Aerbits pays for itself by shortening the days a pile sits on the street, and by giving each cleanup crew back about 38 minutes a shift.
There are two clocks running on every illegal dump in Oakland, and only one of them shows up on a budget line.
The first is the one a resident feels. Every day a pile sits on the street is another day someone walks past it, another day of a hazard on the sidewalk, another day a neighborhood is told to live with it. That clock is the real product of a cleanup program, and until you fly, it is invisible. A 311 request records the day someone complained. It does not record when the pile first appeared, or how long it stayed after the report.
The second clock is crew time. Every minute a crew spends driving to a pile that isn't there, or waiting to come back with the right truck, is a minute that comes out of some other work order. A council member does not have a line item called “dumpsites detected.” They have a line item called crew hours, and one called overtime.
Most illegal-dumping arguments price the second clock and never measure the first. Both matter, and this piece measures both, starting with the one that never reaches a spreadsheet.
The first clock: how long a pile stays on the street
Start with how the count works. Piles are counted once per flight, not once per pile. A pile still on the street at the next pass gets counted again. The corridor flights landed on October 2, 6, 9, 13, 16 and 20, so a pile that survived the month is counted six times.
That is not a double count to net out. It is the count of days. Every time the drone sees the same uncleared liquid, the same construction debris, the same mattress, it is recording another day that material sat on the street. A pile seen on six passes is not six piles of garbage. It is one pile that residents walked past for three weeks.
That distinction is the whole value. A work order records a moment. The flights record duration. Read that way, these counts measure how long illegal dumping stays in place, and the real cost is not the truck that finally hauls it off. It is every day before that, when the jugs and the debris sit on a sidewalk that people use.
This is also where coverage pays for itself beyond crew minutes. The reactive system only knows about the piles someone complained about. Aerbits sees the ones nobody reported, 613 piles and 1,444 cubic yards with no request within a football field, and sees them again on every pass until they are gone. The earlier a pile is caught, the fewer of those days a neighborhood has to live with it.
The cost of an idle crew hour
Oakland's illegal dumping operation is expensive per hour and it runs nearly every day. Public Works field rates for a standard refuse truck and a two-person crew come to about $156 per hour for a single site visit, and a full eight-hour work order with the larger crew and equipment footprint can run to $5,000 or more in man-and-equipment time alone, before dumping fees at the transfer station.
For the ride-along economics, we use the corridor runs we did in October 2026 as the anchor: each CalRecycle corridor run was 32 people-hours and 16 truck-hours to pick up just over 2 tons. At field rates that is roughly $2,500 a day in man and equipment, with dumping fees on top. That is the number to keep in mind, because it is the cost of one crew-day spent loading and hauling rather than looking for garbage to load.
Run the arithmetic across the city and the scale is clear. Oakland receives roughly 29,000 illegal-dumping service requests a year (24 months of 311 records through October 2026, category ILLDUMP, 61,009 geocoded requests). At 2,500 to 3,000 operating hours a year for the dumping program, the operation is on the order of $12 million a year.
Which gives a clean rule of thumb: every 1% of crew time moved from unproductive to productive is worth about $120,000 a year. One percent doesn't sound like much until you write a $120,000 next to it.
Where the idle time actually comes from
This is where the pitch stops being arithmetic and starts being evidence. We checked 1,036 open City work orders against 1,380 accepted piles across 19 flights from September 16 to November 2, 2026, in Oakland. Each work order was resolved into one of four situations based on where the pile was relative to the reported point: nothing within 100 yards, a pile only in the 33-to-100-yard band (wrong location), a pile in place but with kinds the request missed, or confirmed as filed.
Three of those four situations are crew time that is already gone before the crew gets rolling.
Work orders with no pile within 100 yards
37% of all work orders checked. The City was sent somewhere the drone saw nothing, or the pile was gone by flight day. Some of these get caught by a supervisor on a pre-validation drive, which saves the truck roll but not the trip. The rest are a full crew dispatch to an empty lot.
Work orders pointing at the wrong spot
31% of all work orders checked. The pile is within a football field but not where the request said. The crew drives to the reported point, searches, and either hunts for the pile on foot or gets sent back. 1,106.9 square yards and 501.5 cubic yards of garbage sit within a block of where the crew was told to go.
Work orders that under-describe the pile
27% of all work orders checked. The pile is in place and the request named only some of what is there. A crew sized for bags meets drywall. 1,051.4 square yards and 486.1 cubic yards, and the request named cardboard 105 times but missed it 167 times.
Work orders confirmed as filed
4% of all work orders checked. This is the share where the report was right, the location was right, and the kinds were right. It is the only group where the crew's drive produced no surprise.
That is the shape of the problem: 96% of work orders we checked carried some kind of exception that a crew would only discover by going. And one of the most expensive exceptions is not even about location. It is about equipment.
The wrong-truck problem, in numbers
The kinds table from the same 19 flights shows how badly requests under-describe what is on the ground. Among the 1,380 accepted piles:
| Material | Piles showing it | Cu yd in those piles | Requests naming it | Requests missing it |
|---|---|---|---|---|
| Containers (may hold liquid or hazardous material) | 669 48% | 1,877.3 | 10 | 158 |
| Construction debris | 687 50% | 1,816.0 | 106 | 130 |
| Cardboard | 854 62% | 2,185.0 | 105 | 167 |
| Appliances | 215 16% | 514.3 | 32 | 41 |
| Tires | 189 14% | 660.0 | 51 | 44 |
| Mattress | 224 16% | 458.7 | 194 | 30 |
Read the container row with care. 669 piles, just under half, showed at least one container that can hold liquid or hazardous material: jugs, buckets, cans, paint containers. That is a visual tag from the imagery, not a hazmat determination, and it is the loosest label in the table. What holds up is the asymmetry around it. Ten requests named a container. 158 missed one. A crew that arrives for loose debris can end up handling material it was never told to expect, and some of it will not go in the truck it brought.
This is where the equipment math gets real. A city running a mixed fleet has pickups, mini-packers, and loader-equipped trucks. Each has a cost, an availability constraint, and a specific job. Send the wrong one and the city pays for a second trip plus the idle time of a truck that could have been somewhere else. Oakland is running loader units at roughly $7,500 a month each. A loader that is parked because the pile was already handled by a pickup is $7,500 a month of capacity that was bought and not used.
Scattered routes
Reactive 311 dispatch scatters trucks across the city. Each stop is a round trip from the yard, and the yard-to-transfer-station leg alone is 3.5 miles one way. Pile sizes measured from the air let the city batch by area and by equipment, so a loader route is a loader route instead of one truck's second trip of the day.
Dumping growth
A pile that sits attracts more dumping. That is not a soft benefit, it is a compounding cost, and it is why finding piles sooner means smaller loads and fewer trips. The same 19 flights measured something else worth naming: 613 accepted piles, 44% of everything we found, had no 311 request within 100 yards, covering 3,509 square yards and 1,444.4 cubic yards. Some of that is the 40% unreported share the city's own field observation suggests. None of it is visible to a reactive system.
Of those unreported piles, 45 (501.3 square yards) sit within 27 yards of a railway, on railroad or Caltrans land the City does not clean. That is a separate problem, and one worth its own line item, but it is small next to the other 568 piles that are simply never reported.
Turning recovered hours into budget savings
Here is the part that decides whether a contract gets funded, and it is where most proposals fail. Recovered crew hours only save cash if the city actually cuts spending somewhere. A crew that finishes early and goes home early is a labor-savings number that may never reach the budget.
Oakland is currently funding additional crews working overtime on Tuesdays, Wednesdays, Thursdays, and Saturdays. That is the line to pitch against. Every recovered hour is an overtime hour the city no longer needs, and an overtime hour is a line a council member can watch shrink on a budget document. That is a different kind of promise than “more piles found.”
The break-even, stated plainly
At a proposed contract of roughly $960,000 a year, and using $120,000 of value per 1% of crew time recovered, the contract pays for itself when roughly 8% of crew time moves from unproductive to productive.
On an eight-hour shift, 8% is about 38 minutes per crew per day. That is the whole pitch. Not “we find a lot of dumpsites.” Not “we improve detection.” 38 minutes. That is a concrete, checkable, small-sounding number that a public works director can accept or reject on the merits, and a budget office can accept or reject on the dollars.
What we report from the pilot
If the pitch is in crew minutes, the pilot has to be measured in crew minutes. These are the five numbers we report, and they are all measured before and after:
- Dry-run rate. The share of work orders dispatched to a location where no pile exists, measured before Aerbits and after. The baseline from 19 flights is 37% of work orders having no accepted pile within 100 yards. This is the strongest number in the set, because it is the one the City can check against its own dispatch records.
- Work orders closed per truck-shift. Output per shift, which is the metric that captures routing efficiency and equipment matching together.
- Tons hauled per crew-hour. The productivity number. If recovered time is real, this goes up. If it does not, the time was not recovered, it was just moved around.
- Repeat trips caused by wrong equipment. The count of jobs where the first truck could not finish and a second trip was needed. This is what justifies loader utilization claims.
- Overtime hours. The number that closes the budget argument. If overtime does not come down, the savings are theoretical.
How the pricing should be structured
The honest structure is a base fee plus a share of verified crew-hours recovered. The base fee pays for coverage. The variable piece pays for the part where we can prove we made the operation more productive.
That matters because it aligns the two things that usually fight in a contract like this: the vendor wants to get paid for flying, and the city wants to get paid for results. A base fee plus verified crew-time share means the vendor only earns the variable portion by making crews measurably faster. It also means the verification has to be the City's, not ours, which is the right way round for something that touches a public works budget line.
For context on the unit of account, Oakland's 311 records show why wrong-address work is so expensive: roughly 42% of all illegal-dumping requests in a 24-month window were a repeat report of the same pile within seven days, and about 62% within 21 days. The City's own disposition vocabulary puts the explicit duplicate bucket at 18% of requests, or about 5,500 a year. On the crew side, a bad work order costs roughly $185 end to end; a ghost pile that is caught before dispatch costs about $40 in a drive-by; a full return trip for unprepared equipment runs closer to $185. Those are the numbers a per-outcome pricing ladder is built from, and they are covered in more detail in We Price for Outcomes, Not Effort.
The point
Detection is the mechanism. Crew time is the product.
A city does not budget for piles. It budgets for crews, trucks, and the overtime line. So the contract should be written against the thing that actually shows up on the invoice, and the program should be measured on the thing the director actually feels at the end of a shift: did the crew spend its eight hours loading garbage, or hunting for it?
Aerbits can give each crew back about 38 minutes a shift. We have the data to measure it, and the City should have the data to verify it.
quth-gb8e, category ILLDUMP, 61,009 geocoded requests over 24 months. The container count in the kinds table is a visual tag, not a laboratory or hazmat determination, and should be read as "a container is visible," not "a hazardous material is present." Recovery percentages, the $960K contract value, and the annual program cost are planning estimates, not measured outcomes, and should be treated as such until the pilot reports the five metrics above.