Digital Consulting Services
Navigating the vast array of digital technology solutions available can be a difficult task. In many cases, the capabilities promised by the analytics are not delivered by the underlying data quality. We’re all familiar with the phrase “garbage in … garbage out”, and nowhere is it more applicable than in “big data” analytics solutions.
The most expensive projects are the ones that fail, and the primary reason for failure in digital data technology projects is a mismatch between the data quality and the analytic solution objectives.
Athens Group’s Digital Consulting Group guides you through three critical activities so that you ensure your data quality, and analytic solution objectives are aligned.
Establishing the Data Analytics Objectives
- What are you trying to accomplish – what business, operational, or safety objective are you trying to meet?
- What digital technology is available to accomplish the objective?
Establishing the Data Quality
- Can you collect the data necessary to support that objective?
- Does the data represent a stable and well characterized process?
Establishing the Infrastructure and Tools
- What infrastructure do you need to transmit and store the data so it’s available to the right people and analytic tools at the right time?
- What level of cybersecurity is necessary to protect the operation and the data?
- What tools do I need to filter and analyze the data?
The objective and the data quality are intimately tied. The data quality will limit the objectives you can achieve and the defined objective will establish the data quality requirements.
The Data Analytics Objectives are defined by three characteristics: Task, Method, and Proximity:
The Task defines what you are trying to accomplish with the data analytics. This can include:
- Passive monitoring of a process
- Active control of a process
- Decision support for the management and operation of a process
The Method defines how you are going to accomplish the task. This can include:
- Mechanization – where a mechanical device replaces human muscle previously used to accomplish the task
- Automation – where a controller with a pre-programmed deterministic sequence, monitored by a human operator, executes the task
- Autonomous Operations – where a controller with a pre-programmed deterministic sequence executes the task without the need for human intervention
- Artificial Intelligence – where an intelligent control program can make non-deterministic decisions about how a task should be accomplished without the need for any human intervention
The Proximity defines where the task will be accomplished. This can include:
- Adjacent to the process, with physical contact
- Removed from the process, nearby, in the same physical location, perhaps within easy transit or within sight of the process
- Remote from the process, onshore or in a centralized data management facility
Once the three primary characteristics are defined, the data quality necessary to support those objectives can be defined. Data quality has two primary characteristics:
(1) The Data Metrics address the physical characteristics of the data itself. This can include:
- Content, which ensures the data has sufficient bandwidth and contains enough information about the process
- Accuracy, which ensures the data represent the actual condition
- Precision, which ensures the data is repeatable across multiple processes
- Consistency, which determines if the data is repeatable for a single process
- Stability, which ensures the data represents a statistically valid process
- Traceability, which ensures that the data can be identified in both time and space
- Timeliness, which ensures the data is available when it is needed to support the task
(2) The Data Level defines the filtering and conditioning that is necessary to make the data useful at different points in the analytics. This can include:
- Raw data, which is data taken directly from the source sensor
- Conditioned data, which is data filtered to account for any sensor based characteristics
- Processed data, which is data filtered to accommodate the specific requirements of the analytics