Welcome again to the data to product specialization. Over the next four courses, we're looking forward to giving you a glimpse of what data science is and teaching you simple techniques and Python-based tools to generate products from data. So, let's first understand what we mean by a data product. In this lecture, we'll explain many interpretations of data product. Describe what we focus on as a data product in the data to product specialization. Illustrate a typical process to go from data to a predictive insight towards an end goal. Motivate why Python is a good choice for data product development. Simply speaking a data product is the output of any data science activity. We generate actionable insights from big data using data science. That insight is a product of Data Science or what goes into a data products as a data-driven knowledge. But why do we need it? While the answer to this question varies, the commonality between all the reasons is to make the data help with taking actions or more like to take data-driven actions to facilitate an end goal. The former United States chief data scientist, DJ Patil defines a data product as, a product that facilitates an end goal through the use of data. Great definition as it speaks to the heart of the reason why we would like to generate a data product in the first place, or the purpose of today the product activity to put insights to use in production so we can take actions. In this specialization we focused on models as a specific data product that can further be used to generate other user oriented or back-end system levels products. So, we define data products as system models that help us to understand data in order to gain insights and make predictions. Now, let's look at what transformation data goes through from raw data to a model and what products get generated as a result of the data science process. Let's look at this simplified data science process. First, we need to access data from its source which often comes in a form that requires further processing, we call this raw data. If raw data is served as a product it goes through some engineering steps for storage and access. You might have heard the general statement that 80 percent of data scientist's job is cleaning and managing data to prepare it for modeling and analysis which is the fun part. We call the output of this data preparation, derived data. Programmatic access to derived data through application programming interfaces shortly referred to as APIs is an important thing for modeling. An API is a common data product that others who are building data products from the dataset use. Bear in mind that for the derived data to happen the data scientists need to do some processing of their own, like cleaning the data adding new entities to the data like a logical group of segments in the data and labeling them. Some of these additional entities can be modeling products themselves and can be served through the API's as well. In addition derived data can be visualized through Custom Dashboards, Web based interfaces like maps or websites and many other applications. Finally, derived data can be used for generating machine learning models or other algorithms that make the derived data useful in the context of applications that include operations research, decision support, just like the one in the fire example and artificial intelligence. Over the next couple of courses, we will learn how we can generate models, create visual dashboards using Python and serve models in the context of web frameworks. In this lecture, we introduced the concept of data products, and shortly explain how the data and algorithms that come out of the data science process can be deployed as data products within. Next, we will review some examples of enterprise-level data products, but before that we've provided you a number of useful links to read about data products. They are do's and don'ts as well as the role of data science in generating data products. I suggest going through these readings before the examples, so the examples will be a little bit more useful.
