| | April 20179CIOReview"distributed data preparation" or "self-service data preparation," which is typically done by business analysts or data scientists.Gartner predicts that by 2020, self-service data preparation tools will be used in more than 50 percent new data integration efforts for analytics. Both traditional and distributed are necessary, but their mix is highly dependent on the nature of analytic needs and projects. While there are many factors, it mostly depends on the amount of time required to access and cure data, in addition to the data types, the sources and targets, and whether or not a data model is required and desired.Today people seek to visually explore and discover facts and insights from their data as they go along mashing up various sources, analyzing them on-the-fly with data visualization tools and techniques, "because I'll know it when I see it."But visual data discovery isn't the only case for a combined data integration / data preparation mix. The need is arguably greater for more advanced prescriptive or predictive analysis cases, and when the necessary human judgment over ML recommendations is applied.Biggest Obstacles to Access DataMost obstacles toward a data-driven customer experience are related to access to data and timeliness of its analysis­which in part is a reason for the proliferation of AI/ML technologies. With the rapid increase in computational capacity and an abundance of data coming from so many new sources including smart sensors and robotics, the need for speed is obviously greater.But there are two components to this. First, being able to access all relevant sources to support faster, more agile business models is vital. This is the case for a mix of traditional data integration and distributed data preparation that we discussed earlier. It's not just about combining all types of data sources, but allowing people to prepare it in ways that can meet their analysis timeframes.The second part is about having a more efficient way to visually analyze, understand, and interpret those data. Today this is not anymore just about doing it quickly, but having the ability to shift focus of analyses­and the respective mashed data sources­at any point, depending on changing market conditions, or new decisions /strategies stemming from the discoveries of the analysis. Skills to Implement Adequate Strategies Technology leaders today are required to build a diverse set of skills in their organizations that aren't just based on technical experience or business acumen, but related to strategy and data management, as we quickly move into a world of analytics-as-a-service.From a data perspective, successful IT leaders differentiate by treating data as a valuable corporate asset. This implies understanding the business and financial aspects of data-as-an-asset, as well as being able to formulate and implement a corporate data strategy.Alongside, data preparation, data wrangling, and data modeling are necessary, particularly for assessing the mix of traditional data integration and distributed data preparation within a corporate data strategy.From an analytics standpoint, we have the infrastructure and deployment side, where Technology leaders need the skills to implement adequate hybrid strategies for their analytics platforms. This implies understanding how data repositories/sources, analytics platforms, or a combination of both can remain on-premises or be deployed on the cloud, and how will this evolve in time­essential guidance for the organization's journey to the cloud.The other aspect of analytics relates to building and fostering an analytical culture in the organization. The required skills include analytic process design, data visualization, predictive analytics, scenario modeling and what-if analysis, and perhaps one of the key elements of an analytical culture, storytelling with data.Future Technology InnovationAnalytics is evolving, gradually incorporating AI / ML technologies and algorithms into its fabric, resulting in what we call "adaptive intelligence." This complements analytics strategies and can support data-driven customer experience. This is of interest to me because adaptive intelligence is at the intersection of people judgment and machine automation. While machines can ingest more data in one second than what people can in ten years without forgetting it and without fatigue; and automation greatly simplifies repetitive computational deductive or inductive processes, we cannot replace human reasoning. The ability to understand and adjust analytic model inputs and training data, improve data imperfections, and apply ethics to our use and interpretation of data are a few examples of what machines can't completely replace. It's no wonder why analyst firm IDC states that by 2020, organizations that analyze all relevant data and deliver actionable information will achieve an extra $430 billion in productivity gainsJosé Villacís
< Page 8 | Page 10 >