can't refresh all of it in one pass. We needed to keep everything current with many small jobs running on a schedule, but running several at once tended to make them collide or slow the live search to a crawl.

We changed how the refresh jobs pick up work so each one quietly claims its own batch and skips anything another job already has, which lets several run side by side without stepping on each other. We also separated the "when did we last check this" bookkeeping from the data search depends on, so keeping things fresh no longer competes with people actually searching. Each run also knows to stop before it hits its time limit and picks up where it left off next time.

When someone pays, we charge their card through Stripe, but the actual tickets come from a separate supplier, and that second step can be slow or fail after the card is already charged. Payment confirmations can also arrive more than once or out of order, so the same purchase could accidentally get processed twice. Getting the money, the tickets, and our own records to always agree was genuinely tricky.

We rebuilt the payment flow so a confirmation is only ever acted on once, no matter how many times it arrives or how the messages overlap. If the supplier can't complete the purchase after the card was charged, the system refunds the customer automatically instead of leaving them paying for nothing. And before any order goes through, we re-check the live price and availability so nobody is charged the wrong amount or sold a ticket that just disappeared.

A stadium seat map is a dense, awkward graphic, and we wanted people to do more than stare at it: tap a section to pick it, pinch to zoom, and see at a glance which areas are cheaper or pricier. The raw maps we receive are just shapes with no easy way to know where each section sits or how to frame it when someone taps, so making the map feel responsive and readable took real work.

We worked out, from the raw shapes alone, exactly where every section lives so a tap can zoom straight to it and snap back out cleanly. We color the sections by price so the cheaper and more expensive areas read instantly, and the map resizes smoothly as you drag the ticket list up and down. The result is a map you can explore and understand at a glance instead of squinting at a flat picture.

We build the iPhone app, the Android app, and the website from a single shared codebase, which keeps everything consistent but means a problem on one platform can be invisible on the others. A couple of the nicer visual touches turned into real traps: an animated background that froze the iPhone app whenever you moved between screens, and a frosted glass look that ate memory on iPhone and made scrolling stutter on Android.

We kept the shared codebase but gave each platform its own version of the tricky pieces, so every device gets the smooth, native version that suits it. The animated background was rebuilt a different way on iPhone so it no longer hangs, and the glass effect now uses Apple's built-in version on newer iPhones with lighter fallbacks elsewhere. People get an app that feels at home on their device, and we still ship from one place.