Learnings from running an online competition to solve a mining problem

Ore-X_grade_processingOre-X_grade_info

I’ve been asked on many occasions about what we have learned from running an online competition last year. Today, I finally got around to writing this blog so the learnings are written down and passed on.

For those who have not come across it, late last year, GlassTerra teamed up with Unearthed to run an online competition for Goldfields. The goal of the competition was to find a way of turning photographs of the face in a drive into a grade estimate. For Goldfields, the competition was a chance to solve a small problem which was not economically viable if trying to solve in traditional ways but would drive an incremental productivity gain if solved. For Unearthed, it was a chance to extend their product offering; and for GlassTerra, it was a chance to show the capabilities of our new technology to the world.

The competition delivered a success for everyone, but as with all things you do for the first time, we made mistakes and learned a few things on the way. So here it is:

1. When is an online challenge a good mechanism to procure innovation?

When you are trying to find the how but not the what.

  • The answer is an algorithm.
  • The algorithm should it be found will deliver a short payback period.
  • The real world data upon which the algorithm is to operate already exists.
  • End users want the result right away.
  • It is easy to determine if the algorithm is giving a good answer.

2. What are the must haves in terms of running the competition?

Say it loud, say it simple, say it often.

  • Partner with industry and get them to shout out on their email lists and use publicity if you can.
  • Make the reward exciting – money, reputation, ability to create a new start-up, mastery, bragging rights.
  • Don’t ask to own the IP as this will turn away many potential contestants.
  • Explain the problem “like I’m five” if you want to get more people involved from outside the industry. No one else really understands our jargon.  
  • Show pictures and visuals. The majority of mining problems have spatial component, so show it.
  • When you have a winner, deploy the result in some minimal way to the end users straight away. Capitalise on the high.

It is great to see that Unearthed now has another competition underway and this time it is for BHP Billiton.  You will find GlassTerra there presenting the input data in full 3D.  Good luck to everyone who has a go at it.

Till next time,

Christian

 

Leave a Reply