Three Keys To Predictive Analytics And Machine Learning: Openness, Speed, And Security

Tina Tang

Part 4 in the 4-part Predictive Analytics and Machine Learning series

This blog series has focused on the challenges faced by those that seek to use predictive analytics and machine learning (PAML) technology to deliver better outcomes to customers and stakeholders.

This series looks at aspects of a recently released Forrester study, Powering The Intelligent Enterprise With AI, Machine Learning, And Predictive Analytics. In the first, my colleague Andreas Bitterer pointed out that only 15% of organizations have adopted PAML. This is despite the fact that organizations see PAML as critical. Indeed, according to the Forrester study:

  • 88% of companies believe the next generation of enterprise applications will be infused with machine learning and other AI technologies.
  • 93% say PAML is important for building more personalized customer experiences.
  • 91% say PAML is needed to drive efficiency with back-end applications.
  • 91% say PAML is needed to drive efficiency of customer-facing applications.

Where is this discrepancy between adoption and intent rooted? A common obstacle to adoption is that organizations are typically split into two sides of the house: the experimental side and the operations side, often at a rapid clip.

Here I want to further explore the importance of a solid PAML platform, but from a slightly different angle focusing on three core attributes critical for driving success: openness, speed, and security.


Native integration with key open source frameworks for machine learning – such as TensorFlow, R, and Python – is critical for PAML success. The openness of your PAML platform directly impacts productivity and efficiency, allowing models scripted in different languages to be reused in different scenarios.

Another factor of importance is your freedom to deploy in the landscape environment of your choice. Maybe you want to deploy on premises in your own data center. Maybe you want to deploy in the cloud – either public or private. Or – as many companies do today – perhaps you need to operate in a hybrid landscape that mixes aspects of cloud and on-premises. Whatever your choice, your PAML platform shouldn’t hold you back.


The data-processing demands of PAML models are immense, and the speed of processing can make or break their uniqueness and usefulness. With an in-memory database approach, you can process machine learning models and score live data in real time – a virtual imperative when it comes to embedding intelligence into operational business processes. It also enables you to combine other advanced analytics such as streaming, text, or spatial analysis with machine learning algorithms.

But fast PAML isn’t only about processing speed; it’s also about the lifecycle of models as you move through iterations in pursuit of continuous improvement. This is why your PAML platform should support a wide range of capabilities that help developers and the operations teams deploy quickly and maintain models with efficiency.

To this end, a robust set of APIs, function libraries, and core machine-learning models can help tremendously. Critically, these should all be categorized for easy selection by developers. Categories might include models based on machine learning tasks such as classification, regression, segmentation, forecasting, recommendations, or link analysis. From there, developers can train the models on specific data sets for specific purposes.

Secure PAML

According to the Forrester report, “Over half of respondents (55%) ranked concerns over new data privacy and compliance as a barrier to adopting PAML solutions.”

In the current climate of malicious attacks and government-mandated privacy laws, PAML solutions must prioritize security. To move forward with confidence, look for a PAML platform that builds in security from the ground up. Real-time masking and anonymization at the database level are critical. Features like these can not only bring peace of mind but also help improve performance.

Move forward

So, my parting message is that yes, you can move forward with predictive and machine learning technology today. You may feel it is risky to do so. I argue that not modernizing your applications with PAML capabilities is in itself a risk. One key to success is an enterprise-proven PAML platform – one that supports self-service for business people, delivers tools that enable the operationalization of PAML, and supports the openness, speed, and security needed to move forward with confidence. Such a platform can help you seize opportunities to drive value for your customers and stakeholders alike.

For an in-depth look into the intelligent possibilities for your business, review the August 2018 Forrester Consulting study, Powering The Intelligent Enterprise With AI, Machine Learning, And Predictive Analytics, commissioned by SAP.

Tina Tang

About Tina Tang

Tina Tang is senior director in SAP’s Digital Platform division, where she leads SAP HANA machine learning, e-sports, and advanced analytics product marketing. With over 20 years in Silicon Valley, she is passionate about tech and diversity. Tina is a founding member of, and serves as a board member of Tina holds degrees from the University of Texas at Austin.