In this special guest feature, David Winikoff, Senior Director, SteelCentral Product Management at Riverbed Technology, believes that as AI and Machine Learning go from hype to reality, organizations must be cautious in adopting too quickly, as one size does not fit all. David has over two decades of success leading product management teams and as an engineer (both software and hardware) creating offerings for enterprises. Among the innovations he has helped developed: high performance storage subsystems, multi-modal unified communications, QoS-based WAN optimization, and sophisticated network performance management tools. His current role is leading Riverbed’s digital performance management plans for advanced analytics. A key focus of this effort is to combine Riverbed’s industry leading depth and breadth of performance data into a common data source for machine learning and artificial intelligence techniques. In his free time, David is an instructor for the University of California Graduate School of Business, teaching courses related to innovation, entrepreneurship, marketing and product management.
Everywhere you turn, businesses are talking about the benefits of ML and AI – Machine Learning and Artificial Intelligence – technologies that are helping organizations to get more out of their data through deep-dive analysis. The lure of adoption may be the promise of enhanced efficiency and greater productivity but the real added value comes in the form of unexpected data results that become insightful business intelligence through clever analysis.
The truth is that there are no guarantees of what sort of results a deep-dive into the data will provide. While ML and AI are smart technologies with vast capabilities, they can only process the data they’re provided. While many organizations might find value in an analysis of the various datasets within their organization, the insights that provide a competitive edge or shed light into customer habits or industry trends will require an approach to data analysis that goes deeper than what can found within any single dataset.
Beyond “What” to Learn “Why”
To understand the power behind ML and AI technologies, consider the example of a traffic intersection with an unusually high number of collisions. Through a simple analysis of a single dataset – the locations of accidents within a city – officials can identify a problem that otherwise might have gone unrecognized. But to find a solution, traffic engineers must examine other datasets to explore possible causes, including data such as traffic volumes and patterns, lane configurations or signal timings.
ML technologies can help identify the issue, or the “what,” through real-time monitoring of the data – city traffic accidents, in this example. But the real power of the technology kicks in during the pursuit of the “why.” By injecting layers and layers of peripheral data, AI/ML technologies can identify – in real-time – potential solutions that otherwise might have been overlooked.
Big Data Provides Granular Insight
The age of Big Data is one where machines can collect, organize and analyze datasets at a speed and scale that humans simply cannot achieve. This opens the door to greater insights by considering peripheral data that might not previously been examined yet still has relevance.
Consider the traffic example. An injection of data perceived to be unrelated could provide even greater insight about what’s really causing those accidents. Weather data would allow engineers to gauge whether wet roads or poor visibility played a role in the accidents. Data about road construction would reveal if lane closures or detours were contributing factors. Even events, such as ballgames or concerts at a nearby stadium, might come into play. Another data layer could include looking more closely at data related to stadium events, such as event type, alcohol sales and timing.
By adding more data variables to dig deeper into the root cause, the solution may be as simple as an officer directing traffic after an event at the stadium, which is a lower-cost and less-disruptive solution than changing lane configurations, decreasing the speed limit or altering the timing of traffic signals.
Just as more information could make the roads safer and more efficient, businesses also could benefit from new insights on what’s really impacting productivity, sales or customer support. If a network slowdown is keeping mobile customers from placing transactions, the company needs a robust mechanism to both quickly identify the slowdown and explore peripheral data to determine the cause, such as the types of devices impacted, affected versions of the app and regions where the issues are most prevalent.
The Layers of Discovery
The key to successful implementations of ML/AI solutions is recognizing that data discovery occurs in two parts – identifying the problem, then proposing the solution.
A deep data analysis of data can recognize patterns and anomalies that provide predictive insights about possible issues – before they escalate into bigger problems for customers, partners or employees. A simple data analysis may show when a company’s KPIs are less than optimal but without the peripheral data – which have impactful correlations or are predictive indicators – the reasons behind the suboptimal KPIs will continue to be best guesses.
ML/AI allows the analysts to get past the “what” and focus on “why” so that it becomes easier to implement solutions that actually address the causes.Different teams within an organization have different objectives, but it’s important to remember that they’re all interconnected – and analysis based on a variety of data, as well as business objectives, will reveal the relationships between those teams and the greater business success.
ML Needs a Mission
Technologies based on ML and AI can provide valuable insight and understanding – but there is no out-of-the-box solution to help organizations find that nirvana of complete digital understanding. Too often companies that look at ML solutions are exploring the capabilities of the technology, but are not necessarily realizing the benefits. Pursuing the “why” and putting action items into place based on data analysis is more effective for business success than just plugging in ML technologies to see what you get.
AI and ML are robust and capable of handling heavy workloads. With the right strategy for putting those technologies to work, organizations will quickly find that adding more data will power the technologies to provide even greater insight.
Sign up for the free insideBIGDATA newsletter.
Speak Your Mind