How Ombra chooses, calculates and prices
Ombra promises to explain every stop. This page explains the rest: where the places come from, how shade is estimated, what each ranking criterion weighs, and which prices the amounts are calculated from. The limits appear here alongside the calculations.
Where the stops come from
Every stop Ombra knows comes from OpenStreetMap, in an extract covering the whole of France. More than five hundred thousand places are kept from it: picnic tables, rest areas, shelters, toilets, drinking water, car parks, playgrounds, cafés, restaurants and bakeries.
A place enters the database if it offers at least one of those ten services. The three accessibility traits — wheelchair access, changing table, dogs allowed — then count towards the choice, but never suffice to admit a place: pharmacies, town halls and schools carry them as often as picnic areas, and treating them like the others would have turned Ombra into a directory able to suggest a pharmacy for lunch.
Left out are the places none of those services describes, the map tags Ombra does not read, and places drawn as areas when the service only makes sense at a point — a picnic table traced as a polygon does exist on the map, but its centre falls in the middle of a whole rest area.
A place the routing engine cannot attach to the network within a hundred and fifty metres is rejected rather than snapped to some arbitrary road. The detour shown beside it would no longer mean anything, and an unreachable rest area would appear to cost three minutes.
How shade is estimated
Shade is calculated, never measured. Three sources can answer, from the most reliable to the roughest, and the first one to answer wins. Whichever served is shown with each stop, together with its reliability.
- A shelter declared on site
- The map says there is a roof. The most reliable source, though not a certain one: a shelter may have been dismantled, or may cover only part of the place.
- The position of the sun and the mapped obstacles
- The height of the sun at the hour you pass gives the length of every cast shadow; the trees and buildings nearby give it its shape. The place is represented by a disc eight metres in radius, because nobody knows to the metre where the tables are set.
- The kind of ground around
- A safety net, for when nothing is finely mapped: a wood, a park, open ground. We then know the place sits under trees without knowing where the trees are, and the reliability shown says so.
- Sun set, or too low
- Below twelve degrees of elevation, the calculation stops. The length of a cast shadow is the height of the obstacle divided by the tangent of the sun’s angle: at five degrees, a twelve-metre tree casts a hundred and thirty-seven, far beyond the neighbourhood examined. Carrying on would conclude “no shade” at the very moment when everything is in shade. Every place therefore receives the same value, and shade stops deciding anything.
- No usable source
- When none of the three answers, Ombra writes it down and says so plainly beside the stop. An invented average would travel into the score without leaving a trace.
Heights missing from the map are lent: twelve metres for a broadleaf, fifteen for a conifer, ten when the species is unknown, two storeys of three metres for a village house. Dropping a building for want of a tag would amount to claiming it casts no shade at all, which would be further from the truth. A guessed height lowers the reliability accordingly.
When you require shade rather than prefer it, the threshold is sixty per cent of the site’s surface. Any lower, and barely shaded areas would pass for shaded — and that is the mistake you only discover once you are there.
What makes a stop’s score
A stop’s score is a weighted sum of six criteria, and of nothing else. Each one can be read separately, with its value and its weight, for the stop that was chosen as for every stop that was set aside.
| Criterion | Weight |
|---|---|
| The detourWhat the stop costs you in minutes, relative to the largest detour you accepted. The dominant criterion, because it is the constraint you state first and the one you feel most. | 0.30 |
| The shadeThe shaded fraction estimated for your hour of arrival. Removed when you do not ask for it, when the sun has set, and when the sky is overcast. | 0.25 |
| The arrival timeThe gap between the time you will arrive and the time you asked for, relative to the width of the window. | 0.15 |
| The services you wantedThe share of your wishes the stop covers. Services answering the same need count as one: a table, a café, a restaurant and a bakery are four ways to have lunch, and a village with all four does not have lunch four times. | 0.15 |
| The place’s recordWhat the map knows about the place: opening hours, services, accessibility. A complete record scores higher, and anyone may complete any record — enriching one is free and open to all. The day it became paid for, this criterion would be a ranking you could buy. | 0.10 |
| The quietThe distance to the nearest busy road: no score below thirty metres, full score beyond four hundred. The upper threshold stood at a hundred and fifty metres until it was measured — four Auvergne places out of five were past it, hence all tied, while a table set a hundred and sixty metres from a motorway took the maximum score. | 0.05 |
These weights are starting values, which the context adjusts. A criterion with nothing to say disappears from the score instead of counting zero there, and the weight of the others is recomputed so that totals stay comparable from one stop to the next. Its absence can be read in the score’s detail.
You can ask for the chosen stop to be compared with any of those set aside. The answer gives the difference, criterion by criterion, in the same terms as above.
Why no placement can be bought
Those six criteria are the only ones a score can hold. The list is closed in the code: a “partner”, “sponsored” or “featured” line cannot be written into it, and a test checks this on every run of the integration pipeline.
When your stops are being chosen, Ombra does not know which ones are run by a partner. A business can tell us its opening hours and complete its record: that earns it a better record, and you will see it written, but never a better placement.
Which prices the amounts are calculated from
The amounts shown rest on fixed prices, taken by hand on and frozen since. A price refreshed daily would make the cost of a saved trip move while nothing about the trip had changed: a plan reopened six weeks from now shows the same figure as the day you worked it out.
| Assumption | Value used |
|---|---|
| Assumed vehicle | A petrol city car: 6.5 l per hundred kilometres on the motorway, 5.2 on the road. You can declare another — diesel saloon, SUV, electric, campervan, motorbike. |
| Petrol | €1.75 a litre |
| Diesel | €1.65 a litre |
| Electricity | €0.25 per kilowatt-hour, between the socket at home and the rapid charger |
| Tolls | €0.10 per kilometre of tolled motorway, class 1 |
All of these approximations lean the same way, and deliberately so. Motorway fuel costs twenty to thirty cents more a litre than in the village, and the avoided trip is priced at the village rate. A city car burns less than the average of the fleet, so the saving calculated is smaller than yours. We would rather announce a cautious gain and hold to it.
The per-kilometre rate ignores which concession holder you would have crossed: the routing engine does not name it. The twenty French motorways that are free along their entire length are taken out of the calculation, though, and the Millau viaduct is added back separately — tolled in the middle of a motorway that is not, it is paid even when the rest of the trip is free.
What Ombra cannot yet tell you about time
The routing engine computes free-flow durations: it assumes you drive at the posted speed, with nothing taken off for village high streets, junctions and the seven o’clock tractor. On an Annecy — Solignac-sur-Loire, it announces four hours eight where Google counts five and a half. That whole family of engines is optimistic, by roughly a quarter.
The correction is written and waiting for its measurements. No free service returns the durations actually observed, and twelve reference routes still have to be read off by hand. Until then, Ombra shows its engine’s durations uncorrected, and we say so here rather than let you assume a precision they do not have.
This is why the home page gives a ratio and not a duration. Across ninety-one Auvergne trips, routed twice by the same engine — once with everything allowed, as any sat-nav would, once with Ombra’s avoidances — the motorway-free route takes 1.13 times longer at the median. A quarter of the trips exceed 1.30, and the worst approach 1.75, all of them in the Massif Central.
Two reservations about that ratio, worth stating before anyone quotes it. It holds for the Auvergne, where the motorway network is sparse and the relief forces the back roads into detours they would not take on flat ground; a measurement around Paris would say something else. And it understates: the engine’s optimism hits a back road harder than a motorway, whose posted speed is close to the real one. The true gap is therefore a little wider than 1.13.
Distance was measured on the same sample, and it is the more surprising result: the motorway-free route was the shorter one fifty-six times out of ninety-one, and its median distance is 0.95 times that of the reference. The motorway goes around what the back road goes through.