Machine learning with CloudCoins

A few months ago I wrote about how we built our wellness app experiment called “Kubecoin”. It’s a side project, a type of hobby app that we built to take to developer events, to challenge participants to walk more, and to offer an immersive taste of IBM Cloud technology.


We updated the app, and renamed it CloudCoins (because it is built on more technology than just Kubernetes) and we’re experimenting with it at the Cloud Foundry Summit Europe 2018 in Basel, Switzerland, this week. It is built for iPhone and Android.


CloudCoins Concept

CloudCoins is a mobile app, backed by a blockchain system that anonymously converts participants’ steps into a cryptocurrency, as they walk around the conference. Then they can then exchange the digital “CloudCoins” for promotional swag at the IBM booth on the show floor of the event. This time around at the Cloud Foundry Summit 2018 Europe, we’re offering stickers, pop sockets, and charging cables as incentive to walk more, while you think about the technology powering our app!

There are many apps that track steps for points, but I doubt there are many that strive to know as little about the individuals using them, as our app does.

Isn’t anonymity a hindrance to user data research?

Privacy is one of the questions and motivations we’re continuing to explore.

At the beginning of the year, we thought our app would be used for just one conference, but we ended up partnering with an IBM team in Scandinavia, and those developers independently began using it at events in their region.

With their insights and ideas, we’ve been evolving the app a little more. It’s a neat and tidy way to immerse curious people in IBM Cloud. By the end of the year, more than a thousand people will have installed the app, and thousands more will at least be aware of the concept behind the app.

One thing we needed to do was improve the app so that it could work for multiple and simultaneous events. We needed to think more about a distributed hyperledger blockchain fabric, and we needed to segment channels for each of the events (or organizations) on the fabric itself. One event cannot see the transactions of another event.

We can still use the system to collect data – just not personal data. With the collaboration of our Nordic IBM cousins, we’ve been watching the overall usage numbers, footsteps, and the distance walked. So far we are keeping that data in a spreadsheet,  but in the future, we want the app to log this data for us.

What can we learn from the general data?

We started collecting what data we could collect while respecting anonymity:  the number of participants per day, the distance walked collectively, the steps taken collectively, and the total for each phone platform that the app was downloaded to. We wanted to try using some of this data to see if we could make predictions on how many people we could expect at the next event.

Another IBM developer got us started with Watson Studio,  where we could load our participation data and then create a prediction for the next event that is based on the number of attendees and the participation of past events. Here’s what the architecture looks like:


We built a little Cloud Foundry Enterprise app around that prediction data, and we added a new mobile view so participants can see how the prediction measures up to the actual participation at the event this week.


It’s a simple view for now. The original design charted this data for each day of the event, and for footsteps and distance, too. We just didn’t have enough seed data for all that we wanted to do, for this conference. Ultimately, we want to use this data in real time, to automatically adjust the prices of the swag items that we have. We travel with a limited supply of merchandise to exchange for CloudCoins.

The value of participation data

The data that we’re working with can help us automate an economy within CloudCoins. We travel with a finite amount of marketing material to exchange, so if there is a lot of participation, and a lot of walking, we need to adjust the value of each of the items we offer. That way, we can keep the interest alive until the end of the event.

This need for adjustment is no different from any retail experience, with inventory control. It is clear how powerful these technologies can be for maximizing customer interest. As I mentioned at the beginning of this post, CloudCoins is a hobby app. We build it as a side project, with just a few people putting an hour in here or there. Anthony Amanse is the superstar Android and system developer, and Kunal Malhotra is the Watson Studio genius.

Just think what we might be able do if we were full-time!

I’m looking forward to seeing the other work with Cloud Foundry at Cloud Foundry Summit Europe in Basel. If you meet me there, let me know what you think of the app. And look for the CloudCoins app in its next appearance at the EHiN health conference in on November 13 in Oslo, Norway, presented by the IBM Nordic team.