Proposals in under an hour, and a new route before dinner when a pass closes

The situation

The operator sells walking holidays the old-fashioned way: a conversation first, then a proposal built by hand. A typical request reads like a letter. Two friends in their sixties, one with a knee that dislikes long descents, nine days in early September, “somewhere with lakes, nothing scrambly, and at least one rest day with a good view.”

Answering it meant opening the stage spreadsheet, checking which inns were open that week, adding up ascent and descent per day, confirming luggage transfer between the chosen villages, and finding the last bus in case a stage needed shortening. Then writing it up. It took most of a morning, and in spring the team handled dozens of requests a week.

The second problem arrived in season. When early snow closed a high pass or rockfall shut a path, the duty planner phoned inns, re-routed luggage and called walkers, often between 8 and 10 PM, after the guides’ reports came in. Nobody was doing it badly. There was simply no way to see the whole chain at once.

What we tried first

Our first prototype generated complete proposals straight from the brochure and a few hundred past itineraries. The prose was lovely. The routes were not: two of the first ten drafts sent walkers into a village on the inn’s closing day, and one placed a 1,100-meter descent on the morning after the longest day of the trip.

That told us the order was wrong. Language comes last. A route has to be feasible before anyone writes a sentence about it.

We also dry-ran sending re-plan suggestions straight to walkers. It was fast, and it was a mistake: in the test inbox, alternatives arrived before the inns had confirmed rooms. We switched to drafts that a planner approves, and that is how it has stayed.

What we built

A stage graph, not a stage spreadsheet. Roughly 180 stages became a routable network: distance, ascent and descent, surface, exposure, shade, inn and hut options, luggage-transfer cut-offs, and public-transport bail-outs for shortening a day.

Energy budgets per walker. Route Engine holds each traveler to a daily effort budget built from distance and climb, and places rest days where the cumulative load says they belong. The knee that dislikes descents becomes a constraint, not a note in the margin.

Drafts with reasons. Itinerary Studio turns the request into a draft proposal with a line of reasoning for every choice (“Rest day in the lake village: day four is the trip’s biggest climb”). Planners edit, reorder and rewrite; the engine re-checks feasibility after every edit.

Re-planning that waits for a person. Concierge watches the operator’s own closure reports, guide notes and weather data. When a stage closes, it drafts a revised route for each affected group, requests holds from inns by email, and puts the whole package in front of the duty planner. One approval sends walkers the new plan by WhatsApp or SMS, with luggage instructions changed to match.

What changed

The numbers are above. What the planners noticed first was different: a morning of arithmetic became twenty minutes of reading a draft and a longer, better conversation with the traveler. Proposals went out the same day more often than not.

Re-planning moved from the phone to a queue. On the evening that snow closed the season’s highest pass, four groups had revised routes approved and delivered before 7 PM, and the duty planner made two phone calls instead of eleven.

What we’d do differently

Clean supplier data before anything else. Inn closing days lived in three places, and they disagreed. We lost most of week three reconciling them. We now open every pilot with supplier truth.

Let guides define exposure. Our first exposure ratings came from map data. The guides corrected about a fifth of them in a single afternoon. That afternoon belonged in week one.

Measure per planner, not only per team. The median hid a spread: two planners sent drafts nearly as written, one rewrote almost everything. All three were right for their travelers; looking sooner would have helped us tune the drafting.