The Play of Predictive Analytics

By:  Bhawna Manchanda, Big Data Architect, Product Engineering Services, Happiest Minds Technologies

Driven by the combination of ever-growing market competition and evolving consumer demands, there is a tremendous pressure on software engineers and stakeholders. With no scope for development delays and production failures, there is a huge drive towards making these processes automated. Engineering decisions have become more critical and always carry a huge impact on businesses. However, how many times do we take these decisions based on any relevant data?Not even once.

The paradigm change introduced with Digital Transformation has challenged the traditional software engineering approaches across many dimensions.

Customer Interface: Users are now accustomed to the instantly responsive experience from leading market players that provide real time value, mobile functionality, and a user-friendly interface. All time availability and real-time scalability have become an implicit non-functional requirement and are no more a luxury.
Technology Stack: Complex interfaces, volumes of data and real time need leads to too many assembled components. This coupling further complicates the Automation task.
Delivery Timelines: With Innovation being a game changer, time to market is a crucial key to success for any business. Hence, delivery timelines are getting skewed. Too many assembled components put extra pressure on testing because of many corner cases.

All these challenges call for an automationand control around the Software Delivery process. There is a definitely a need for making decisions proactively rather than taking actions reactively.Predictive Analytics can support that shift towards the strategic management of IT production by providing development team, testing team and related managers with communication, planning, and monitoring capabilities.

Many other industries have made use of advanced analytics, but software development is the way to go. Predictive Analytics can be an apt fitment to solve problems in the software development lifecycle (SDLC). Throughout the SDLC, developers produce data as they plan, build, test, and deploy their software, whether as part of an agile development environment or a traditional waterfall. This data is hardly used to make any proactive delivery related decisions.If planned strategically, this data has a potential to help different stakeholders across entire SDLC.

If strategized well in advance, software development logs and related tools can use Power of Predictive Analytics to achieve more efficiency in their process and procedures. Some high level examples can be:

• Predictive Analytics helps Developers in the SDLC
  Get warning if a code check-in isa problem
  Raise alert if a code does not follow Nonfunctional requirements
  Rank code reuse or function/class usage
  Alerts for interface noncompliance
  Interesting implementations can be integrating the data from the application log, server logs,test logs, machine performance logs to provide exhaustive dashboard advising developers and managers on the stability of the application.This can also help in identifying any anomalies in a particular stage of process flow or at a particular page or even predicting a performance bottleneck.

For one of our client, based on the customer-browsing pattern, site page navigations map we could find anomalies / failure points on certain pages of their e-commerce site, which inturn helped them prioritize their task as per their Product roadmap.Integrating Predictive Analytics with project plan data can also train the system to give data-driven effort estimates for future requirement for similar type of activities.

• Predictive Analytics helps Testers in the SDLC
  Identify most crucial test cases based on frequency , impact, time required to solve and criticality
  What bugs should be prioritized based on their past impact
  How critical is individual component
  What are most crucial test cases
  Corner case identification
  Maximum benefits can be made by defining data drive test optimization plans. From defining the use cases, implementing the same and figuring out corner case leakages to planning and estimations for testing based on previous runs, everything can be made data driven. This can reduce the number of late stage surprises and planned management of engineering resources.

Predictive Analytics will help not only in alerts, but also in root cause analysis. Are there some specific scenarios relationship between data items which lead to failure?

• Predictive Analytics can also help Maintenance/Support Staff in the SDLC
  If there is going to be a performance bottleneck
  Predictive Load balancing
  Predict site / server  breakdown
  Raise Security Alerts in case of Security intrusion
  Invalid usage of Roles and responsibilities
  Predictive alerts based on application usage and data usage pattern
  Predicting device failure or Network failure is the most common use case of Advance Analytics for supporting staff. Any deviations from the set patterns can be easily picked by Machine Learning algorithms ensuring no downtimes at all.

Implementing all these things needs strategic inclinations and decisions in the same direction. Therehave to be planned efforts around.

Predictive Analytics helps Integration in the SDLC
Integrating the data from multiple planning and development phases like system logs, application logs, application navigation maps, network logs, test management systems etc. Put together they can make entire delivery data driven.

Predictive Analytics helps in Visualizing in the SDLC
Identify the main KPI for each phase and have an exhaustive Dashboard to display data like warning counts, error counts, application failures, test failures, Bug resolving efforts, networks stats etc.

Predictive Analytics helps do Root Cause Analysis the SDLC:
Drill down to the root cause with detailed data analysis of each phase or process. This can be useful for future improvements and planning.

Looking at the fast-paced penetration of Big Data, powerful tools and techniques is surely going to make Predictive Analytics an indispensable component of an organization’s decision-making framework. The accurate forecast by Predictive analytics system will help organizations identify their loopholes and provide a constant support to the above-mentioned building blocks of a Software Engineering Automation process.This in return will improve estimation of time, effort, and cost; increases the quality and relevance of the software that is delivered to the end customers.