Getting data management right is becoming increasingly critical. Those that position their organizations to manage data correctly and understand its inherent value will have the advantage.
A recent Forbes Insights report, “The Data Differentiator: How Improving Data Quality Improves Business,” sponsored by Pitney Bowes, examines the key role of data quality. A previous blog summed up the benefits of good-quality data and the costs of poor-quality data. This one provides advice on choosing a data partner.
Any data partner brought on board needs to have a clear and accountable plan for working with an organization to improve data quality. Some other considerations in choosing a data partner:
Put in your own auditing practice. Businesses really need to have their own auditing practice in place for evaluating external sources of data from vendors. It might be fine for the vendors to maintain the databases and platforms for simple email or mailing campaigns, but for a really holistic view of the customer, that data needs to be brought in-house. All of the matching logic, all of the hierarchies, enrichment rules, all of the business logic and clean metadata that the service provider is doing on your behalf needs to get translated back into your organization.
Reputation. Reputation and longevity aren’t everything when it comes to selecting a data partner, but they are important to get a sense of in the decision-making process.
Metadata. What does the vendor provide in terms of metadata about its data set? The latency of its data, how often it is maintained, the number of records it contains—and, ideally, there’s the opportunity to evaluate it in a proof of concept phase before writing any checks. The existence and availability of metadata is, in itself, a good sign.
Supplier commitment over time. Data is in a state of constant flux, meaning that what was true and accurate in the past—whether six months ago or six minutes ago—may not still be true in the present. Even beyond the currency of the data, the world around us changes, and so does the industry. Consequently, it’s important to partner where there’s a commitment to maintaining data and bringing new innovations and products to customers over time, and not form a partnership that ends when the main implementation is complete.
Source authority. What is the authenticity of the data that is required and the reliability of the provider? Not all data sets are created equal. Take satellite data, for example—with a multitude of satellites in operation in Earth’s orbit, many owned and operated by different governments, research facilities and corporations, any given satellite could be considered to be an authoritative source. Some applications of satellite data, however, may require specific spatial and spectral resolution. Existing satellites in orbit may not be necessarily fit for each purpose or application. Therefore, more flexible platforms, like unmanned aerial vehicles that are more ubiquitous, are being considered but may not be operated by an authoritative source.
License structure and wind-down clauses. When bringing in external data sets, businesses should understand the commercial terms of use and termination—and in particular, what rights remain to use work derived from that data after the end of a contract.
Total cost of ownership and time to value. Total cost of ownership is about more than the price of purchase. Buyers should also factor in the cost of keeping software maintained and updated—and how well it supports agile, iterative methodologies that speed time to deployment. Time to value depends strongly on the particulars of a solution and an organization’s needs, but agile solutions typically see at least some partial deployment in a working environment within three to six months, versus the typical 18 months to three years for a more traditional waterfall implementation method.
Customizability. Some software comes shipped with preset schemas, which may or may not fit an organization’s purposes over time. Some vendors have a “black box” approach that allows little, if any, modification, so the ability to customize schemas should be ascertained up front. Canned schemas almost always wind up having to be modified extensively—not necessarily the easiest thing to do.
