fieldwork
The Route Density Math of Working a Storm
Why the shape of a hail swath, not the size of the market, decides what a canvassing day is worth, and why county lines are the wrong boundary
Two numbers nobody writes down
A canvassing day contains exactly two activities. Your people are either in front of a homeowner or moving between homeowners. Everything else about the day acts on those two buckets.
The ratio between them is not a property of the market. It is a property of the shape of the storm. A swath that runs for miles as a narrow ribbon across thinly settled country produces a day that is mostly driving. A swath of the same area sitting over a platted subdivision produces a day that is mostly conversations. Same acreage, same storm, and a completely different economic day.
Almost nobody logs those two buckets separately, which is why a bad week usually gets explained by the script or by the market. Often the real explanation is the geometry the crew was handed that morning.
The boundary you canvass to is probably wrong
Here is where a day most often gets lost before it starts: the crew canvasses a boundary that was never a storm boundary.
The National Weather Service upgraded warning capabilities on October 1, 2007. Before that, warnings were county-based, which meant a warning "encompassed the entire county regardless of what portion of the county the storm or threat was located." Under storm-based warnings, the polygons "only cover the portion of the county actually threatened by the storm." The same handout notes that NOAA Weather Radio All Hazards continues to alert the entire county, while several vendors alert only if you are inside the storm-based polygon.
So two different mental models of the same event have been circulating for a long time. One of them is a county. The other is a polygon. People who grew up on the county model canvass counties, partly because the radio still tells them a county.
The handout includes a worked example that shows how far apart the two models can sit. Tornado warnings were in effect for Hale, Bibb, Perry, Tuscaloosa, Jefferson and Shelby counties. The cities of Tuscaloosa, Birmingham and Calera were not in the polygons, and therefore were not under a tornado warning. Three named cities, inside warned counties, outside the warned area.
That example is a tornado warning and not a hail swath, and the hazards behave differently. The geometry lesson is what transfers: the warned area and the county are different shapes, and the place names everyone recognizes sit wherever they happen to sit.
Failure mode one: canvassing a political boundary
Two failure modes follow from the mismatch, and they cost in opposite directions.
The first is canvassing a political boundary: a county, a city limit, a ZIP code. The addresses are easy to obtain and the list comes out long, and a meaningful share of it was never under the storm. Reps spend the day driving to doors with nothing to look at, hear nothing worth hearing, and come back with a read on the market that is really a read on the list. The damage doubles, because you lose the day and then you draw a conclusion from it.
A ZIP code deserves a specific warning. It is a postal routing boundary, and nothing about how it was drawn relates to where a storm went. A swath can clip the corner of a large ZIP and miss everything else in it.
Failure mode two: canvassing only the middle
The second failure mode is quieter and more expensive. A crew finds the obvious center of the swath, the street where every roof is visibly hit, and works it until it is exhausted. That is the right place to start and the wrong place to stop. The center is where the damage is loudest, so it is also where competitors are thickest and where the homeowner has already heard the pitch twice. Meanwhile the edges of the same swath go unworked, and whoever is reading the geometry rather than the street picks them up next week.
The center sells itself. The edge is where a route is actually won or lost.
Why the edge is worth the most and is known the least
The edge is also the part of the map you can trust the least, and it is worth knowing why.
Most of the footprint you work from is radar-derived. NCEI's Severe Weather Data Inventory, which carries NEXRAD Level-III hail signatures among other products, states plainly that it adds no quality control beyond archival processing, that missing data does not mean no severe weather occurred, and that much of the automatically derived data is radar-based and represents probable rather than confirmed conditions. That caveat applies to the whole product, and it bites hardest at the margins, where an estimate is deciding between some hail and none.
The research on radar hail estimates points the same way. MESH, the maximum estimated size of hail, estimates hail size from the reflectivity properties of a storm above the environmental 0 degree C level, and it is used across the NWS to diagnose expected hail size. Over a study period of 2012 to 2019 across the contiguous United States, severe hail hours estimated from MESH ran 2 to 4 times greater than those estimated from Storm Data, even in plains areas where population density is relatively high, and the differences were larger where population density is low. The same paper notes that lack of radar coverage, from beam blockage or widely spaced radars, explains part of why that gap narrows in the Intermountain West, the Appalachians and the Northeast.
Two readings of that gap are both partly true, and the authors say so: severe hail is plausibly going underreported where population density is low, and MESH likely overestimates severe hail to some degree. For a canvasser, either reading leads to the same handling of the edge. It is the highest-value part of the route because nobody has worked it. It is the lowest-confidence part of the route because the estimate there is least certain. So you go, you expect a higher share of doors with nothing on the roof, and you do not let that share rewrite your read on the center.
Density is an ordering problem, not a radius
The habit worth replacing is the circle on the map. A radius around a report, or around a signed job, is a shape chosen because it is easy to draw. A storm footprint is not a circle, and it is rarely even convex.
The alternative is to hold the footprint as an actual geometry and treat the route as an ordering of addresses inside it. The format for this is not exotic. GeoJSON defines seven geometry types, including Polygon and MultiPolygon, and in a polygon the first ring is the exterior ring while any additional rings are interior rings, meaning holes (RFC 7946). A linear ring is a closed LineString of four or more positions whose first and last positions are identical. Coordinates are WGS 84 decimal degrees, and a position lists longitude first and then latitude, which is the single most common way a working swath file gets silently mirrored to the wrong side of the country.
MultiPolygon and interior rings are the two features that matter most for a storm. A real swath often arrives as several disconnected pieces, because the storm pulsed as it moved. And holes are real too: a band inside the outline where the estimate fell under the size you care about. A circle cannot express either one, so a circle forces you to choose between including addresses that were never hit and excluding addresses that were.
Once the footprint is a geometry, route density stops being a feeling. You ask which of your addresses fall inside it, and then you order those addresses for travel. The first question is a containment test. The second is a routing problem. Neither one is answered by a radius.
What to do with this
Stop canvassing the boundary in the headline. The county, the city and the ZIP were drawn for other purposes, and the Birmingham handout shows three named cities sitting outside polygons that covered parts of their own counties.
Work the loud center first, then plan the edge on purpose, with the expectation that it produces a worse hit rate per door and far less competition at the door.
Hold the footprint as a geometry rather than as a list of place names, so the question "was this address inside it" has an answer that does not depend on who typed the list. Storm Stream is an API over the same public feeds: it builds swath geometry and answers which of a list of addresses fell inside it.