Rules For Developing Analytical Capabilities
Wed, April 14, 2021

Rules For Developing Analytical Capabilities

Businesses must know that there is no “big data” and “enterprise data” / Photo Credit: Wright Studio (via Shutterstock)

 

Ramanujam Rao of business news website Forbes enumerated a few rules for developing analytics capabilities. First, businesses must know that there is no “big data” and “enterprise data.” They tend to split analytics into the traditional data warehouse (DW) enterprise data and predictive-analytics-focused big data platforms. Frankly, value comes from all data, particularly when unstructured data merges with enterprise data such as customer interactions, finance, and more. We are currently deviating from DWs and multiple repositories of data to moving toward aggregated data.

Second, traditional DWs are irrelevant as they require too much structure and are complex to develop and maintain. Besides, traditional DWs don’t help in progressing businesses. Creating new platforms on data warehouse architecture is not a sustainable investment. Instead, businesses should opt for a single, data-lake style repository that can store all enterprise data, including high-volume and high-velocity data. Migrating to a new architecture involves hard work, but the returns are definitely better. Third, AI and ML adoption are evolutionary. However, these technologies can only be adopted once a strong foundation of quality data has been established. That data should be curated, managed, and secured. Powerful solutions like enterprise information chatbots, enterprise cognitive capabilities, and the like can be easily developed once a robust data foundation is created in the enterprise and when the business is confident enough in its business intelligence and reporting capabilities.

Fourth, the primary value of a data lake is analytics, not operational reporting. Most software-as-a-service (SaaS) providers deliver reporting capabilities out of the box. These are the best places to develop operational reporting. A data lake is not a viable location as it’s “built into the cost we already pay to the provider, who will be glad to offer more powerful solutions.” This has nothing to do with the lake’s lack of capabilities. In fact, using a lake for operational reporting should be done in “very targeted unique scenarios that do not duplicate vendor-offered solutions.”

Finally, data mining and science are non-deterministic. Data mining efforts can yield wonderful results or nothing at all. Hence, enterprises need to be comfortable in experimenting with new things and accepting failure.