The Age of Machine Learning
We are increasingly surrounded by Machine Learning solutions that assist us with making decisions on everyday tasks that were done the old-fashioned way not long ago. From "movie recommendations" on Netflix, to the "connected home" via Amazon-Alexa, to "driver-assist technologies" within today's Smart Vehicles, decision-making in our everyday lives has taken a giant leap forward in recent years.
Despite these advances, it is often incorrectly perceived that Machine Learning is not suited to the Small & Medium-Sized Businesses (SMB) sector for a variety of reasons. We dispel SMBs' perceived barriers to adopting Machine Learning (ML) and instead, demonstrate the contrary – that the SMB sector is in a prime position to embrace ML and capitalize on its benefits.
What is Machine Learning?
Before we get to establishing why SMBs should be actively considering ML, it is beneficial to set a foundation by first defining ML. An often-quoted definition, attributed to Arthur Samuel defines ML as "the field of study that gives computers the ability to learn without being explicitly programmed." Samuel's definition dates to 1959 and does not factor in the technological advances that have enabled the paradigm shift that makes ML prevalent in today's world.
We will elaborate further on this point in a future blog. For now, it suffices to say that "learning" begins with a business process or a pain point (i.e. a use case). Practitioners isolate the use case's "inputs" and the "outcomes" generated from different values taken by the "inputs." Learning algorithms then seek to predict the outcome given a set of values against the inputs, after first identifying the subset of inputs (called "features") that will deliver an accurate prediction. Statistical Learning and Machine Learning are the two broadest categories of learning methods. ML's rise to prominence is attributed to the collective influence of several factors: significant advances in computing power, elasticity offered by cloud platforms and the explosion of structured and unstructured data assets (both internal assets and external assets accessible via open APIs).
Perceived Barriers to Machine Learning in the SMB sector
1. Data Volumes:
The argument here is that SMBs typically deal with small data volumes against most business processes, suggesting that ML may be an overreach. While this may be true of traditional transactional assets, they do not account for the rapidly expanding digital interaction assets (web, mobile, social media, spatial, etc.) or the wealth of open-access external assets (macro-economic indicators, market or sector open data, news feeds, regulatory compliance, etc.) that may be relevant to a business process. Data width is as important as data depth when assessing data volumes. SMBs looking to gain a competitive advantage over their peers are taking steps towards enriching their transactional assets with internal digital assets and available open-access external assets. ML employed against these collective assets will deliver business performance driving insights for these SMBs.
2. Hardware and Storage Requirements:
SMBs already subscribe to cloud-based business applications like Office365, SalesForce.com, Concur and Namely. Yet they do not draw the connection to Cloud services when challenged by hardware and storage requirements to house relevant data assets for their business. Aside from the elasticity that Cloud provides, Cloud also allows businesses to convert what would otherwise be a capital expenditure to an operational expenditure.
3. Develop, Maintain and Innovate Cloud technologies like Microsoft Azure and Amazon Web Services are rapidly expanding to provide infrastructure support for ML. Both platforms include a suite of proven ML development tools together with niche turnkey Machine Learning as-a-Service (MLaaS) offerings that address targeted business processes or use cases. Further, services vendors like WorldLink are quickly filling the ML resource gap faced by SMBs through Managed Services offerings. These include the selection, running and maintenance of proprietary or MLaaS solutions and complementing these with advisory and development services to meet ongoing innovation and business transformation needs.
4. Guided Diagnostics Need: ML algorithms scale well in high-volume, low-latency scenarios requiring use of unstructured data on a small foot-print. On the downside, ML algorithms do not allow for interpretation of the relationship between important features in the data and the predicted outcome. This makes them unsuitable in their native form within scenarios that require guided diagnostics (e.g. Healthcare). However, where ML algorithms falls short, traditional Statistical Learning methods come to the rescue. While Statistical Learning design relies heavily on business Subject Matter Expert (SME) collaboration, it excels at providing guided diagnostics support. Consequently, hybrid (machine and statistical) learning approaches are attractive options for SMBs who are typically looking to employ ML as a productivity aid to assist their SMEs. In turn, the SMEs facilitate the "learning" within these algorithms through their extensive business expertise and knowledge base.
Explore what Machine Learning can do for your SMB
Follow us through our series of blogs on Machine Learning to learn more about how ML can help your SMB exceed its objectives. WorldLink offers a suite of advisory, discovery, development and managed services together with knowledgeable resources who have a proven track record of deploying ML solutions for Fortune 500 business. We combine our Machine Learning prowess with business, technology and cloud acumen to help you realize a best-in-breed ML solution for your chosen business process.
By Brian Monteiro - Director, Analytics Practice at WorldLink