New product
Which properties were hit, and how badly, within a day of the storm
Betterview (now Nearmap) · Insurance · Geospatial · Machine learning
When a hurricane or wildfire hits, insurers need to know fast which policyholders were affected and how badly. I designed CAT Response, from a one-month MVP that shipped as a new revenue stream to a second version shaped by what customers asked for next.
Act 1
The MVP
Shipped
Impacted properties on the left; pre-event and post-event imagery on the right, viewed as an overlay or side by side.
CAT Response began as a proof of concept, shaped by our own team’s ideas and by roundtables with our existing customers. From the first idea to a working MVP took about a month.
When a hurricane, wildfire, or other catastrophic weather event hits, insurers need to know quickly which policyholders were affected and how badly. Within roughly 24 hours, flyovers captured high-resolution imagery of the area, and our machine learning and machine vision models graded the damage to every property. That triage told insurers where to send adjusters first. Before relying on it in a live event, we tested the technology against historical CAT events, comparing its grades with their known outcomes.
I designed the product, building on patterns we’d already proven in the Changes panel. Adjusters could view a property’s pre-event and post-event imagery as an overlay or side by side, so a single glance showed whether it was damaged and how badly.
Each property carried a damage dial on a five-step scale: No Damage, Affected, Minor, Major, Destroyed. The model’s grade was a starting point, not a verdict. A predicted dial shows an edit icon; once an adjuster confirms or corrects it, a needle takes its place, so anyone can tell a machine guess from a human decision. An info tooltip defines each level, so every adjuster grades damage the same way.
Every property in the list also carried its value, which set up the question customers asked next.
Predicted dials carry an edit icon; confirmed dials show a needle.
Outcome
CAT Response shipped as an add-on to our existing offering, opening a new revenue stream for the company.
~1 mo
idea to MVP
~24 h
event to imagery
5
damage levels
Hurricane Idalia (2023) in the event view: the storm’s path, boundary layers, and the impacted properties with their damage state.
Act 2
Version 2.0
Designed · paused
After launch, conversations with our customers kept arriving at the same next question. Knowing which properties were damaged wasn’t enough. They wanted to know what an event would cost them, and who on their team should handle each property.
Version 2.0 was the answer, reorganizing the product around four levels, from every active catastrophe down to a single address, and carried the damage scale and the machine-versus-human distinction from the dials through every one of them.
Status
Version 2.0 was shaped by discussions with our customers and fully designed. Betterview was acquired by Nearmap while it was in progress, and the project was put on hold.
1
Event list
→
2
Event summary
→
3
Property table
→
4
Single property
The new starting point: every active catastrophe on one map.
Properties sorted by reconstruction cost. When estimated damage exceeds a property’s value, the figure turns red. Separate “High Damage” and “Verified High Damage” filters keep the machine-versus-human distinction.
One property: before and after imagery, value beside estimated loss, a severity control, and a next step in the same place.
Building footprints filled with their damage level, with a wind-speed scale for context.
Team assignment: properties divided across adjusters, with each person’s workload visible at a glance.
Version 2.0 screens are design comps. Event names, dates and dollar figures are sample data.
Next case study