Dirty Data Practice No. 1
You think buying the coolest business intelligence tool is all you need. 数据挖掘研究院
It may be a truism that your BI reporting tools are only as good as the information you feed them (that is, "garbage in, garbage out"), but that doesn't mean that the right actions are a given. Since most organizations still take an isolated view of data, data governance remains a difficulty, says Ian Charlesworth, principal analyst with IT consultancy Ovum. Data is all too often siloed in different business units and is entered, treated and viewed differently, making "one version of the truth" impossible. 数据挖掘研究院
Clean It Up
Know your data. 数据挖掘研究院
The first step of data governance is to establish a clear view of your data; find out what you have, how reliable the information is, what data is beneficial but previously unused, which data is corrupted and which IT projects are duplicating information. And be sure to communicate to stakeholders the cost of not having data governance and the value of creating it. 数据挖掘研究院
Dirty Data Practice No. 2
You procrastinate until you can do a complete overhaul.
An all-or-nothing approach is almost guaranteed to fail. For starters, bringing all data under control in one fell swoop is not realistic given time and money constraints, and in organizations where such an overhaul is possible, user resistance is almost a given.
Clean It Up
Start small, think big.
Instead of all or nothing, prioritize the most crucial aspects of data governance, in keeping with your overarching vision. For example, Charlesworth recommends focusing on four key areas. 数据挖掘研究院
- Create data quality processes and procedures, and where possible embed these at the point of data creation or capture. For example, create a data validation routine in an order entry system or establish a corporate standard for name and address nomenclature.
- Assign a data steward. This person should be someone from within the business who can champion and enforce data quality practices throughout the business. This person should have an intimate knowledge of how and where the data will be used by the business, and who can act as a liaison between the business and IT.
- Create a master data management solution. For starters, this means assigning unique identifiers to core information assets across the business, such as service codes, customer definitions and so on.
- Integrate metadata. Metadata gives important information to both IT and the business, puts complex information into layman's terms and relays vital information about underlying data syntax, semantic correctness and so on.
Clean It Up
Establish a culture of data governance.
Ongoing training and key milestones that measure data governance's benefits can help keep quality control on users' radar. Successful data governance also depends on dedicated sponsorship from someone in top management. Charlesworth says the CIO is often the perfect person for the job due to a CIO's likely combination of forward-thinking and a focus on efficiencies around process, money and technologies. Some companies even create a C title specifically for the position, such as chief data officer or chief data steward.
Dirty Data Practice No. 4
You let red tape suck the life from your efforts.
Charlesworth says many data governance efforts fail to show positive change, and instead stall in meetings and bureaucracy. But if you don't focus on action and demonstrable wins, users won't feel the positive benefits firsthand, making user commitment unlikely.
Clean It Up
Deliver quick wins.
To get user buy-in and commitment, you must create, demonstrate and internally market the positive changes won through data governance. For example, one measurable benefit to focus on initially could be improving validation of order entries to reduce errors. 数据挖掘研究院
Dirty Data Practice No. 5
You make ROI the be-all and end-all. 数据挖掘研究院
Can you accurately isolate investment benefits and attribute them to a particular project? In today's multifaceted, complex business environment, this is not likely, says Charlesworth. Calculating ROI on a particular investment assumes that everything else in the business either stood still or had no influence on the benefits, he says.
Clean It Up
Create a clear picture of success. 数据挖掘研究院
Charlesworth recommends looking to other metrics such as internal rate of return (a measure of an investment's efficiency or the rate of growth a project is expected to generate) and economic value added (estimate of true economic profit). However, the most important thing isn't the calculations per se, it's the discussion around defining success—what it looks like and how you know when you have it, says Charlesworth. This is especially important in terms of measuring value of data governance at various phases and levels of granularity to make sure you stay on track, and, if not, making corrections. His examples of such metrics include a data quality dashboard that displays the accurateness of data processing, data consistency and reuse of rules/measures, and project-specific metrics such as standardization of product master data elements.
