“…interviews with 325 IT and data decision-makers across Financial Services in Europe. The presence of a data decision gap is often due to inaccurate and incomplete datasets, which ultimately impact a businesses’ bottom line”
After all the investment in data management infrastructure, firms are still unable to use the data to make better decisions. A reason for this is the accuracy of the data. Data accuracy is binary, it’s either right and can be relied on to make decisions or it’s wrong and you can’t.
Only recently has there been the realization at the enterprise level that data needs to be treated as a “first-class citizen” that should be managed and cared for.
Quality data is at the core of any successful application development project, so here’s a software engineering perspective on what it takes to produce accurate data:
Data needs to be everybody’s job and there needs to be a clear process around curation, how the data will be used and by whom. There needs to be policies and procedures in place that employees must adhere to. The performance review process needs to be amended to include measures around data citizenship.
The same thing needs to be called the same thing. The different entities and actions in your organization need to be modelled and represented in a single data model. It’s unrealistic to convert all the legacy data siloes and so a clear way to join the data between old and new is required.
Bad data is worse than no data, so a reliable process to detect if the data is either incomplete or incorrect is paramount. This is a thorny issue; how can you guarantee that data from new systems are integrated? How can you guarantee that data isn’t lost during outages system recoveries? Making sure your data is depended on as part of normal daily operations is a useful trick.
Ultimately there needs to be a leader within the organization, who is senior enough to make cross divisional change happen and to be able to break ties on whether something should be modelled this way or that.
Link to the original article here.