How RouteTidy calculates

This page explains, in a way you can check, where the numbers in the planner come from, which methods are behind them and where their limits are. The sample routes on this website are measured, not estimated.

Where drive times come from

The basis is the OpenStreetMap road network. Driving times and distances by car come from openrouteservice (HeiGIT gGmbH), with Geoapify as a fallback. For each route we fetch a matrix: drive time and distance from every stop to every other stop.

The matrix can differ by direction, so one-way streets and turn restrictions count. Live traffic is not included; congestion and road work are handled by your navigation app while driving.

How the order is calculated

  • Up to 12 movable stops (not counting the start and a fixed end), RouteTidy calculates exactly: a dynamic programming method (Held-Karp) checks all orders in a combined way and is guaranteed to find the best one for drive time or distance.
  • With more stops that would take too long. RouteTidy then starts with the nearest next stop and keeps improving the route until no improvement is left: reversing sections (2-opt) and moving chains of one to three stops to a better place (Or-opt). The result is very good, but not guaranteed to be the best.
  • With time windows, the first goal is as little lateness as possible, then drive time or distance. The search starts from two solutions (best order without windows, sorted by window end) and is time-limited – again an approximation.

Schedule and arrival times

  • Arrival = departure + drive times + time at the previous stops. If the route arrives before a time window opens, the planner assumes waiting.
  • Arrival links show customers a range of ± 15 minutes around the calculated arrival, rounded to 5 minutes and never before the departure. The vehicle is not tracked.

Splitting and the check list

  • To split, the planner sorts stops by their direction from the start, cuts at the largest gap and divides them into equal parts. It then swaps stops between routes as long as that makes the areas more compact. This runs in the browser on straight-line distances, without a request to the routing service.
  • With a capacity, it cuts the parts by quantity instead of by number of stops: a route is full when the next quantity would exceed the capacity or the route has its share of the total; stops without a quantity count as 1. Swaps then only happen as long as every route still fits. If the chosen number of routes is not enough, the planner tells you how many you need.
  • When you re-optimize the rest of a route in progress, the start and the stops already done stay at the top; the planner optimizes from the last stop done across all open stops, with the same calculation as always. Arrival times still count from the planned departure, not from the actual time of day.
  • The check list flags addresses where the search found a different house number, and stops that are more than 16 mi and more than four times the usual distance away from the center of the route – typical of a same-named street in another town.

How we measure the sample routes

Every sample route uses fictitious stops: random points with a fixed seed around a place, so the measurement can be repeated. Points that cannot be reached by car or that lead implausibly far by road are drawn again.

“Before” is the order of the list, the way orders often come in; “after” is the order optimized for drive time. Both values come from the same provider matrix. Across 19 routes with 6 to 20 stops (measured October 30, 2026 – November 11, 2026), the median route got 44% shorter and 38% faster; the range runs from 23% to 64%.

We report the median because a few very unsorted lists would pull the average up. If you already sort your list well by hand, you will save less.

Limits

  • No live traffic, no time-based driving bans
  • No road restrictions by weight or vehicle height; load only as a capacity when splitting
  • Splitting by number of stops or quantity, not by working time

All measured sample routes

Each row links to the page with map, order and measurement date. The “Cities” rows were added later and are not part of the 19 routes above.

Use casePlaceStopsBeforeOptimizedSaved
Delivery drivers Denver 18 75.9 mi 27.0 mi 64%
Couriers Chicago 12 55.8 mi 22.4 mi 60%
Beverage delivery Atlanta 14 58.0 mi 33.6 mi 42%
Meals on wheels Columbus, Ohio 12 39.1 mi 18.6 mi 52%
Produce boxes and farm deliveries Portland 16 157.0 mi 57.2 mi 64%
Furniture delivery Kansas City 7 132.8 mi 68.3 mi 49%
Home health and home care Minneapolis 10 30.9 mi 19.7 mi 36%
Home-visit therapists Indianapolis 8 28.6 mi 16.1 mi 44%
Contractors and service techs the Phoenix area 7 45.0 mi 31.8 mi 29%
Cleaning services Charlotte 9 50.1 mi 28.5 mi 43%
Lawn care and landscaping the Raleigh area 9 57.5 mi 39.6 mi 31%
Property maintenance San Antonio 9 71.1 mi 45.1 mi 37%
Chimney sweeps Boston 14 75.0 mi 35.7 mi 52%
Field sales the Dallas area 10 125.5 mi 73.4 mi 42%
Real estate agents Nashville 7 50.0 mi 30.2 mi 40%
Home inspectors and appraisers Salt Lake City area 6 86.3 mi 66.7 mi 23%
Flyer distribution Boise 20 32.3 mi 15.3 mi 53%
Campaign signs Oklahoma City 16 61.1 mi 24.1 mi 60%
Food banks and food rescue Pittsburgh 11 58.8 mi 30.8 mi 48%
Cities Los Angeles 15 91.4 mi 53.5 mi 41%
Cities Houston 15 107.2 mi 58.3 mi 46%
Cities Austin 15 105.3 mi 40.5 mi 61%
Cities Philadelphia 15 95.3 mi 41.5 mi 56%

Questions

Is the optimized order really the best one?

Up to 12 movable stops, yes – the method is exact. Beyond that and with time windows it is a very good approximation; the planner tells you which case applies.

Why do the times differ from my navigation app?

The planner calculates without live traffic. Your navigation app accounts for congestion and road work while driving; the order still makes sense.

Can I check the measurements?

Yes: every sample route names its place, measurement date and provider; the table above lists them all. The stops are fictitious, the road values are real.

Plan a route · Savings calculator · What’s new