Blockchain, IoT, AI, machine learning, deep learning.
Major investments are being made into these technologies; or, at least, whichever solutions are branded as such. Most people do not even know what the terms mean, or how the tech behind them works. In the US, recently, the Long Island Iced Tea company tripled its stock price in a single day by simply tacking on the word “Blockchain” to its name.
As the digital race heats up, a number of companies and government agencies are now putting together big digital transformation and data science initiatives. Most have these at the top of the agenda as a strategic priority with full executive support.
And most will fail.
More than seven out of 10 large IT projects fail—either miserably by not achieving their intended outcomes, or by simply missing targets and not delivering a return on investment.
Most fail for the same reason: behavioral science.
To be clear, it is beyond any doubt that data science and digital technologies are the technology components of an industrial revolution that will fundamentally change how markets, companies, and governments work.
As co-founder and chairman of the data science non-profit Flowminder.org, I’ve had the opportunity to participate in pioneering data science projects, such as the first uses of mobile big data for earthquake response (Haiti 2010, Nepal 2015) and the use of mobile and satellite data to predict the spread of infectious diseases (cholera, Ebola).
These collaborative projects and related academic research gave us privileged access to the big data units of UN agencies as well as some of the largest multinational telecom groups; and how they leverage data to get relevant insights or to transform their business models.
Since 2010, we have been exposed to numerous multimillion-dollar digital transformation and data science initiatives in the private and public sectors. The common patterns between successes and failures has confirmed a realization. Successful digital transformation is not about understanding technology, but human behavior.
Two Common Pitfalls
Behavioral science is the study of human behaviors and decisions, including irrational ones. To illustrate, here are two common pitfalls that we have seen drive digital transformation initiatives into the ground; and also some thoughts on how to avoid them.
First, too much focus on tools.
Hadoop or Spark cluster? Supervised or unsupervised learning? Is it even worth trying to do anything before a cloud solution is in place? Because it is technology, it is very easy to suffer from technology myopia.
My own experience is that tradecraft is much more important than the technology. As an example, in 2010, our team supported the UN Haiti earthquake recovery effort by providing population maps that continuously computed the anonymous location data of three million Digicel subscribers.
On a laptop.
Every computation took a few days, but we got the job done. Could we have done it faster with more hardware? Of course. Would it have mattered? Probably not, as the UN had weekly decision cycles, and we knew from close collaboration what information they needed.
Our tools and analyses were fit for purpose.
Is it more important whether it is a convolutional neural network or a random forest computational model? Or has it more to do with the business relevance of the insights?
For a majority of companies, first defining the business case will determine which tools and technologies are fit for their purpose.
To achieve this, executives need to have deeper understanding of technology, and data scientists need to have a deeper understanding of the business models. And most importantly, they need to understand each other.
The second pitfall is going too big.
A classic decision trap is overdoing transformation and innovation efforts. It appears to be intuitive; the more important the topic, the bigger the need for coordination and planning.
Grander visions, longer timeframes, the most expensive international consultants, get more people involved, earmark bigger budgets, all to minimize risk and guarantee a successful project.
Just avoid talking to some boring management professor who will tell you the research indicates a strong correlation between project size and too much planning on one hand and risk of failure on the other. Total buzzkill.
One solution is to allow a portfolio of independent and sometimes overlapping and even competing initiatives and then scale the ones that get traction. It’s classical risk management as it diversifies the risk. I will bet that the most successful projects will not be the ones that were predicted to be so from the beginning.
Case in point, even a global leader such as Google shuts down 100 unsuccessful innovation and transformation projects each year. They should know what they are doing, and experimentation is the norm.
Don’t have the capacity to do this internally? The Asian Institute of Management’s new data science research laboratory, ACCeSs@AIM, would make a great collaboration partner. Launched on 08 March 2018, it is powered by a 500-teraflop Acer supercomputer (or the computing power of 250 high-end laptops). Some of the top young minds in the field will work on the real-world problems of top corporations in the lab.
These are just a few insights I gathered from working with some of the largest companies and organizations around the world on how to develop strategy and business cases around their data assets. How do they compare to your own experiences?
I am always interested in discussing specific digital challenges and opportunities and hope to meet you in an AIM caseroom very soon.
Dr. Erik Wetter is a visiting professor at the Asian Institute of Management and faculty member of the Stockholm School of Economics (SSE) where he supported technology companies and corporate innovation as Director of SSE Business Lab. He works with ADB, the UN, World Economic Forum, and multinational companies. Catch Erik speak about Data Science Strategy on 22–23 March 2018 (for details, e-mail JEspanol@AIM.edu or call +63.2.892.4011 ext. 2804) and Negotiation and Sales on 20–21 March 2018 (MMartinez@AIM.edu; +63.2.892.4011 ext. 2818); or e-mail Erik at EWetterMT@AIM.edu.