In-Depth
How to Sell a Data Quality Project
It’s a tough time to cost-justify BI projects and get buy-in from upper management. Here’s how to get your data quality project approved.
By Gian Di Loreto, Ph.D.
Quality. It’s a word in our language that, for me, invokes images of well-made hand tools that won’t give out when you are under the car and bust your knuckles, or of the stitching on the seats of my F-250 Super Duty pickup truck. If feels good to talk about high quality, and it makes your clients feel good when they can be assured of it.
As business became more dependent on data, it seemed natural to begin to talk about data quality. Well, almost natural. The quality of something tangible such as a house or a vehicle is easy to measure and understand. The quality of artistic endeavors such as music or films is harder to quantify because there is no universally agreed-upon scale.
Data poses a similar dilemma. We lack an agreed-upon standard of measurement for data quality, and, in fact, it is often (if not always) the case that the quality of a particular data set depends entirely on its purpose. This lack of standards also makes data quality projects hard to sell to upper management.
I became involved in data quality following a career that started in physics and migrated to software development. Data quality seemed natural to me because I am interested in computers and programming and because I am neat and (some would say obsessively) clean in my personal life. For me, it’s pleasing for things in general to be in order, to be accurate, and to be correct. The data in my life should be no different.
When I must explain the reasons to fund a data quality project to a CFO or other high-level corporate officer, I can’t justify the expense because it’s nice to have things in order. Telling them you like order or that you like your home clean (though with three children, it isn’t always spotless) isn’t going to convince Time Warner to hire you to clean their enterprise data warehouse (I know this from experience).
That wasn’t always the case. When I started working in this business in the late 90s, companies spent money on data quality even though they saw it as a luxury. Those days are now long gone, and data quality practitioners need a carefully constructed and compelling argument for why their bosses, users, or clients should care about the quality of their data.
The Carrot or the Stick
There are two clear paths you can take to sell a data quality initiative: a carrot (money) or a stick (compliance).
The age-old question in our business is “How much money is our poor data quality costing us?” To answer this question, you can simply add up the amount of man hours and associated dollars spent on manual data cleanup efforts currently underway (or which should be underway based on known issues). This is almost certainly an underestimate but it might be sufficient to convince those who control the purse strings that it would be worth attacking the problem with a more robust solution.
From there, make the connection between poor data quality and dollars spent gets trickier. Your approach essentially boils down to estimating the amount of bad data in your warehouse (you are usually safe estimating that between 30 and 50 percent of all database records contain a material error that will cause a downstream problem) and then further estimating the costs associated with those downstream errors.
This is very fuzzy logic, and I’ve been on many a sales call where an astute executive saw it for the guessing game it is. That executive would usually voice a preference for the predictable expense of a manual solution even though it was just addressing the tip of the iceberg. This is hard to argue with, so your best bet is to conduct as much research as you can and solidify your math, shore up your calculations, and try to convince executives how much money they will save with good data.
Fortunately, there is hope for us data quality practitioners. In 2002, a great thing happened to us: Sarbanes-Oxley (SOX).
Sarbanes-Oxley requires not only that all organizations maintain accurate information (read: clean data) but that they be able to prove it. This is a great thing for data quality practitioners. We can now say that it’s no longer sufficient to manually clean bad data as it shows up -- if this data is tied to the financial picture of the company in any way (and what data isn’t?), it needs to be provably clean. If you work to build and enforce data quality, you will be giving your clients, managers, or customers exactly what they need to show SOX auditors that their enterprise is addressing the issue as required by law.
Now data quality goes from being a luxury that few understand to a necessity that every organization needs for compliance. I can tell you from personal experience that it’s a much easier sell using the SOX approach.
Why do I care about data quality? My answer is that I like things to be clean and I like using computers to solve complex problems -- and that it pays the bills. Why should an organization care about data quality? That’s easy -- because they have to comply with SOX.
Gian is the owner of Loreto Services and Technologies, where he is responsible for everything from sales to implementation to payroll. You can contact the author at gian@loretotech.com.
.