Machine Learning - eBook (EN)

BUILDING A MODERN STRATEGY FOR ANALYTICS AND MACHINE LEARNING SUCCESS Inhibitors to success Monetizing data offers a potentially significant payback in business value. However, many organizations have struggled to harness their collective data to get the accurate and meaningful results needed to achieve their goals. Several issues stand in their way: Analytics at AWS. "Their data warehouse might not be able to meet the growth and scale of data coming in. So they struggle with scale, performance, and the operational costs of maintaining it in their own data centers." Data silos. Corporate data tends to be walled off along departmental lines, accessible only to select groups of users. Data is also stored in multiple places, including your data warehouses, data lakes, and data- bases. Analytics projects relying on data that's incom- plete because of these limitations often fail to produce the insights and the returns their designers anticipate. To gain new value from their data, organizations need the ability to break down silos so they can combine and analyze all relevant data regardless of where it lives. "Without sharing, there can be redundancy of data, redun- dancy of cost, and an inability to see the full picture," says Gabriel. "This creates a disadvantage to the company as well as to the individual department." Data gravity. As data continues to grow, it becomes harder to move around. This data "gravity" limits the comprehensiveness of analytics run against a particular domain and can degrade reliability of the insights and decision-making based on them. "To make decisions with speed and agility, customers need to acknowledge data gravity by easily moving the data they need between data stores in a secure and governed way," says Pathak. 3 Data growth. The amount of data created in the world continues to escalate: IDC predicts the amount of digital data created over the next five years—rising from 64 zettabytes in 2020 to 180 zettabytes in 2025—will be greater than twice the amount of all the data that's been created since the advent of digital storage. Rapidly rising volumes and types of data are increasingly difficult to manage using older on-prem- ises infrastructure and manual processes. "Few companies truly understand all the data that they have—whether in the cloud or stored locally," says Michael Gabriel, a Partner at Fortium Partners who has held global CIO positions at HBO and the National Basketball Association. "So knowing the data you have available or could have available, and how it needs to be utilized to support analytics, is problematic." Outdated data infrastructure. Three-tier, on-premises data infrastructure lacks the scale and performance capabilities needed to manage growing data—and requires constant configuration, manage- ment, and capacity planning. Aging infrastructure also makes it difficult to support ML and other advanced analytics capabilities that are critical to delivering action- able insights across the organization. Organizations need data systems that can scale as needs change and data volumes grow. "We see a lot of customers who have an on-premises analytics system and have hit the limits of what they can do," says Rahul Pathak, Vice President for

• Cover