Group tour quotes in 20 minutes, down from three days

The situation#

Group travel is an assembly problem. A single enquiry for thirty people across eight days combines hotels, transport, guides, entry tickets and meals, sourced from suppliers who share nothing and answer email at their own pace.

Consultants did it by hand. Check availability across supplier inboxes, price each component in a spreadsheet built from last year's spreadsheet, assemble an itinerary document, format it, send it.

Three days at median, longer in season. The operator had assumed they were losing on price. Reviewing lost enquiries showed something else: customers had contacted four operators, and the first credible quote set the itinerary vocabulary and the reference price that everyone else was then compared against.

What we did#

Weeks 1–2 — separated assembly from judgement. We sat with the consultants and mapped what a quote actually involves. Roughly eighty per cent was repeated assembly — availability, component lookup, formatting. The remaining twenty per cent was genuine commercial judgement about margin, supplier relationships and how hard to push on a particular booking. Only the first part was a candidate.

Weeks 3–7 — built the quoting engine. A component catalogue with supplier rates, availability checks against the suppliers who had any interface at all, itinerary assembly, and a formatted document generated from the result. Suppliers with no interface stayed manual, and the system was explicit about which components were confirmed and which were provisional.

Weeks 8–12 — the booking side. Holds, partial and staged payments across a group, amendments, and the cancellation logic that group bookings need when four of thirty people drop out three weeks before departure.

Weeks 13–14 — load and launch. Season traffic modelled and tested, with the supplier-facing integrations given timeouts and fallbacks so a slow supplier degraded one component rather than the whole quote.

The boundary we kept#

Pricing stays with a consultant. The system assembles the itinerary, applies standard rates and presents a suggested total; a person sets the final number before it goes out.

That was the operator's decision and we agreed with it. Margin on group travel depends on supplier relationships, repeat-customer history and how much the operator wants a particular booking — none of which is in the data, and a model optimising a suggested price would have been confidently wrong in exactly the cases that matter most.

What we would do differently#

We built supplier integrations before establishing which suppliers actually mattered.

Three weeks went into interfacing with a supplier group that turned out to represent under four per cent of bookings, because they had the best API and were therefore the easiest to start with. Meanwhile the two suppliers behind roughly half of all itineraries had no interface at all and were handled last.

Starting from booking volume rather than integration convenience would have delivered the same value in half the time. We now rank integration work by share of transactions before looking at how easy each one is to build, which sounds obvious and was not what we did here.

Where it went next#

The operator moved to a managed retainer. Direct booking for their standard packaged tours launched the following quarter, which removed a further tranche of consultant time on enquiries that never needed a human.

The remaining manual supplier components are now tracked with a cost attached, which has given the operator a concrete number to take into supplier negotiations — several have since provided a feed rather than lose the volume.