Enterprise Business Intelligence (BI)

Customer Opportunities & Challenges

Large distributed enterprises, and increasingly small and medium enterprise (SMEs), are looking to drive performance through data analytics to better understand customer needs, create better products and services, and/or improve efficiencies. Some applications of predictive analytics that are of interest across many different vertical market segments include targeted marketing campaigns, behavior based advertising, customer and employee churn risk estimation, predicting product wear or failure, and detection of threats, fraud, or malfeasance to business systems and data. Companies often have only a fragmented view of their customers, spread across internal and external silos. Increasingly, customers are looking to push data to the cloud, where it can be pooled into a “data lake”, integrated for modeling, and then processed with cloud-based analytics services with scaled-up processing and take advantage of more sources of data, and get analysis products faster and ultimately in real-time, even when computing on very large datasets.  They are also eager to take advantage of the latest methods in artificial intelligence, machine learning, and high performance computing at scale. This is all about gaining superior insights with more speed and agility, and being incisive in taking advantage of tight windows of business opportunity when they are presented.  

However, there are numerous potential hurdles that enterprise customers can face in achieving their goals:

• Struggling with the complexity of their own systems, which may reflect a distributed collection of incompatible hardware and software applications and data formats, that are continually evolving through restructuring and M&A, which greatly adds to the project costs and makes it difficult to predict analytics success prior to the spending outlay
• Legacy systems, regulations, security, or proprietary considerations constrain part of the data to reside on premises, while other data is moved to the cloud, and the analytics could usefully leverage all the data together
• Data integration and cloud-based analytics project costs can run so high that analytics need to deliver unexpectedly and possibly unrealistically high performance to meet ROI
• Analytics workflow cycle times can be longer than required due to the time it takes to pull, aggregate, and clean the necessary data, and then optimize high dimensional centralized models

Our Value Proposition for Enterprise BI

Because Collaborative Analytics (CA) is based on a natively distributed analysis framework, it is agnostic to where data resides, and also to what methods are used to analyze it locally, which gives it flexibility and agility.  It provides a solution fabric that rides logically on top of the physical infrastructure and native software applications, abstracting away those details. For these reasons, CA can help make analytics pay in the enterprise with the following benefits:

• Take advantage of data wherever it resides, including in partial states of migration between private premises and cloud; support the whole range of hybrid public-private cloud implementations
• Provide significant reductions in cost and latency (time-to-analytics) by avoiding integration of information from disparate systems and in different native formats into massive data lakes
• Maintain flexibility to move data and migrate systems at will and the analytics will “ride along” and continue to operate; CA provides an integrated predictive analytics capability without requiring interoperability of systems, and that does not require redesign as the systems underneath are migrated or applications upgraded or replaced
• Accelerate the use of information residing in new systems that are acquired through M&A activity, prior to the expense and time of integrating systems, and help to prioritize what information would be most useful to integrate first
• Continue to extract value from legacy systems, even as those systems are being phased out of primary business operations
• Enable leverage of external sources of data without having to mix that data with existing data
• Facilitate experiments with business questions and the availability of support in the data to quickly determine project viability and project attainable levels of performance to ensure efforts will lead to actionable insights
• Validate other models that are combining multiple sources of data
• Quantify the value of information; prioritize efforts around sources that provide the most value, only use sources as long as they are valuable and then disconnect from them
• Reduce the analytics cycle time required for model/evaluate/deploy/validate to accelerate time-to-analysis results
• Take advantage of the latest methods in machine learning and artificial intelligence to mine local data stores optimally and independently of one another
• Empower different functional groups (e.g., sales, service, support, finance, operations) to query available data from its perspective concurrently, and also tap the data from other parts of the organization without modifying it or moving it, thereby facilitating cross-functional collaborations
• Keep talent and expertise requirements manageable by assigning personnel to become expert in local sources without requiring them to understand everything about all the data across the silos
• Reduce costs of data security, redundancy, and governance since data is never centralized; eliminate the recurring costs of data management and governance by enabling use of data without “onboarding” it

We are in a period of significant innovation in methods for machine learning and artificial intelligence, substantially based on leveraging large-scale computing. CA affords businesses the latitude to explore innovations in analytics, in terms of information sources and methods, while not forcing entirely new infrastructure investments.  Investigation can be done with managed cost and complexity. The core reasons for this derive from the distributed design (loose coupling) and the level of abstraction provided. For example, customer image repositories could be mined with deep neural networks, while sensor data repositories could be mined with support vector machines (SVM), and text data could be mined with random forests, and CA would compose those local analytics into a globally optimized and unified system.

CA provides superior business intelligence by using more sources of data, at lower total cost of ownership and with less latency, through composed analytic models.  It allows businesses to future-scale and positions them to make the analytics pay.