Case Study · Aerial imaging
DB
Scaling aerial survey processing
Parallel tiling and spot compute cut processing time by 64% and spend by half.
Tools
- Figma
- React
- Python
- AWS
Time duration
16 weeks
Overview
DB’s aerial survey product had to explain the service in three beats (book, map, deliver) while the backend kept pace with gigabytes of imagery per hectare without burning reserved compute.
Problem Statement
Field crews finish a flight and wait hours for tiles and reports because pipelines run sequentially on reserved instances. Meanwhile the marketing site talks past operators who need a booking path that works on a phone at the edge of a field. Slow processing and unclear booking both stall revenue.
Solutions
- Queue-driven fan-out on spot capacity with checkpointed resumption.
- Produce web-ready image pyramids during ingest instead of on demand.
- Redesign the public site around book → map → deliver with a field-friendly contact path.
- Overlay analytics (vegetation, topography) tuned for low-bandwidth rural links.
Design process
Empathize
- Operator interviews
- Flight-day observation
- Cost audit
- Competitor platforms
Define
- Jobs map
- Latency budget
- Booking funnel brief
- Risk list
Ideate
- Pipeline architectures
- Booking flows
- Map UI concepts
- Status notifications
Design
- Service pages
- Dashboard states
- Map surface
- Mobile booking
Test
- Load tests
- Field bandwidth trials
- Pilot surveys
- Ops handover
-64%
Processing time
-51%
Compute spend
1,200+
Surveys / month
Discovery insights
Survey responses gathered during discovery that ranked jobs, friction and priorities before we committed to flows.
When do you need survey results after a flight?
Before packing up on site
46%Same day
32%Within 48 hours
15%Within a week
7%
Biggest friction booking an aerial survey today?
Unclear scope / pricing
38%Slow back-and-forth email
29%Hard to schedule windows
22%Unsure of deliverables
11%
Audience research
How people split across channels, screens and decision criteria. Used to weight information architecture and first-release scope.
Which deliverable do you open first?
- Orthomosaic / map49%
- Analytics overlay28%
- Raw stills14%
- PDF report9%
Where do you usually book field services?
- Phone41%
- Web form33%
- WhatsApp / chat18%
- Email only8%
Outcome
What the client said, and how the engagement landed after launch.
Field teams now see results before they've packed the drone away.
The previous pipeline processed flights sequentially on reserved instances. We rebuilt it as a queue-driven fan-out on spot capacity with checkpointed resumption.
A tiling service produces web-ready pyramids during ingest rather than on demand.
The result is a platform that gets cheaper per survey as volume grows.