Analytics is a wide scoped field that is composed of many fields. It includes business intelligence and data science. It includes data mining and machine learning. Analytics is the widest category of using data in a way that transforms businesses. Analytics is a key component if not the main of digital transformation. When going digital, a key component is visualizing the business by identifying key performance indicators. Not only do we showcase the data, but how we can document our data in order to better analyze the data. Then having good concise data will help us using popular algorithms to predict future performance and even further test various success factors.
Analytics Roadmap divided into 4 phases:
Descriptive Phase:
State business metrics
Identify data required
Extract and prepare data
Analyze data
Present data
Diagnostic Phase:
Identify trends, thresholds, and anomalies
Further studies
Drill into the analytics (discovery)
Determine casual relationships
Predictive Phase:
Identify business outcomes
Determine the data required to train
Determine types of analysis
Validate results
Test predictions on performance
Prescriptive Phase:
Build a business case
Define the model's rules
Test, Test, Test
The journey of analytics starts with describing the data. In this stage, we ask the question of what happened. In this phase, we are starting to extract available data and trying to present it in a usable matter. This phase also includes setting up auto-refresh for our data. At this stage, our business has identified the importance of digital representation of our data. The business case identifies key performance indicators and the data analyst, data scientist task it focused or showcasing the data using various tools in close to real-time. Most popular amongst them is Microsoft Power BI Desktop which is free of charge.
The second phase of the journey is diagnostics. Here the business has benefited from real-time monitoring and wants to dig deeper into the causes of these key performance indicators. The data scientist in this phase will focus on connecting data together in order to answer the question of why the data is the way it is. Why do we have a drop in sales, by connecting to production data, such connections provide deep dives? Tools used here are Influencers, and Decomposition tree in Power BI, and anomaly detection models in Azure Machine Learning Studio Classic, both of which are free to the public.
Third and critical to many managers and planners is what will happen next? Business owners want to forecast future performance to plan for it; drop in sales due to seasonality, and the likes. The predictive phase focuses on using the now available data to predict future trends. This is done with the help of business owners. Identifying which variables are dependent on which variables are key to formulating the best models to forecast future sales, breakdowns, and so on. The tool recommended to use here is the Azure Machine Learning Studio Classic, KNIME, or IBM Modeler.
Finally, the fourth phase is known as the prescriptive phase, just like a physician would prescribe medicine to heal, and medicine to prevent; this is the goal of this phase. Tools such as Azure Machine Learning Studio Classic, IBM Modeler, and KNIME come in handy here. Also, simulation and optimization if possible can be done here. Business case has to be transformed to fit popular models. These models are then tested, using test data to showcase how different combinations of factors will make better our status quo. An example of this is market segmentation and marketing promotions offered to different categories of customers.
Note close working with business owners is key to success in all stages. All tools suggested here are free tools except IBM Modeler. Also commonly known that with analytics comes programming skills. Tools such as Azure Machine Learning Studio integrates with excel and allows for use of models easily. Power BI also makes visualization simple task of drag and drop. The need for programming is decreasing and we are moving more and more towards the codeless world. It is common for a data scientist to have knowledge of one or more programming languages. It adds to their tools-kit, however not critical for successful analytics projects.
We hope by the end of this article you have gathered a clearer idea of Analytics Roadmap.
- For questions and comments please email ali.ahmed@symbios-consulting.com
Mobile-WhatsApp: +201022002506
Comments