Why "Exact Location" Is the Wrong Goal

Open any safety app’s website and you will find the same promise: precise location. Pinpoint accuracy. Exact coordinates.

We have been building a responder network for months, and the more real emergencies we thought about, the clearer it became that this promise is answering the wrong question.

Here is the question it is actually answering: where is she, roughly?

Here is the question the person walking towards her has: which door?

Those are not the same problem, and no amount of GPS precision solves the second one.

What a pin does not tell you

Your phone does not know exactly where it is. It knows where it probably is, within a radius, and that radius changes constantly. Outdoors with a clear sky it might be 5 metres. On a street with tall buildings on both sides, indoors, in a basement, in the rain — it can easily be 60 or 80.

Here is the part that matters: on a map, those look identical. A pin is a pin. It does not get bigger when the phone is less sure.

So a responder looks at a confident dot on a screen and walks to it. On a dense street, 80 metres of error is four buildings and both sides of the road. They stand in the wrong place, looking at the wrong entrance, while the person who needs them is around the corner.

Nobody did anything wrong. The app just never mentioned that the dot was a guess.

So we made the pin admit it

ORBII now records how accurate each emergency location actually was, and tells the responder in words rather than making them infer it from a dot.

They see one of these:

• “Pin is good to about 12 m” — walk to it.
• “Pin could be off by 70 m — call her when close” — do not trust the dot, ring her from the corner.
• “Accuracy unknown” — we genuinely do not know, so we say that instead of pretending.

This does not make the location better. Our app was already doing the right things to get a good fix: it uses the highest-accuracy mode, waits for the reading to settle, and keeps the best one it saw. What changed is that it stopped throwing away the one number that tells a human being how much to trust the result.

The behaviour it produces is the whole point. At 12 metres a responder walks to the pin. At 70 they pick up the phone. Before this, the app could not tell those two situations apart, so neither could they.

And then the part GPS can never fix

Say the fix is perfect. Five metres. You are exactly where the dot says.

The responder still does not know whether you are on the third floor or the ground floor, whether it is the gate on the main road or the one down the side lane, whether you are inside the shop or in the stairwell beside it. A coordinate cannot express any of that, at any level of precision, because it is not a precision problem.

A photograph can.

ORBII now sends a few photos from your back camera during an active SOS, so the person coming to help can recognise the place instead of guessing at it. Three frames over the first minute. It replaces the video we used to record, which was bigger, slower to send on a weak mobile connection, and less useful to somebody glancing at a phone while walking fast.

The rules we put around it, and why

A camera that takes pictures without anyone pressing a shutter is a serious thing to put in an app. So it comes with limits, and we want to be direct about each one.

It only runs during a real emergency you raised. Never before it, never after it ends, never for a test alert.

Your screen tells you while it is happening. There is a line on the SOS screen saying photos are being sent, with a switch to stop it, right there — not buried in settings, because settings is not where you are. It is on by default, because in an emergency you are not going to go hunting for a toggle, and a responder who cannot find the door is the failure this exists to prevent.

Only a responder who already said yes can see them. This is the one we thought hardest about. When you raise an alert, it goes out to ORBII users near you — before any of them has agreed to come. Putting photographs of your surroundings in that broadcast would hand pictures of your home, the people around you, and possibly the person you are frightened of, to every stranger with the app open.

So the photos are not in it. They unlock only for someone who has accepted and is on their way, and only while the emergency is still open. When it closes, so does their access.

They are yours. You can look at them and delete any of them, whenever you like. We delete all of them after 90 days, the same as SOS audio. There is no argument for keeping a photograph of somebody’s worst night longer than that.

On screenshots, the honest version. There is no save or share button anywhere in the responder’s view, so nothing lands in their gallery. We also block Android’s screenshot button while a photo is open. What we will not tell you is that the images cannot be captured — because someone can point a second phone at a screen, and no app on earth prevents that. A promise we cannot keep is worth less than a limit we can describe accurately.

What this is not

It is not more precise. We want to be clear about that, because “precise location” is the phrase everyone reaches for and it would be easy to claim.

The location is as good as it was. What changed is that the app now tells the truth about how good, and shows the one thing that actually gets a stranger to the right door when coordinates run out.

We would rather ship a system that admits its uncertainty than one that looks confident and sends somebody to the wrong building.