<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=95044&amp;fmt=gif">
Back to blog

Accelerate your Data Value Journey with Automated Machine Learning

Tuesday, March 27, 2018

Companies that are serious about getting real value out of their data make substantial investments in analytics. A standard data and analytics solution typically involves three components:


[1] An Enterprise Data Warehouse (EDW), data repository, or something analogous to that: basically the place where you store all your data.


[2] A tool or process to Extract, Transform, and Load (ETL) data: there needs to be a system in place for pulling data out of source systems, transforming (cleaning, filtering, validating, etc.) that data, and loading it into your data warehouse or data repository.


[3] A visualization tool: even clean, filtered, validated data that has been whipped into a consumption-ready format provides no value sitting in tables in your data warehouse. The value of data comes in the form of insights into your business, and the natural way to understand your data and gain insights from it is by visualizing it.





One thing is clear: a data and analytics solution based on these three components is focused on the past; it is a rear-view mirror look at your performance. Deep-diving into your historical data may give a good understanding of what has been happening with your business, which is very important but, unfortunately, not enough to compete in today’s world where competitors are using increasingly more accurate predictive analytics to make smarter business decisions.



Gartner has created a framework to describe the main stages of the Analytics Value JourneyThis framework describes four stages in a path towards increasing the value an organization extracts from its data.


The first two stages, the DESCRIPTIVE and DIAGNOSTIC stages, involve analysis of historical business data to understand what has happened and why it happened. These are the stages where most companies that have made a standard investment in analytics (EDW, ETL, Visualization) live. The next two stages, the PREDICTIVE and PRESCRIPTIVE stages, are forward-looking. They are about using accurate predictive techniques to anticipate what will happen and to prescribe how to influence those outcomes. These are the stages that add the most value in the Analytics Journey. They are also the stages that are the hardest to break into because of the difficulty involved in deploying good predictive techniques.


Companies reaching the PREDICTIVE and PRESCRIPTIVE stages understand the challenges: Data Scientists with enough skill in Machine Learning and state-of-the-art predictive techniques are expensive and hard to find. These individuals must possess the intersection of at least four specialized skill sets:


[1] Data Skills (SQL): They must be data ninjas that can wrangle data and do ETL.


[2] Business Acumen: They must have domain knowledge of your industry and understand how to ask the right questions to drive business results. They must also know how to frame those questions in a way such that they can be tackled through Machine Learning.


[3] Math & Statistics: They must be deeply familiar with a variety of Machine Learning algorithms and statistical techniques.


[4] Hacking Skills (R & Python): They must be able to implement a variety of algorithms in code by training, tuning, and deploying them to make relevant predictions.


No doubt it is difficult and labor intensive to break into the final PREDICTIVE and PRESCRIPTIVE stages in your Analytics Value Journey. The good news is that there is a path of least resistance: leveraging an Automated Machine Learning platform.


An Automated Machine Learning platform takes care of the heavy lift involved in building and deploying Machine Learning models. The idea is that analysts with the skills to move data around and with an understanding of their business can take advantage of an Automated Machine Learning platform to build and deploy accurate predictive models. This gives rise to the concept of the Citizen Data Scientist. A Citizen Data Scientist does not need to have a deep mathematical understanding of multiple Machine Learning algorithms, how these algorithms are implemented in code (trained, tuned, deployed), and how to evaluate the performance of these algorithms to pick the best one for specific use cases. Instead, a Citizen Data Scientist capitalizes on their intuition and business understanding to ask the right questions and prepare a dataset to address those questions. An Automated Machine Learning platform can then consume that dataset and:


  • Pick the most appropriate Machine Learning algorithms to use for that use case
  • Train and tune models based on those algorithms and evaluate their performance
  • Select the best performing model
  • Deploy that model to make future predictions on new data


Organizations that are trying to maximize their data and analytics investment by breaking into the world of predictive and prescriptive analytics can capitalize on their existing analysts working in conjunction with an Automated Machine Learning platform to help them get there. An Automated Machine Learning platform can also help full-fledged Data Science teams scale across their organizations by empowering analysts, who were not previously working on Machine Learning, to build and maintain a library of curated predictive models that can then be reviewed by more experienced Data Scientists.


An Automated Machine Learning platform is the path of least resistance to enhance a traditional data and analytics investment. Such a platform can complement your EDW, ETL, and Visualization tools by adding predictive and prescriptive capabilities to your business, allowing you to extract the most value out of your current data solution.


Want to dive a little deeper? Jorge will be hosting a free webinar Wednesday, April 18th at 10 AM MST. Sign up now to reserve your space!


Register Now


Jorge Zuloaga, Sr. Director of Data Science at Big Squid

Written by Jorge Zuloaga, Sr. Director of Data Science at Big Squid

Jorge earned his MS in mathematics from University of Waterloo and Ph.D. in computational physics from Rice. His work includes particle physics & quantum plasmonics. After academia, he focused on ML applications in healthcare. Jorge joins us as Sr. Director of Data Science. Jorge lives in Utah with his wife, enjoys all things math-related & playing squash

Recent Posts

Comment Below

Subscribe to Email Updates