How heads-up alerts work
When you are out on a bike or on foot, a notification you have to stop and read is no use to you. So RideByWeather says it instead — a short sentence through the earbuds you are already wearing, about the stretch of road in front of you. This page is the whole of it: what we say, where every fact comes from, and where the data stops being trustworthy.
What it tells you
Five things, each of them a fact about a place rather than an opinion about it. Road conditions and reported incidents are two separate switches, and both start switched off.
Fast road, no bike lane
CyclingA road tagged 45 mph or faster with no bike lane and no shoulder mapped on it.
Source: OpenStreetMap
No sidewalk
Running & walkingA road mapped as having no sidewalk, where traffic is tagged 35 mph or faster. Only when the map explicitly says there is none — see the limits below.
Source: OpenStreetMap
No street lighting
EveryoneA road mapped as unlit, and only once the sun is more than 6° below the horizon. It is not news at noon.
Source: OpenStreetMap
Few turn-offs
EveryoneFewer than four mapped roads within half a mile — a long way to the nearest turn-off if something goes wrong.
Source: OpenStreetMap
Reported incidents
Opt-in, off by defaultA robbery, assault, sex offence or bike theft reported to police near this spot, in the last few months, at a similar time of day to when you are out.
Source: City police open data
The rules that keep it bearable
An alert layer that talks too much gets muted on the first ride and never turned back on. These limits are not settings — they are how it is built.
Once each, and never again
An alert is announced a single time per activity. No repeats, no nagging.
One every 45 seconds, at most
If several things stack up, the most serious one is spoken and the rest stay quiet.
Only what is ahead of you
When we know which way you are travelling, anything behind you is dropped. Being told about a corner you already passed is noise.
Only within 300 metres
About forty-five seconds of riding — close enough to act on, far enough to react.
Always on screen too
Everything spoken also appears as a card with its source and date, so it lands with your sound off.
About the reported-incident alerts
This is the part that deserves the most explaining, so here it is plainly. Turning it on adds incidents that were reported to the police and published by the city as open data. Today that means New York City, Chicago and San Francisco — the cities that publish incident-level data with coordinates, promptly enough to be worth reading.
We report incidents. We do not rate neighbourhoods.
Every alert is a single reported incident, at a location, on a date — and it always tells you that date. What you will never get from us is a score for an area, a “safe/unsafe” verdict, or a colour-coded map of where to avoid. Reported-crime data records where crime is reported as much as where it happens: it moves with how heavily an area is patrolled and how much people there trust the police. Turning numbers like that into a judgment about a neighbourhood and speaking it into somebody's ear is not something we are willing to stand behind. A dated incident at a corner, we will.
What gets filtered out
- Most crime types. Only robbery, assault, sex offences and bike theft come through — the ones that change what somebody alone on foot or on a bike might do. Burglary, fraud, drug and property offences are dropped. They tell a runner nothing, and including everything is how one of these maps quietly turns into a map of who lives where.
- The wrong time of day. A 2 a.m. incident is not information for a 7 a.m. run. Incidents only surface if they were reported within about three hours of the time you are actually out.
- Anything about the people involved. These datasets carry suspect and victim age, sex and race columns. We never read them, never store them, and never show them.
- Stale sources. Los Angeles' dataset has not been updated since December 2024 and Austin's carries no coordinates, so neither is wired up. Silence beats a confident answer from dead data.
Where this data stops
Worth knowing before you lean on it:
- The map is incomplete, and we treat it that way. Road detail comes from OpenStreetMap, which is volunteer-mapped and patchy — in central San Francisco only about a third of streets record whether they have a sidewalk at all. An unmapped sidewalk is not the same as a missing one, so we only tell you a road has no sidewalk when the map explicitly says so. That means real hazards will sometimes go unmentioned.
- Silence is not an all-clear. No alert means we found nothing recorded, not that there is nothing there.
- Police data lags. New York publishes roughly a quarter behind; Chicago and San Francisco are usually within days. Recent events may not appear at all.
- Outside the listed cities there are no incident alerts. Road conditions work anywhere OpenStreetMap has coverage, which is most of the world.
- It is not a substitute for your own judgment. You can see the road; we can only see what has been written down about it.
Turning it on
Open a ride, run or walk, go to the settings panel, and switch on Warn me about what's ahead. Road conditions and reported incidents have their own switches, and you can keep the alerts on screen only by turning off Say it out loud. Nothing is announced until you actually start recording.
Your location is used to look up what is near you and is not stored for this feature. See the privacy policy for how location is handled across the app.
Ride with it on
It costs nothing and it stays quiet until there is something worth saying.