What problem did the client have?

Problems were spotted late, through field walks.

What were the results?

Time to spot a problem area down from about 7 days to 1 day; field visits targeted to flagged zones.

CASE STUDY · AGRITECH / PRECISION AGRICULTURE SOFTWARE

CLIENT · ANONYMISEDAGRITECH / PRECISION AGRICULTURE SOFTWARE

01 · PROJECT OVERVIEW

An anonymised agribusiness company. Name withheld under confidentiality.

Satellite imagery processed into vegetation-health indices, with alerts for stressed zones.

WHAT THE CLIENT WAS DEALING WITH

Problems were spotted late, through field walks.
Agronomists covered large areas on foot and by vehicle.
A pest outbreak or a patch of water stress could grow for days before someone saw it.
Walks were scheduled by habit, not by need, so time was spent on healthy fields while troubled ones waited.
The company wanted a way to see the whole farm at once and direct people to the places that mattered.

02 · WHAT WENT WRONG

Every engagement hits problems. These were the ones that mattered here, and what we changed so they did not come back.

PROBLEM 01During the first weeks, cloudy conditions left long gaps between usable images, and the system raised alerts that were really artefacts of cloud edges and shadow
Agronomists stopped trusting the alerts. We improved cloud and shadow filtering, showed the age of the last clean image on every field, and suppressed alerts based on a single poor image. We also added a way for agronomists to mark an alert as false, with a reason, which gave us a clear list of what to fix next.
PROBLEM 02We also added a way for agronomists to mark an alert as false, with a reason, which gave us a clear list of what to fix next
PROBLEM 03A second hurdle was field boundaries
Some of the boundary data the company held was out of date, so a few zones included track, trees or a neighbour's crop, which distorted the indices. We used the agronomists' knowledge and recent imagery to redraw the boundaries, added an inner buffer so edges are ignored, and gave them a simple tool to correct a boundary themselves.

03 · THE PROCESS

We built the system in layers and tested each one with the agronomists.

01
Discovery
we mapped field boundaries and crop calendars, and sat with agronomists to learn which problems they cared about and how they described them.
02
Imagery pipeline
we set up regular intake of satellite images, cloud filtering and the calculation of vegetation indices for each field.
03
Calibration
agronomists visited zones the system flagged and zones it did not, and told us whether it matched what they saw. We tuned the thresholds from those visits.
04
Alerts and rollout
we added alerts and a dashboard, then rolled out to more regions in steps.

We held a weekly review with the agronomy lead, focused on false alarms and missed problems.

04 · RESULTS

Measured against the way the work ran before the engagement began.

NOW 01
Agronomists start the day from the map, not from a fixed route.
NOW 02
They visit flagged zones first and use the rest of their time on planned work.
NOW 03
The findings they record after each visit improve the rules.
NOW 04
The company keeps a season-by-season history of each field, which helps them plan inputs for the following year.

05 · WHAT WE BUILT

Images are pulled on a schedule, clipped to field boundaries and screened for cloud and shadow.
Vegetation indices are calculated per zone and compared with the same crop at a similar stage, and with the same zone earlier in the season.A sudden drop matters more than a low reading.
Alerts are raised only when a change persists across more than one image, to cut noise.
The dashboard shows a map with flagged zones, the trend for each, and a place to record what the agronomist found on the visit.
The recorded findings feed back into tuning the rules.

AFTER LAUNCH

Monitoring and controls
Handover: We documented the pipeline, the thresholds and how to add a new crop. An agronomy analyst on the client side made the first threshold change with us, and now owns the tuning.

06 · QUICK ANSWERS

What problem did the client have?: Problems were spotted late, through field walks.
What did Brandsmashers build?: Satellite imagery processed into vegetation-health indices, with alerts for stressed zones.
How was the work run?: We built the system in layers and tested each one with the agronomists.
What were the results?: Time to spot a problem area down from about 7 days to 1 day; field visits targeted to flagged zones.

YOUR TURN

Tell us what is stuck. We will come back with a plan and matched, pre-vetted engineers.