Skip to content

Under the hood

There's no route to build

Snowmass replaces hand-built routes with live dispatch. Here's what actually happens during a storm — and what's running underneath.

During the storm

  1. 1

    Drop your properties once

    Place your properties on a map — or import a whole book of business from a CSV. That's the setup. There are no routes to draw, no per-operator lists to click together, and nothing to rebuild when the season changes.

  2. 2

    Every property is scored live

    As tractors move, Snowmass continuously scores each open property for the whole fleet — weighing each machine's position, the real road travel time to the property, and the direction you've chosen to sweep. Each property is scored the moment a machine needs its next stop — driven by live GPS and completed stops — so an assignment always reflects where the fleet is right now.

  3. 3

    Each tractor claims its best next stop

    The best-positioned machine atomically claims a property — settled in the database in a single step so two tractors can never double-book a property. The fleet fans out across the perimeter on the opening move, then sweeps dense blocks before moving on, without a dispatcher grouping stops by hand.

  4. 4

    When a plan breaks, the swarm heals

    A machine goes silent, stalls, or gets recalled for a breakdown — and its properties are released back to the swarm automatically within minutes, picked up by whoever's now best placed. The moments a static route falls apart are the moments Snowmass absorbs on its own.

  5. 5

    Plow-ridge recheck closes the loop

    When a crew clears a property but the road plow re-buries the entrance, the operator flags it — and Snowmass holds the property through a cooldown, then re-queues it so the swarm sends someone back. The re-accumulation a static list can't even express is handled for you.

Under the hood

What makes the swarm feel intelligent

Snowmass dashboard showing active passes and live run counts

The scoring — position, road time, and direction

Every few seconds, each open property gets a live score for every available tractor. Three things drive it, and none of them is a route you drew in advance.

  • Position: how well-placed a machine is relative to the property and the block it sits in — so tight clusters get swept together.
  • Real road travel time: actual drive time over the road network from a live routing engine, never a straight-line guess that ignores a river or a highway.
  • Direction: your chosen sweep — outside-in, inside-out, or nearest-first — so the whole storm flows the way you want it to.
Pass detail with deployments, service visits, and crew confirmation

The atomic claim — why two machines never double-book

Scoring decides who should take a property; the claim makes it real without a race. When a tractor takes a property, that claim is settled in the database in a single atomic operation.

  • Two machines reaching for the same property in the same instant is resolved cleanly — exactly one wins, the other simply scores its next-best stop.
  • No dispatcher arbitration, no double dispatch, no two crews arriving at one property to find each other already there.
  • Because claiming is race-free at the database, the fleet can re-optimize continuously without ever stepping on itself.
Operations command dashboard with per-machine liveness

The behaviors that keep coverage smart

A raw "nearest available" rule would scatter a fleet and waste minutes between stops. A handful of production behaviors shape the swarm into something that covers ground the way an experienced dispatcher would.

  • Start-spread fan-out sends the opening moves wide across the perimeter instead of piling onto one block.
  • Density cohesion keeps a machine sweeping a dense block to completion before it leaves — no ping-ponging across town.
  • Anti-steal keeps a machine from grabbing a property right next to a peer already working that block, so crews don't overlap.
Operator messaging with live per-operator conversations

How it heals when reality diverges

The whole point of a swarm is what happens when the plan breaks. Snowmass treats failure as normal and routes around it without a human re-planning the storm.

  • A machine that goes silent past its window has its claimed properties released back to the swarm automatically — a breakdown or dead battery doesn't strand a route.
  • A dispatcher can recall a broken-down machine on purpose, and its work flows straight back to the fleet.
  • Wedged claims — where one deployment ends up holding a property it can't finish — are detected and healed so work never quietly stalls.

Only possible because there's a real engine underneath

Live ETAs, what-if simulation, self-healing, plow-ridge recheck — none of these are features you could bolt onto a static route. They exist because dispatch is a living system, re-deciding the storm every few seconds. That engine is the whole product.

Ready before the next storm

Run a simulation scaled to your fleet, or talk to the team about the season.