Why Smart Talented People Build Bad Data Science Solutions
Nobody sets out to build a bad data science solution. Data Science, Machine Learning and AI are truly disruptive technologies set to change the world. However even the brightest minds with the best training can often struggle to produce real business value. Here's why.
Ultimately technolgy is only valuable when it solves valuable business problems which are well defined
The problem is when
- Business don't enough understanding or experience with new technologies to properly define or manage data science projects effectively.
- Data professionals don't understand the business well enough to realize that many project specifications need to be changed to meet the true business objective.
Often stakeholders or practioners focus too much on algorithms and optimizing for accuracy.
Accuracy is often the worst metric for success
Would you rather be accurate or profitable?
On the face of it a more accurate system is better.
If you had a trading system which was 10% accurate vs 90% accurate which would you pick?
You can't really answer that question until you know how much you gain from being right vs how much you loose from being wrong.
How much is the cost of running the trading system?
What is the cost of obtaining the data and is it reliable?
Can you scale and put things into production?
These are just a handful of the questions you should be asking yourself when you set out to solve a business problem.
Do you want help to reivew your data science projects?
I provide advisory to tech startups and mentoring to data scientist to help review this and many other business and technology considerations.
I help data science teams to produce better results that are inlined with true business needs. And I help businesses uncover their true business needs and properly define data science projects to fulfil them. Happier business, happier data scientist.
If you'd like to get on the waiting list for a free consultation please send me a message using the form below