You come off the motorway at two in the morning running on fumes, follow the pin to a forecourt that has shown up on every map you own for years, and find a boarded-up building with the pumps torn out. The map was not lying. It was describing a place that used to sell fuel, which is not the same claim as a place that sells fuel now.
The base map records what exists, not what is open today
Almost every fuel and EV map on your phone, including this one, ultimately traces back to a small number of geographic databases. OpenStreetMap is the biggest open one, and its schema for a filling station is deliberately static. A station gets amenity=fuel, a name, a brand, maybe fuel:diesel=yes or fuel:octane_95=yes to say which grades it sells, and an opening_hours string. Read the tagging page yourself and you will not find a field for current stock, a field for whether the pumps are switched on, or a field for the price on the sign. Those tags describe capability, not condition. They answer "can this place sell diesel" and leave "does it have diesel right now" to somebody else, because nobody has ever agreed on how a volunteer-edited map would keep that up to date. A map built on this data can tell you a station exists and what it is supposed to sell. It cannot, from the base data alone, tell you anything about today.
A closed station does not delete itself
OpenStreetMap and every commercial map built on similar crowdsourcing have no expiry date on a point of interest. A record persists until a person visits, notices the change, and edits it, or until an automated conflation process cross-checks it against something more current and flags the mismatch. Neither happens quickly for a fuel station on a road most people drive once. Local editors tend to fix what they see near home; a rural forecourt an hour from the nearest active mapper can sit demolished for a long time before anyone bothers, because updating a map is unpaid, invisible work that benefits strangers. The convention some mappers use, prefixing a tag with disused: to mark something closed but not yet deleted, only works if a human applies it. Nothing in the data model forces the correction. So the pin you are navigating to might reflect a satellite pass, a survey, or an app update from three years ago, and there is no flag on it that says so.
"Available" from a charging network means the box reported in, not that it works
Public chargers report their own state over the Open Charge Point Protocol, and the status string "Available" has a narrow, specific meaning: no vehicle is currently plugged into that connector and the charge point has told its network it is ready to accept one. It says nothing about whether the screen has crashed, the card reader will take a payment, or the cable is intact. The gap between that self-report and reality is now measured, not guessed at. Field inspections by ChargerHelp across 4,800 charge points found that more than one in ten showed as online in the software while actually failing a real test charge. Aggregated uptime data from Paren put true station uptime at 84 percent against a 92 percent figure the same stations reported about themselves, an eight-point gap between what the network says and what a car can actually do there. J.D. Power's regular survey work puts the failure rate for a public charging attempt at around one in five, and breaks the reasons down: the charger was simply out of service 62 percent of the time, no charger was available or the queue was too long 18 percent of the time, payment was refused 11 percent of the time, and a damaged cable or connector accounted for 9 percent. Harvard Business School researcher Omar Asensio, after mining over a million charging reviews, put the average reliability score at 78 percent and made the comparison that sticks: at a normal fuel station, if two pumps out of ten were broken, drivers would not tolerate it for a week.
A card transaction only proves somebody paid for something
Some apps infer that a station has fuel because a card was authorised there recently. That is weaker evidence than it sounds. A pump authorisation can happen and then fail partway through a fill if the tank runs dry mid-transaction. A transaction can be for the shop, the car wash or a bag of ice, not for fuel at all, and the point-of-sale system that generated it usually does not distinguish. A driver can also pay and get nothing, then walk inside for a refund, and the original charge still shows up in whatever feed an app is buying. None of this means transaction data is worthless. It means a payment event is evidence that a card machine worked, not evidence that a specific grade of fuel came out of a specific pump ten minutes ago.
Opening hours are a small programming language, and most parsers get it wrong
The opening_hours tag looks like plain text and is actually a formal grammar with its own precedence rules, exceptions and edge cases for things like public holidays and hours that cross midnight. The OpenStreetMap wiki says plainly that "the syntax is quite complex, but it allows even some highly complicated rules," and then documents a long table of common mistakes mappers make when writing it, from using 0600-1800 instead of the required 06:00-18:00 to mixing up the comma and semicolon that separate rules with different meanings. The specification itself notes that many applications implement only a limited subset and simply fail, silently or otherwise, on anything more elaborate than a single daily range. So even where the underlying hours were entered correctly, the app rendering them to you may not understand the string it is holding, and will either show nothing, show it wrong, or quietly drop the exception that mattered, such as the one that says closed on a specific public holiday.
The price on the sign and the price in the feed are rarely the same number
Most published fuel prices are averages, not readings off a specific pump. The US Energy Information Administration samples around 1,000 stations out of roughly 130,000 nationwide every Monday morning and publishes a volume-weighted national and regional average on Tuesday. That is a genuinely useful economic indicator and a poor way to know what a specific forecourt is charging this afternoon. A smaller number of countries do better by law. Germany's Markttransparenzstelle für Kraftstoffe requires every public station operator to report each price change to the regulator in real time, and that feed is what powers German price-comparison apps down to the individual station. Most of the world has nothing like it, which means a price shown on a global map is often a regional estimate wearing the appearance of a specific quote.
What this site does, and does not, get right about all of it
Refuelia is not exempt from any mechanism described above. It is built on the same base map data with the same absent stock field, it shows chargers whose network told it "Available" with the same narrow meaning, and it cannot inspect a cable or a card reader from a server. Where it tries to do something different is in what it shows you about its own uncertainty rather than pretending the uncertainty is not there. Every status on the site carries the minute it was recorded, not just a colour. A report from nine days ago is labelled as nine days old, not repainted green to look current. Confidence in a report decays with its age on a curve, so a fuel report from an hour ago and one from a week ago do not look the same even if both said "available" at the time. Where nobody has filed a recent report for a station, the map says exactly that, "no recent report," rather than inferring a hopeful status from stale base data because a green pin is more satisfying to look at. This does not fix the underlying problem that closed stations linger in map data or that charger backends over-report their own health. It means the site tries not to add a false layer of confidence on top of data that was already uncertain before it arrived. Treat any fuel or EV map, this one included, as a starting point for a decision you still have to make with your own eyes at the forecourt, not as a promise about what you will find there.
Sources
- OpenStreetMap Wiki: the amenity=fuel tagging schema has fields for brand, opening hours and which fuel grades a station sells, with no field for current stock, availability or price
- OpenStreetMap Wiki: the opening_hours syntax is described as quite complex, with a documented table of common mapper mistakes and a note that many applications only implement a limited subset of the grammar
- Harvard Business School Institute for Business in Global Society: analysis of over one million charging reviews found an average US charger reliability score of 78 percent, roughly one in five chargers not working
- Kelley Blue Book, citing J.D. Power's E-Vision Intelligence Report: about 20 percent of public EV charging attempts fail, 62 percent of failures because the charger was out of service, 18 percent no charger available, 11 percent payment refused, 9 percent damaged cable or connector
- Canary Media: ChargerHelp field inspections of 4,800 charge points found more than 10 percent reported online in software while failing an actual test charge, and Paren data put true uptime at 84 percent against 92 percent self-reported by stations
- US Energy Information Administration: weekly retail gasoline prices are sampled from about 1,000 of roughly 130,000 US stations each Monday and published as a volume-weighted national and regional average the next day
- Bundeskartellamt: German law requires operators of the country's roughly 15,000 public fuel stations to report every price change to the Markttransparenzstelle für Kraftstoffe in real time