2 Reasons why a Covid-19 Bluetooth app isn’t built over the weekend

Gepubliceerd op april 21, 2020

Over the last 10 months the developers of the Dutch start-up Airchip gained hands-on experience working with Bluetooth. We are building an app that helps you get drinks faster while visiting an event by reducing queues. Now we want to share our code and knowledge to help fight the spread of the Coronavirus. If the goal is to get a technically good app? Let us help!  

Initially our goal was to build a very reliable mobile order and payment app for events and festivals. After investigating many Bluetooth SDKs we built in-house a cross-platform native Bluetooth controller to exchange data when 4G fails, like it often does at busy events. This spring Airchip was ready to show the Dutch event industry what we can do and prove our market fit… And then came the Coronavirus. Like it did for so many startups, it pulled the rug out from under all plans.

But here comes the new Covid-19 tracking app from the Dutch government and now Bluetooth is suddenly the talk of the town, and Airchip, by coincidence withexpert knowledge. This is what we learned:

  1. The signal strength of Bluetooth is a bad indicator for close proximity
  2. Big data analysis is needed to correct for device and mobile phone operating system variations

     

So, together with more than 600 other companies, we rushed to prove our value by applying for the Dutch government Covid-app tender. An extremely hectic selection procedure followed where companies suddenly had to prove technical excellence in days but also reply to deep societal questions that arise with these sorts of applications. In hindsight, a recipe for unstructured frameless assessment and a lack of technical depth from the applicants.

 

 

Airchip’s Contribution

While the public opinion was (and still is) debating about privacy and the hasty procedure, Airchip just started to build on the technology we know well, Bluetooth. Within 4 days we were ready to test our stack because we believe that any Bluetooth role for contact tracking should be based on scientific feasibility assessments.

On Friday 17th of April 2020 Airchip conducted dozens of controlled field tests with the purpose to gather RSSI (Received Signal Strength Indicator) data and research the statistical correlation with proximity. Multiple devices, distances, passing speeds, and over 40.000 BLE (Bluetooth Low Energy) data packages have been recorded. Ready to be analysed in depth and contribute to a proper working application.

The first results affirmed our presumptions that even during controlled conditions with devices on fixed positions the average signal strength fluctuates heavily. With these small distance variations it is hard to detect a clear correlation between proximity and signal strength. The fluctuations can be the results of attenuation, reflection, hardware anomalies and much more. To train an algorithm the difference between 2 centimeters and 200 a lot of data is needed.

That same afternoon we heard that Airchip was not one of the elected seven that could go and participate in the ‘Appathon’ following that weekend – in retrospect maybe better due to the negative reviews independent experts outed over the procedure and the lack of depth.

But should this be the end for us?

Open for collaboration

Airchip cannot decide alone about the moral complications this app brings and we think we do deserve a seat around the table concerning the technical feasibility. Sharing and comparing app data is crucial to validate if the app works as expected, but this requirement radically contradicts the privacy of the user. Therefore all validated test data should be harvested and aggregated, even from competitors and start-ups.

 

"We will gladly offer our insights and open source our codebase to whomever wants to help solve this crisis."

We strongly believe that our knowledge, experience and data could be of great value to the selected app developers, many of whom do not have any proven experience in the field of Bluetooth. We will gladly offer our insights and open source our codebase to whomever wants to help solve this crisis.

Airchip hopes the well intended initiative of the Dutch government will not be eclipsed by the ethical discussion. A discussion that needs to be held but is a separate one from the technical feasibility research. The big tech companies may have won the contract but it is their social obligation to reach out to friends and foes to make this app work.