01 · The last 230 feet
Pins point at roofs. Parcels go to doors.
Every map your fleet already uses was built to find addresses, and an address is a point in the middle of a parcel. For a house, close enough. For everything with a loading bay, a side gate, a leasing office that signs for things, or a door that is legally on a different street, close enough costs minutes per stop and gets paid for in missed windows.
The average lie on Maple Court. On this sample building the address pin sits 230 feet from the door deliveries actually use, around the back, past a gate the pin has never heard of. Your buildings have their own versions; the distance is different, the shape of the problem is not.
What a new driver does with a roof pin. Circle once following the arrow, once looking for numbers, once on foot. The van idles somewhere it should not be, and the parcel room was behind them the whole time.
What has to be learned, per building. Where to stop the van so it is legal and close, which entrance takes parcels, and the walking line between them. Three facts. A building only has to teach them to one driver, once, if something is paying attention.
02 · How the door gets found
Your drivers already survey every door.
Rightdoor's SDK rides in the courier app you already ship. It watches the shape of a stop, not the driver: the van stops, the phone walks, the phone pauses at a door while a handover happens, the phone walks back. A few stops agree, and the building has a learned door with a parking bay attached.
Park. Where vans actually stop for this building, clustered over real stops, weighted toward stops that ended in a fast handover.
Walk. The on-foot line from bay to door, gates and cut-throughs included, because couriers found them before any map did.
Door. The pause point that keeps agreeing with successful deliveries becomes the pin. When a building moves its parcel room, the pauses move, and so does the pin.
Three sample buildings, three different ways of being wrong. Switch the sign:
03 · What the driver sees
Signage, not suggestions.
Next stop · Maple Court · 3 parcels
Park Alder St bay → Gate B → Door 4
Code on file, shown only on approach. Parcel room inside, left of the mailboxes. If the bay is full, the second legal stop is marked two doors down. Everything readable at arm's length, in gloves, in rain, because that is the job.
No new workflow. The stop simply navigates to the learned door in the app your drivers already use. There is no Rightdoor screen to learn, which is deliberate.
Offline like a tunnel. Learned doors for the day's route ride on the phone. Basements, garages and dead zones get their directions before they eat the signal.
04 · What operations gets
Seconds, added up honestly.
Drag your own day through the arithmetic. The seconds-saved figure is the one argument every vendor inflates, so here it is as a slider you control, with our price held against it at the same size.
146 driver-hours a week handed backAt 140 stops and 30 saved seconds, across 25 vans, five days. Time a route can spend delivering instead of searching.
$700 a week to RightdoorFour cents per delivered stop, same arithmetic. If your seconds estimate halves, both lines halve together, which is the honest way to buy this.
Stop timing without surveillance theater. Park-to-door durations per building show which stops eat routes, so planners fix buildings and time windows instead of blaming whoever drove them.
Exceptions with a location attached. A failed delivery records where the driver actually stood. "Customer not home" at the wrong door stops being indistinguishable from the truth.
05 · Limits, in large type
What a pin cannot fix.
New buildings start unknown.The first drivers teach the system, and early stops there look like today's stops. Density is the product; a two-van rural route learns slowly and gains least, and we will tell you that before you pay.
Gates change. Codes rotate.Learned doors decay on purpose. A door whose pauses scatter or fail gets demoted fast, and the next driver gets the honest "entrance unconfirmed" instead of a confident old answer.
Doors, not people.The SDK learns stop shapes on the device: park, walk, pause, return. It does not need a driver's identity, break habits or life story, and fleets can hold the raw traces on their own side of the API.
"Our new starters used to bleed twenty minutes a route on three hospital stops. The doors were never where the pins were. Now the van stops once, in the right lay-by, and the walk is a straight line a child could give directions for."Invented for this fictional site: Renata K., depot lead at a 40-van parcel fleet in Ohio
06 · Price and starting
4¢ a stop. Live by Friday.
Per delivered stop, counted by your own scan events, with a monthly cap per vehicle so a heroic December cannot produce a heroic invoice. No platform fee, no per-seat anything. Failed stops are not billed; a pin that did not get anyone to a door earned nothing.
An afternoon of SDK. A few lines in the courier app, a test route around your own depot, and the first learned doors appear on stops your drivers were making anyway.
One depot pilots. Two weeks, your worst buildings first. We report seconds saved with the method shown, including the stops where we saved nothing.
Roll out by route density. Buildings your fleet visits daily learn in days. The long tail keeps improving quietly behind them.