2018-02-22

SharePoint a small data tool

SharePoint is created on top of SQL Server and integrated with other Microsoft applications such as Office and cloud services. This definition places SharePoint among small data applications since SQL only works on structured data. The reasoning behind it was to create a product that was already BI-ready, without having dedicated in-house experts.

However, in the last decade, Big Data has risen as a central point in IT development, and unstructured or semi-structured data carries more valuable information for companies than neatly tabulated records since it can power machine learning algorithms. Can SharePoint, as a professional content management system, rise to the challenge of creating a welcoming environment for Big Data or it will become obsolete due to this shift in data magnitude and organization?

The role of machine learning

First, we need to understand how can machine learning and data science outsourcing be used in a corporate environment and what are the problems it could help solve. It uses data to forecast trends dynamically. Compared to regular statistical tools, machine learning looks at vast amounts of data and aims to uncover patterns. By extrapolating these behaviors, new data points are created.

Problems to be solved

Brandon Rohrer described the five questions machine learning can help answer and which in fact, represent the five basic ML algorithms.

The first one is classification into two or more groups. In an organization, you might have tasks or groups of employees that you need to divide into similar groups.

Anomaly detection is concerned with identifying those data points that stand out from the crowd, the so-called outliers. This approach could come in handy during results or performance evaluation, especially if the amount of data from which a conclusion needs to be extracted is overwhelming.

Regression analysis aims to offer a numerical answer to a question concerned with size. Sales prognosis, revenue, and costs can all be successfully predicted with this algorithm.

Identifying the way data is organized represents one of the core powers for machine learning algorithms. Clustering is all about defining a center and deeming all the elements within a certain distance from that pole as part of the same cluster. This is a way to measure similarity, without being identical, defining families of items. It is useful to create client target groups and constitutes the base for recommendation engines.

Lastly, reinforcement learning algorithms are those that learn from past behavior, either supervised or unsupervised. These are well-fitted for smart systems that need little human input, like those controlling sensors. This could be useful for an organization relying on automation.

From descriptive to predictive

To remain competitive, organizations are moving from a reactive to a proactive approach. This translates into the tools they use. While the 90s and early 2000s focused on reports which used historical data, the next decade was all about BI, which offered real-time diagnostic analysis. In this context, SharePoint became the staple mark of collaborative platforms which also provided dashboard features.

Machine learning is supposed to take this a step further and show glimpses of the future based on existing data. Integrating the previously discussed algorithms with SharePoint offers companies a central repository of past experiences and predictions. These will act as a roadmap for employees who can now evaluate at each moment how their work is helping them achieve the proposed goals.

SharePoint applications of ML and Big Data

From the information presented above, it follows that the marriage between SharePoint and Big Data could be beneficial to the organization, but it has yet to be used to its full potential. This is due to some inbuilt limitations, to which solutions have become available. 

Limitations and solutions

The main weakness is related to the file object size of 2 GB restriction. On its own, this is incompatible with Big Data, but through the Open Data Protocol which helps to get information in and out external workbooks, this inconvenience can be overridden.

Another way to work around the file size limits is to use an analytical workbook which is only restricted by the server’s capacity and process the data externally. Then, find ways to bring the results in files that comply with the 2 GB requirement.   

It might not be a full-scale use of Big Data, but it could be enough to give an organization the answers it is looking for regarding the relationships between teams, results, and documents.

Text analytics

One of the best applications of Big Data for SharePoint is related to text analysis. Since the platform holds numerous documents, it would be helpful to classify these, tag them accordingly and create management systems that help fast and accurate retrieval of information.

Through the Text Analytics API from the Azure ML, the algorithm can take any piece of text and perform either sentiment analysis, keyword extraction or even tone evaluation.

For an organization, such tools have a plethora of applications, including HR application evaluations, social media monitorization, internal documents classification and more.

Customer journey

By importing data about the customers at each stage of their relationship, the company can evaluate the impact of their actions including marketing efforts, pricing decisions, and distribution channels. The algorithms allow companies to create client personas, regression gives sales estimates while anomaly detection could point out either excellent opportunities or dangerous pitfalls. A unified view can evaluate the value of the customers through each phase of the lifecycle. We can expect an end-to-end approach to become the norm.

The reign of data

Data has become a currency in the business world, and the tools companies choose should be able to make the most out of it. Since SharePoint is the most likely choice of an organization for BI and content management, it should be enhanced with the capability to process this kind of information and turn it into actionable insights, ready to be consulted at a glance. Machine learning algorithms can act as the data processors, while SharePoint is an excellent repository and dashboard environment.

About the author 

Maria W