Spatial analysis

Know where to build next.

AI agents score every block on your own data and public data, from the census to the FCC broadband map, and show the method. A person decides where you build.

The questions it answers

Six questions planners ask before they build.

Coverage planning, cost, demand, BEAD, routes and poles. Each answer below is a real result from one demo project, with what it read.

  • 01Coverage gaps and unserved locations

    Which locations does no working network reach?

    874 of 874locations with no in-service network

    Every network feature with a lifecycle stage, 1,954 of them, is planned, so none counts as coverage today. Run it as if the plan completes, and a drop reaches every location.

    Read: the lifecycle stage of every network layer, layers/addresses.geojson

  • 02Serviceability and cost per passing

    Can we reach every home, and what does each one cost?

    $772per home passed

    All 874 locations sit inside the 60 m drop limit; the longest drop is 59.9 m. The draft design builds for $703,863 across 912 homes.

    Read: the draft design's drops and its build plan, materials, labor and 10% contingency

  • 03Demand and take-rate scoring

    Where will enough people sign up?

    81of 100 for 48th Avenue & 212th Street, the top block

    9 accounts already wait on its 49 units. At the demo's 30% take-rate policy, a campaign there opens at 15 sign-ups. 132 waiting accounts, scored across 45 blocks.

    Read: layers/customers.geojson, layers/addresses.geojson, on an H3 grid at resolution 10

    • Unservedunder 25/3
    • Underservedunder 100/20
    • Served100/20 or better

    04BEAD eligibility tiers

    Which locations count as unserved or underserved?

    874locations to tier against 25/3 and 100/20

    Lux reads the FCC broadband map cell by cell: serviceable locations, each one's best advertised speed, wired service apart from satellite. The demo project holds no FCC data, so it shows no tier counts.

    Read: FCC Broadband Data Collection, per H3 cell of about 0.74 km²

    • On poles · 804 m
    • Sidewalk · 694 m
    • Road crossing · 8 m

    05Route options

    Which way should the fiber run?

    1,506 mfeeder route, picked from 2,666 street segments

    Every segment is an option: aerial on the existing poles, a sidewalk trench or a road crossing. Each is weighed by length and kind: aerial at half the cost of a sidewalk trench, a road crossing at three times.

    Read: street-graph (OpenStreetMap streets with your poles and spans), compute/routing-feeder.json

    • Make-ready · 12
    • For a person · 2
    • Clear · 272

    06Poles and make-ready exposure

    Which poles hold up the build?

    12 of 286poles need make-ready

    272 clear. 2 are set aside for a person: a steel pole the method does not cover, and an attachment height a model placed and no person has confirmed. The 2 priced so far need a down guy and anchor, $1,850 each.

    Read: layers/poles.geojson, poles/*/loading.json, _meta/pricing.json

Ask the map

Ask in plain words. See what it read.

Pick a question

YouWhere should we build first?

  1. Readlayers/customers.geojson524 accounts
  2. Readlayers/addresses.geojson874 addresses
  3. Readpoles/*/loading.json286 verdicts
  4. Readlayers/central_offices.geojson1 office
  5. ScoreFive weighted criteria30 · 20 · 15 · 5 · 5
  6. TestTop weight moved 10% either wayleader holds

Answer

48th Avenue & 212th Street · 81 of 100Waiting demand, units, distance to the office, poles needing make-ready and past churn, each scaled 0 to 1 across the blocks.

Each block is one H3 cell at resolution 10, about 1.5 hectares. The counts come from the project's own files, on the same grid the analysis agent scores.

Your data and public data

Your files and the public record, on one map.

Every answer names the sources it read: the public record an ISP or broadband operator plans against, and your own GIS.

Public dataIn every project

  • US Census

    Population, households, housing units, median household income, renter and owner share, from the American Community Survey.

    Block group · tract · county
  • FCC broadband map

    Serviceable locations, which provider offers what technology at what speed, and how many reach 100/20 over wired service.

    H3 cell · about 0.74 km²
  • OpenStreetMap

    Buildings and streets for the area you select, and the street network a route runs on.

    Building · street
  • US parcels

    Parcel lines from Regrid, with the owner, land use and zoning of a parcel on a click.

    Parcel
  • Census boundaries

    Counties, tracts, block groups, places and ZIP codes from the TIGER files.

    Boundary

Your dataStays in your workspace

  • GIS files

    Shapefile, File Geodatabase, GeoPackage, GeoJSON, GeoTIFF, Parquet, CSV and Excel, up to 5 GB a file.

    Upload
  • Live services

    Esri Feature Services, WMS, WMTS and WFS, STAC catalogs and Google Sheets.

    Connect
  • Databases

    Tables from PostgreSQL, BigQuery and Snowflake.

    Connect
  • Poles

    Your pole records, or the poles your crews capture on the phone. Each one gets a loading verdict.

    Per pole
  • Addresses and accounts

    Address lists placed on the map, and customer accounts with their status.

    Per location
  • Your network

    Existing and planned plant, kept apart by lifecycle stage, so planned work never counts as coverage.

    Per feature

Your data is encrypted in transit and at rest, walled off per organization, and never used to train models.

From analysis to design

The area you accept goes straight to design.

The analysis goes with it. The design agent drafts the network on its own branch, and every step after it follows from that design. Nothing changes until your team adopts it.

  1. 01

    The area you accept

    Bayside, Queens: 874 buildings and 912 homes, with the ranking and its method attached.

    A person accepts

  2. 02

    Network design

    3 cabinets, 193 closures and 874 drops. The worst path loses 22.26 dB of a 28 dB budget; 0 rule violations.

    The design agent drafts

  3. 03

    Splice plans

    Fiber by fiber through the office, 3 cabinets and 193 closures, with 162 splitters placed.

    Drawn from the design

  4. 04

    Make-ready

    286 poles judged: 272 clear, 12 need make-ready, 2 set aside for a person to decide.

    Checked on every pole

  5. 05

    Build plan

    $703,863 to build: 679 work orders and 5,237 crew hours, priced line by line.

    Your engineer approves

The numbers are the app's own outputs for this design.

See network design

Questions

What planners ask.

Do we have to replace our GIS?

No. Keep it and connect it. Lux reads your GIS files and your Esri and OGC map services, and puts them on one map next to census, FCC broadband and OpenStreetMap data.

Which public data does Lux read?

The Census American Community Survey by block group, tract or county; the FCC broadband map, per H3 cell; OpenStreetMap buildings and streets; Census TIGER boundaries; and US parcels from Regrid. Every answer names the sources it read.

Can it help us plan for BEAD?

Yes, on the planning side. Lux counts locations against the 25/3 and 100/20 lines from the FCC broadband map, cell by cell, and the draft design prices each area per home passed. If you hold a Location Fabric license, your premises can be matched to Fabric location IDs, and an unclear match waits for a person.

How do we know a ranking is right?

Every answer shows its method: the sources, the criteria, the weights and the grid. The same data and the same weights give the same map. The biggest weight is moved 10% either way, and a ranking that flips is called a close call.

Who decides where we build?

Your team. The agent ranks areas and proposes. Promoting an area to the expansion plan waits for a person, and nothing is built until a campaign and an approved design say so.

Bring one area and one question.

Send us your serving area and the question your team asks every week. We will answer it on the map in the demo, with the method shown.