Self-Service Data Analytics and Governance for Managers. Nathan E. Myers

Чтение книги онлайн.

СКАЧАТЬ audiences, and a portion of them may find their way to external stakeholders and regulators. A whole book could be written on how to make a thoughtful selection of KPIs for a given process, in order to convey health across a number of dimensions. However, for purposes of this book, we will assume that these have been arrived at separately and are effectively conveying business performance to allow for active and rigorous management. What we do want to cover in this section is the ways that KPIs can be compiled and displayed efficiently through the use of dashboards and visualization tools.

      Most, if not all of our readers will be familiar with common temporal data visualization formats – bar charts (to show value comparisons), line charts (to show time-series movements), scatterplots (to show large numbers of observations), and sparklines (for trending). Some may be familiar with hierarchical visualizations like tree diagrams, sunbursts, and ring charts. A more select few will be familiar with multidimensional data visualizations that can communicate more than one variable for each observation. Examples of multidimensional data visualizations include pie charts and stacked bar charts that show observation values relative to the whole, and Venn diagrams that can show observations that meet either one or both of two population definitions or constraints. Visualizations are widely in use to make both interpretation and comparison of KPI observations as easy as possible to understand at a glance, or at least in a handshake, rather than after a prolonged study.

Dashboards build on data visualization by pulling together all critical KPIs that are necessary to manage an enterprise, a business, a department, or even a process, in a single view. Measuring and displaying multiple business indicators together can offer context that communicating a single fact or value alone can fail to do. They are built to deliver the key and necessary performance metrics in a visually rich frame that can provide stakeholders with an instant comprehension and more complete understanding of the relative health of business processes.

      Anyone who has ever owned the production and distribution of a metrics dashboard or scorecard would likely tell you that they are surprisingly thought- and labor-intensive to produce and maintain. From gathering and agreeing the KPIs that best convey the full picture, to the design and layout of each individual component metric in a visualization, to structuring the page layout to feature the most key of the key metrics prominently – all of these design steps represent a lot of work. When there are multiple recipients, it is always a challenge to navigate the conflicting preferences of each, and we all know that recipients are never bashful about suggesting additions or format changes. We have all been to meetings that have been sidetracked completely by the one audience member who spends the bulk of the session asking questions surrounding the format of the visualizations or the array of components and their order on the page, rather than engaging in a productive discussion on how to improve any of the key metrics. Beyond the design, perhaps even more time-consuming are the maintenance steps that are required each time the metrics dashboard is to be communicated. The slides must be dusted off and refreshed with the updates that have occurred across any of the dimensions from any of the various data sources. From there, date headings must be refreshed, any changes that have been requested must be made, and of course commentary must be updated, before it is sent off.

      Over the last decade, a number of dashboarding and visualization vendor tools have emerged to simplify dashboard design, to enable the efficient capture and assembly of KPIs, and importantly to allow for low-latency refresh of visualizations on demand. Key among them are Tableau, QlikView and Qlik Sense, SAP Business Objects, IBM Cognos, Microsoft Power BI, and Oracle BI. This is an ever-moving list, but these are names readers should recognize, as they represent prevalent and widely subscribed visualization platforms – and they are increasingly tied to business intelligence and data analytics.

The evolution from flat reports on continuous form paper that had the perforated strips with holes on each side, if anyone remembers folding them over, licking the edges, and tearing them off (sorry team, yes one of your authors licked the edges to get a cleaner tear, while your second author claims not to have and cringes in disgust— you'll never know which was whom), to brighter reports featuring better fonts and some color graphs and visuals that we got accustomed to in the 1990s and 2000s, to the highly flexible, visual, dynamic, and interactive digital dashboards we have today that convey business intelligence insights is startling – and game-changing. We cannot introduce significant emerging and enabling data analytics technology without making mention of the advancements in data visualization and dashboarding that puts key information in the hands of end-users and decision-makers in an intelligent, versatile, and insightful way.

      Discussion with Paul Paris – CEO, Lash Affair

       “How did you first learn of the AI components and technologies that are Social Listening?”

      I first learned about Social Listening AI technology during a lecture given by a PA-based firm called Monetate back in the Spring of 2015. At the time of the lecture we were still essentially a start-up and we were more focused on foundational steps to build our company. However, AI captivated me from that moment onward, and I began to watch developments in the space much closer. The potential to get inside of our customers’ heads to improve our products and services sounded like a game-changer. As our company grew to a stage where we were ready, we immediately plugged in. Frankly, we knew that to be a trend-setting and best-in-class company, we needed to be on the forefront with cutting edge technology like AI Social Listening. How could we not?

       “How have you employed artificial intelligence and applied data science for Social Listening?”

At Lash Affair, anything we can do to isolate and uncover consumer trends and brand sentiment will give us an edge. Our company is growing quickly, and more than ever before, we need our hands directly on the pulse of our customers. We want to be the very best in our industry and want to set the bar very high when it comes to exceeding customer expectations. How do we understand those expectations? Well, of course, we don't have time to devote to personally trolling social media sites for posts related to our industry, products and services, and our brand in particular (although we admit to trying every day). However, we have armed ourselves with a serious data analytics dashboard, which gives us the reach to be able to scour the enormous social media landscape for relevant data points that can help us to understand how we are doing with our customers – and with influencers.

       “What specific AI components are employed for Social Listening?”

      Web-data capture is instrumental to helping us to cast a wide net based on search criteria we use to define relevance. This technology is important to allowing us to capture an enormous number of target observations. From there, we work with machine learning analysts and consumer intelligence experts to extract and understand the tone of posts. With enough timely observations, we can interpret and even get ahead of sentiment about both our brand and trends in the industry. Once we have relevant data, NLP technology is enlisted to translate the informal vernacular of social media participants in posts. After all, few bloggers use straightforward and easy-to-interpret affirmative statements like, “My brand sentiment for Lash Affair products and services is extremely positive – on the very far right of the brand sentiment continuum.” Instead, we need to be able to analyze a vast number of sentiment observations and classify each as “positive” (+1) or “negative” (–1), or somewhere in between along the spectrum. Imagine an algorithm that assigns numerical values to the series of observed adjectives being used to describe excitement about our brand, recent customer experiences, and customer loyalty to help us gauge prevailing consumer sentiment. By charting sentiment observation values in time-series, we can spot trends. To get at it involves feeding tons of training data through the machine learning classification algorithm, so that it begins to interpret observations as predictably and reliably as if I personally was in the chair, reading each post, and graphing each sentiment observation as it comes in.

       “What do you do with the summarized customer sentiment information?” СКАЧАТЬ