The Opportunity
Maven has developed a sensor that turns any pet collar into a wellness monitor: tracking activity, sleep, resting respiratory rate, heart rate, water intake, and scratching behaviour for dogs and cats. The challenge wasn’t the hardware. It was enabling the software layer to understand when something actually mattered and explain it to owners without creating noise or anxiety.
Alerts without context lose meaning fast
The existing notification system was hardcoded: thresholds triggered without considering the pet's current overall wellbeing and environment. Owners were being bothered by notifications that didn’t add much value and caused disengagement with the app, such as increased water intake on a hot day.
Sensor data alone doesn't tell the full story
A single metric rarely means anything in isolation. Disrupted sleep combined with elevated resting heart rate and reduced activity reads very differently than any one of those signals on its own. The platform needed a way to correlate signals across dimensions and across time.
Pet wellness conversations are high-stakes and hard to scope
Owners want to ask questions about their pets. But a wellness platform isn't a veterinary service, and the line between data interpretation and clinical advice is one the system had to hold reliably, across tone, memory, and scope.
The Solution
The existing setup worked on a simple rule: if a metric crossed a threshold, send a notification. No context, no memory of relevant events, no awareness that a heart rate spike during a run is unremarkable while the same reading during sleep is worth flagging.
For this initial phase, we built two workflows on top of Maven's existing data infrastructure. The first runs periodically across all of a pet's signals and looks at them together rather than individually. Before generating a notification, it weighs the pet's personal baseline, profile, history, and current environment. Each notification is classified by urgency and written in a tone that Maven’s team helped calibrate: informative and calm, never alarmist. Furthermore, its structure allows the veterinary team to add notification rules, thresholds and criteria without touching the underlying logic in the future.
The second workflow is the conversational layer inside the app. Owners can ask about their pet's data in plain language, for example: how their pet spent the night, whether the last week of activity was normal for the season, what the resting respiratory rate trend looks like over the past month. The system fetches only the data relevant to the specific question, responds with concrete figures, and is explicit when coverage is insufficient to give a reliable answer. After each conversation, meaningful events - symptoms mentioned, patterns flagged, context the owner shared - are stored and carried into future sessions, building continuity over time rather than starting from scratch. When questions cross into clinical territory, it redirects clearly to a vet.
The architecture was built to grow: new signal types, new clinical rules, new input formats and real-time data streams can be added without reworking the foundation, as the platform moves from controlled pilots to broader deployment.
The Impact
Pet wearables are a growing category, and most face the same ceiling: a struggle to turn rich hardware data into relevant and reliable insights for the pet owner. Pet wellness is evaluated through multiple lenses, but those perspectives are rarely connected or considered alongside the pet's overall wellbeing and environment.
What Maven now has is a foundation designed to support a holistic approach to pet wellness, by treating notification intelligence and owner communication as design problems worth solving properly - one that can scale as the platform grows, as the veterinary rule sets evolve, and as owners come to expect more from the data their pet's collar is already collecting every day.
