Vink Intelligence

Case study · Energy · Hospitality

Asset Circles

An AI-driven energy and comfort platform built for a holiday-park operator, turning live sensor and weather data into proactive, plain-language advice, not just another dashboard nobody has time to watch.

The problem

Comfort and energy waste go unmanaged in lightly-staffed vacation properties. A holiday-park operator with a main building and several individually-sensored guest lodges had climate sensors and energy meters in place, but no one watching them. Waste and comfort risk built up quietly, unit by unit, night after night.

What we built

Instead of another dashboard, we built a system that watches the data continuously and tells staff exactly what to do and when, such as "close the blinds before 07:30," with an estimated cost and comfort impact attached to each recommendation.

Architecture

A queue-based ingestion pipeline, organization-agnostic and configured per client through data rather than code.

  1. 1

    Sensors, meters & weather

    Climate sensors, energy meters, and weather forecasts feed the pipeline continuously.

  2. 2

    Queue-based ingestion

    Raw payloads land unmodified in a queue, so historical data can be reprocessed safely.

  3. 3

    Normalization

    Raw readings become per-building, per-metric measurements: energy, temperature, CO₂, humidity.

  4. 4

    Daily rollups

    Per-building energy, climate, and occupancy summaries are aggregated every day.

  5. 5

    Occupancy detection

    A CO₂-based model infers whether a unit is occupied, cross-checked against humidity and temperature.

  6. 6

    Proactive advice engine

    Weather-aware thermal modeling and per-advice detectors generate ranked, plain-language advice.

  7. 7

    Portal & feedback loop

    Advice reaches staff through a REST-backed portal; accept/reject decisions feed back into confidence scoring.

Eight kinds of proactive advice

Each advice type is its own detector, drawing on thermal modeling, weather forecasts, and learned per-building baselines.

Close curtains before heat

Predicts when indoor temperature will cross a comfort threshold and advises closing curtains ahead of the peak.

AC pre-cool for comfort

Pre-cools a unit ahead of a predicted hot afternoon when occupancy is likely.

AC eco mode on low-risk days

Switches AC to eco mode when both peak temperature and occupancy risk are low.

AC setpoint raise

Raises the setpoint during expensive electricity periods when comfort risk is low.

Close sunscreen before solar gain

Infers window orientation and solar exposure, and estimates the temperature rise avoided. The most sophisticated detector, with a self-learning feedback loop.

Night cooling before a hot day

Detects cool, dry overnight windows suitable for ventilation ahead of a forecast hot day.

Reduce night baseload

Flags unusually high overnight electricity draw against a weather-normalized historical baseline.

Vacancy mode for lodges

Identifies the statistically quietest booking window per unit and recommends automatic HVAC setback.

Results

127

advice templates across 8 categories

~€30–40

per week estimated savings at one example property

0.71–0.78

typical occupancy-detection confidence

15 min

sync cadence between sensor reading and pipeline update

Want something like this for your business?

Asset Circles is one example of what a connected, detection-driven AI system looks like in production. If your business has data nobody has time to watch, we'd like to hear about it.

Get in touch