COVID-19 has changed Predictive Analytics more in the last 6 months than the evolution did in the last 6 years

It’s nothing new 

It is December 1993 Susan, a sales manager in a major manufacturing facility has a great idea to improve team efficiency by introducing a new report on the company’s mainframe.  Susan understands what she wants and already dreams of the new, bright future. She takes her idea to the IT department. Much to Susan’s surprise, they do not see this as a priority.

After some months, George, a clever software engineer, gets going with assembler and some mainframe tools. One month later the report is ready (he even worked on it during the weekend). It’s not what Susan expected, making this a disappointment. George feels he did what Susan asked for, especially given the time pressure he experienced, also feeling this is a disappointment.

This story is the classical business-IT alignment story. It’s difficult to bridge the gap between IT and business professionals, because of differences in drivers, culture, and a mutual ignorance for each other’s knowledge.

Yes, I can already hear you saying: “Now we have new tools like DevOps, making this a thing of the past – trust us all is going to be fine… in the (or is it ‘a’) next sprint.” The reality is that new methods and tools were developed to empower the business people to do much of the work themselves, freeing up time for IT to work on the heavy lifting. 

Let’s fast forward to 2020. Susan has achieved her efficiency dream by implementing the report. Susan opens the companies’ favorite business intelligence platform. The company has gone through a selection process, looking at tools like Microsoft Power BI, Qlik, Tableau, Holistics… No, she is no longer going to ask to develop the report in assembler, wait for months, and then come to the conclusion it isn’t implemented the way she wanted it to be. By her mid-morning coffee break, Susan finished the report she dreamed of and she can share it with her team. She feels like an empowered user. 

This is the story of descriptive and diagnostic analytics. It centers around looking at how things happened. 

What if we bring this approach to predictive and prescriptive analytics? It’s pretty clear that the business opportunities in the field of predictive analytics are enormous. This industry is estimated to account for USD 23,9 Billion globally by 2025. There are predictive and prescriptive analytics use cases with immense value in any industry.  

Can this story apply to predictive analytics? What would you be offering your business users? 

Mathematicians and IT professionals have come up with the Machine Learning required to deliver predictive analytics. Machine Learning allows people to create models based on historical data, and use these models to predict what is expected to happen and what can be done to influence the outcome. Initially, this required technical programming and mathematical expertise; skills that are often not easy to pick up for business users. It feels like making a mainframe report in the 90’s. 

For business users, it is hard to see the value Machine Learning can bring. The path to actual results is usually long and complex. Many companies report they find it difficult to be successful through predictive analytics. 

At Tangent Works, we are currently offering a complimentary Gartner research on this topic. Highly recommended to check it out! 

 

Structural Change and COVID-19 

So, you are proud! Your company has created a set of Machine Learning models. It was not easy and it took a long time, but experiments show they are pretty effective. Then the COVID-19 pandemic hits… 

COVID-19 seems to be causing tremendous data drifts, rendering masses of Machine Learning models obsolete. COVID-19 is a structural change. Companies need to deal with the new normal and need to do so fast. In today’s rapidly moving market, this isn’t the first structural change and it certainly won’t be the last. 

Businesses realize the need to redefine their current Machine Learning models and strategies after being affected by data drift, changes in value propositions and changing capabilities. They cannot rely on the “assembler–like development approach” of the reporting story above, as the resulting process is too labor-intensive, too expertise-intensive and far too slow. They need better ways to deal with their challenges, ways that are agile (building new models fast) and scalable too (dealing with many models). 

Machine Learning is being democratized and brought very close to the business users. It’s now often called ‘augmented’ Machine Learning. It has become easy to use, just like the business intelligence tools, empowering business users to live up to their dreams. 

 

 

In times of structural change, people become aware that a handcrafted modeling approach does not provide the necessary scalability and agility. It has become impossible to manually build and fine–tune all the models a company requires, to maintain them, and to find enough competent data scientists to go through this process (while also dealing with all other data-related challenges). 

A popular approach to overcome these obstacles is AutoML (Automated Machine Learning). This approach relies heavily on brute force (and expensive) computing. While improving the situation significantly, not all problems are remedied by it. In many use cases, especially those impacted by structural changes, this approach still doesn’t offer the speed that is vital for businesses to anticipate what comes their way and handle quickly. Moreover, this approach often struggles with difficult feature engineering, a main aspect of many business use cases. 

The recent structural changes have shown that so-called augmented, democratized Machine Learning, has become a real necessity for supporting business users and (citizen) data scientists in their analytical projects. This is why in the last six months, COVID-19 has advanced predictive analytics to a larger extent than the evolution of the last six years have managed to do. COVID-19 forces us to be more agile, flexible and scalable to deal with data drifts, dynamic value propositions and changing capabilities. But how do we accomplish this? 

 

How Tangent Works InstantML helps 

How do you easily set up a robust, scalable and agile predictive analytics solution for time series data? 

This is the question drives the Tangent Works Research and Development team. 

The Tangent Works solution, TIM, uses its InstantML technology to offer predictive analytics for time series data, such as forecasting, anomaly detection and classification. This is what it entails: 

• A robust engine, offering augmented predictive analytics in a lightning-fast, automated, accurate and explainable way.
  Let’s break this down: TIM automates both feature engineering and model building. This allows TIM to significantly surpass other approaches in speed, without sacrificing the accuracy. On top of this, TIM’s models are explainable, presenting users with an under-the-hood view of what the resulting outputs are based on. This is all available through a single REST API.
• Easy accessibility from various platforms.
  TIM seamlessly integrates in existing cloud infrastructure, business intelligence tools, data integration tools, IoT landscapes and more.
  If you’re eager to find out how this works, keep an eye out for this blog; we will update with a post focusing on this soon. 
• Straightforward integration in your current data architecture.
• An emphasis on business value.
  Predictive analytics is all about the business value of the actual use cases. This is why Tangent Works develops and maintains the TIM Use Case Library and the TIM Value Framework: to demonstrate the business value of predictive analytics.
• A partner-driven approach.
  Tangent Works has grown – and continues to grow – a global partner network of companies with data and AI experience as well as vertical domain knowledge. This greatly increases the chances of successful implementation.

If you want to be convinced of the value of AI, ML and predictive analytics even more, check out the Harvard Business Review article on Building the AI-Powered Organization.

To learn more about this Machine Learning technology revolution, check out Scott Bergquist’s, VP Business Development Tangent Works US blog post.
 

What are you providing to Susan to give her the wings she needs to drive business value with predictive analytics? 

 

Dirk Michiels is the former CEO for Tangent Works. He has a 30+ year track record in management and leadership of sales, product and business development.  Over the years, Dirk worked on strategic projects in oil/gas, manufacturing, energy industry on vertical business applications, IoT (SCADA) and data and AI. He worked on shaping a global solution for customer care and billing for the energy sector. Dirk believes in partnership as the fuel for mutual growth. Prior to Tangent Works, Dirk worked for Nokia, Ferranti, Innogy and as an IT/business strategy consultant.