Top-down Data Quality Approach with Conformed Dimensions

In a previous blog we discussed the Top-down versus Bottom-up DQ approaches. We defined some aspects that are unique to each approach but it was high level, so in this post we’ll dig into how we use the dimensions of data quality during the Top-down approach.


As a consultant and educator you want to provide resources to your audience in an understandable way that progressively enables their organization to achieve higher levels of data quality. As such the following has some logic to the progression, but you can pick and choose how you want to use the Conformed Dimensions based on your organizational readiness.


Relating to the Top-down approach, I find that if the business audience has absolutely no exposure to the dimensions of data quality a Jumpstart is helpful. DQMattters offers a class that not only steps through the Conformed Dimensions so that your team can communicate effectively. This also will help you know when to use them and when they might over complicate things.


Activities Conducted During a Top-Down Approach Using the Conformed Dimensions:


Activities Conducted



Educate all audiences (likely separately based on their needs) about the dimensions of data quality and the three levels (Dimension->Underlying Concept->Metric).

Separate department audiences: Sales, Finance, Product, Marketing, Application Development, Systems Administration…etc.


Identify key business areas facing data quality issues (e.g. department, sales funnel, billing…etc)


Business areas might be the same as departments (above) but it could be a cross functional process like Sales funnel that includes product teams, sales persons, marketing, distribution and IT


Identify stakeholders of these areas (2) and interview them for top priorities

Interviews could be individual but to save time we often use a focus group approach


Define a scope of data quality work. In other words, are you forming a data quality function with team members or only going to research and rectify a tactical issue?


Sales Example: Improve data quality of customer data collected in the sales funnel across countries.


Document a clear objective for each of the areas (4). Form a statement and measures of success you expect when done.


When done we will have defined which attributes of customer data are required and implemented controls for these in all systems. We must be able to join sales data for all countries by accounting close and sales incentive periods via the data warehouse. Sales dollars must reconcile between both time periods.

Some would argue that educating our audience about the dimensions of data quality (step 1) before we’ve defined the business area and issue is pointless or even confusing. On the contrary, I have found that by giving everyone the language to communicate data quality they come to a consensus on the subsequent steps faster because they are using the same language which helps them more effectively prioritize issues and focus.


By educating all audiences about the dimensions from the beginning, people naturally start to identify ways to improve data quality using the dimensions of data quality, like:

  • Identification of forms that collect the same information but in different ways. 
    • Validity- They understand that valid data must be ensured via drop-downs at data entry
    • Completeness (Attribute Population)- They identify fields that aren’t required on a form, but are required later during reporting
  • Identification of the correct source of data.
    • Source Documentation- Often intermediately savvy data consumers pull data from intermediate sources (either because they lack greater permissions, or simply because they don’t know about more advisable sources).
  • Identification of poorly defined term between different stakeholder groups
    • Metadata Availability- Only with detailed business vocabulary can users agree upon what they need and set desired DQ levels/thresholds.


Obviously, this isn’t a comprehensive list of CDDQ underlying concepts used during a Top-down approach, but you should be able to see how the knowledge of the dimensions opens possibilities previously inconceivable. The best part is that these discussions with business audiences enable more constructive subsequent conversations with IT members who have conducted Bottom-up approaches to data quality.


In the next blog, we’ll discuss where the Conformed Dimensions are used during the Bottom-up approach. You’ll immediately see why the most successful companies implement DQ using both approaches and leverage DQ findings from one approach within another. It’s like a puzzle with only half the pieces if you only use one approach.

Did you enjoy this blog? If so, help out by taking the annual Dimensions of Data Quality Survey open now!