#How does it work?

Sensors to detect the number of coke in each facing row is detected using ultrasound sensors and stored in the processor attached to the cooler. If any form of networking is detected, the data is sent out from the cooler to the server for processing, analysis, notifications of restocking and for other uses.

As the cooler is being restocked, the expected restocking counts in the cooler are compared to what the restocker is entering. This allows the system to determine the purity of the cooler and adherence to the planography.

# What does it measure?

It measures the count of products in each facing row every few seconds (configurable time), allowing to capture the sales at very fine granularity. It also measures the purity of the cooler at the time of restocking.

Does the retailer/bottler/distributor have to do anything different to make your solution work? (if yes, please explain)

The sensor and processors will have to be installed on the back wall of the cooler. Other than that, the current workflow processes that are being followed by the bottler, retailer and distributors should not have to change.

What materials does your solution require?

  • Ultrasound sensors
  • Processor, capable of talking to sensor, storing data and networking (one or more of WiFI, BLE, Cell etc).
  • A mount to mount the sensor and processor
  • Wire harness and wires for connection
  • Power supply

Best guess on cost to implement

RETAIL: CPU – Raspberry pi zero = $5 network card (USB) = $5.35 Sensors = $30 (24 facings) Total = $40.35

BULK (at coke scale) $20 or so

Challenges you/your team ran into

  • Time to build the features that we thought would make the solution very attractive was not there (analytics, machine learning, planogram recommendations per cooler etc)
  • Building and wiring the sensors reliably is a challenge, especially when it has to be done in a day

What you/your team learned

  • it is so much fun to work in a Hackathon

Next steps

  • Build and integrate it with a backhand platform to reliably for
    • provisioning each cooler
    • Analyzing and presenting data from each cooler/groups of coolers
    • Learn off the data and make Planogram recommendations
    • Integrate notification and alerts
    • Possibly allow retailers to make a maintenence call over the same system
    • Too many other use cases

Anything else you want to add

Our company builds IoT systems that learn from the IoT data. We would have loved to demonstrate that capability for the cooler. But due to the short timeframe, we could not generate enough data to create a machine learning model on a per cooler basis. This project is hosted locally on our laptops for now but can be hosted to a cloud, if needed

Share this project:

Updates