While there are many sources for data, it is typically applications that bring the disparate datasets together to leverage their value and provide services. Data is not currency; data is the fuel that makes the data economy run, and applications are the machinery that converts data into something valuable. In turn, applications are also a source of data in their own right, often by interpreting, transforming and integrating data from other sources.
The chief contribution of applications is to take low-level sensor or input data and elevate it by applying logic and context. A health application might combine altitude, heart rate, location, and elevation sensor data to create data related to fitness and exercise trends over time, or to spot disease by identifying patterns within the input data streams. A way-finding application might take the location and acceleration data from multiple cars and drivers and combine it with mapping data to provide traffic routing and travel-time estimates. A browser might take web-site visits, search queries, and device identifiers to identify user interests and deliver personalized advertising. Thus, the data that applications produce tends to have more informational content than the raw datasets used as inputs.
Applications also sometimes pass-through data that is obtained from other sources, often in combination with new data as described above. It is not uncommon for your calendar or contact data to be replicated and housed in multiple applications and used in different ways. This pass-through and duplication function can make it difficult to draw a boundary around original data sources and secondary or tertiary datasets which might contain the same information.