Their AWS Bill Was Higher Than Their Drone Fleet
A forestry mapping company was uploading 2.7TB of drone imagery to AWS daily during fire season. The compute and egress costs hit $127K per month. Meanwhile, clients waited 2-3 days for orthomosaics that could have been done locally in 4 hours.
Evidence boundaryHistorical case-specific report. Figures describe the deployment documented at the time and are not current universal product guarantees.
Client
North American Forestry Services Company
Industry
Environmental Services
Use Case
Drone Imagery Processing & Forest Mapping
Products
Expanso
Timeline
Pilot in 3 weeks, 8 field offices in 6 weeks
ROI
$1.36M annual AWS cost reduction
/ Challenge
The Challenge
Fire season meant flying 47 drone missions per day across 8 field offices. Each mission generated 58GB of raw imagery. The workflow was simple and expensive: upload everything to S3, process in EC2, download the results. Their cellular provider loved them.
47 daily missions generating 2.7TB of raw imagery during peak season
AWS bill hit $127K per month - more than their entire drone fleet cost
Clients waited 48-72 hours for orthomosaics and change detection maps
23% of imagery failed QC after cloud processing - wasted compute
Field offices had M.2 NVMe drives sitting idle while uploading to S3
One analyst spent all day reviewing imagery quality - before processing even started
/ Solution
The Solution
We put GPU-equipped workstations at each field office running Expanso. Drones land, cards pop into the workstation, processing starts. Orthomosaics ready in 4 hours. AWS only sees the finished products - 58GB of raw imagery becomes 340MB of deliverables.
Field Office Processing
Each office has an RTX 4090 workstation running the full photogrammetry pipeline. Raw imagery goes from SD card to orthomosaic without leaving the building. Processing starts as drones are still landing.
Auto-QC Before Processing
ML model checks each image for blur, exposure, GPS accuracy, and overlap coverage before processing starts. Bad flight? Know in 8 minutes, not after a $200 cloud processing job.
Deliverables-Only Upload
Raw imagery stays on local NAS. Only finished orthomosaics, change detection maps, and flagged problem areas upload to AWS. 58GB becomes 340MB.
/ Results
The Results
Clients now get maps the same day the drone flies. The forestry company stopped apologizing for 3-day turnarounds during fire season. AWS bill dropped from $127K to $14K monthly. The one analyst doing manual QC now does actual analysis.
89%
AWS Reduction
4 hrs
Delivery Time
99.4%
Data Stays Local
3 weeks
Pilot to Value
AWS monthly bill dropped from $127K to $14K - 89% reduction
Client delivery time improved from 48-72 hours to 4 hours
23% imagery rejection rate dropped to 4% with pre-flight QC
One analyst saved from full-time QC - now does forest health analysis
8 field offices deployed in 6 weeks after 3-week pilot
Local NAS storage cost: $0.02/GB vs S3's $0.023/GB + egress
First fire season: processed 4,127 missions without cloud delays
/ What's next
Uploading imagery to process it?
If your field teams are waiting on cloud processing, we should talk. We've deployed at forestry companies, mining operations, and construction sites.
/ Typical engagement
30 days
Free assessment
90 days
Optional pilot
Measured
Baseline vs. after
