Technically Right, Effectively Wrong: Why 85% Data Science Projects Fail

Technically Right, Effectively Wrong: Why 85% Data Science Projects Fail


Special: Value Inspiration Podcast

This is a a special guest blog post written by Ton Dobbe – Chief Inspiration Officer and host of the Value Inspiration podcast. Ton recently interviewed Brian on his show, and wrote up his reflections on the podcast below. Read the highlights or listen to it now.

Every week I interview entrepreneurs and experts from around the world to 🎧 share their big idea about new forms of value creation and the potential we can unlock when technology augments the unique strengths of people to deliver remarkable impact.

Engineering for success

I got inspired by the big idea behind Designing for Analytics; hence I invited founder Brian T. O’Neill to my podcast. We explore why so many AI, analytics, and big data projects fail, and what needs to be done differently in order to be successful – creating software products that people find worth making a remark about.

The thing that triggered me most from my interview with Brian

“Sometimes we forget about the value of fun and different ways of engaging with people”

Why did this trigger me? What's the bigger value here?

For as long as I’ve been involved in the business software industry I’ve dealt with product development and design. And I must agree – too often the focus is literally on the checking off ‘the specs’, like ‘It needs to this, and this, and this’, rather than ‘What’s the big picture of what the user trying to accomplish.

I can recall that, over the course of a decade, we developed at least 4 new user interfaces for our time-sheet application. Yes, visually it changed, but users still hated it – simply because they don’t like doing their timesheet. Until that moment where we took that mindset and said: ‘What if we’d remove the need to do timesheet all together.’

So instead of creating User Interface #5 and following the traditional process, we used the data inside the system, connected dots we previously didn’t connect through analytics, and proposed people’s timesheet at the end of the day in their favorite communication app (Slack, WhatsApp,…). The only thing they had to do was to approve it with a simple ‘yes’ in chat or their voice. Fun, engaging, and adoption guaranteed.

What’s the more significant question/opportunity that raises?

Many companies state they are customer centric. However, do we challenge ourselves enough if we really are? True customer centric organizations turn customers into advocates. And one thing they do is ask different questions. Take for example AirBnB. Brian Chesky, one of their founders once explained in an interview that he doesn’t ask customers about the product he already built. He asks about the product of their dreams.

‘We’d ask these questions like, “What can we do to surprise you? What can we do, not to make this better, but to make you tell everyone about it?” And that answer is different. If I say, “What can I do to make this better?” They’ll say something small. If I were to say, “what would it take for me to design something that you would literally tell every single person you’ve ever encountered?” You start to ask these questions and it really helps you think through this problem.

This is exactly what we talked about in our interview. I especially applaud for Brian’s advice on ways to fix the problem:

  1. Define who owns value creation in your product team i.e. who has the responsibility, the accountability, for analytics and data science solutions to produce value?
  2. Spend time developing the soft skill of your engineers.

As he said:

“We have lost the humanity aspect in solution design. We underestimate how important the soft skill aspect is to business and product creation. The moment we fix that we’ll get to see real wins”

Listen to the episode with Brian now

...and why it has the potential to transform the way product teams can create products people find worth making a remark about.