Data: The Good. The Bad. The Ugly.

The topic of data fuels our industry, and our fascination with the concept of what makes good, bad and ugly data, drove our Data Analytics committee to assemble case study examples of when industry took bad / ugly data and turned it into good data.

What surprised even the savviest of data gurus was the consistent theme that emerged as being the most important component of good data: Naming Conventions.

To get there, we recognized that ugly data doesn’t get worse than fraud – so don’t turn a blind eye to what is and isn’t fraudulent data.

Additionally, we need to ask ourselves whether or not we should be updating groupings, while asking more complex questions:

  • What is social as a category? Is it too vague?
    • Perhaps we need to start breaking this ‘category’ down into more granular components
  • Does prospecting or retargeting get the same weighting, or does it deserve different weights?
  • Are you driving a new consumer or just getting to a site from a search action?
  • Not all impressions are created equal. Were there viewable impressions? Were they delivered on target?

As you work on digging into these questions (and yes, there are more) you can start to build out custom clusters to gain richer insights. But overall, what fundamentally remains is that of proper ‘Naming Conventions.’

Across all case studies shared, what mattered the most was setting up proper, robust and consistent naming conventions, as it helps with the granularity of the data – which, in turn heightens the output. With proper naming conventions you allow for:

  • More models / model curves
  • A better understanding into the best, most efficient multimedia to use and buy against
  • Greater consistency with partners

How does one set up proper Naming Conventions?

For the case study examples presented by industry specialists and participants of the Data Analytics Committee, the following is a sample of steps taken:

  1. When starting, demand the required time upfront so as to be ahead in the long run!
    1. Cleaning data can take up to three months to do effectively. But, no matter when this process happens, time is saved in the long run
    2. Must take on an engineer mindset (create or change a current process or product so as to get a better outcome)
  2. Assemble a team that understands what each department wants and needs (i.e.: planners want brand safety, fraud protection; another team might want direct response metrics etc.) – be collaborative so as to have a clear understanding of each departments overall goals and objectives
  3. Once a framework has been established, feed data through
    1. Weed through anything that does not fit into the specified criteria established in step #1 and/or the framework, making necessary modifications along the way

In the end you are left with a more granular framework to plug data points into that has been built upon better naming conventions, allowing for greater consistency across the organization and all data partners!

Looking to Learn More?

In the meantime, join our MIXX event whereby we unpack the implications that the new Direct to Consumer economy has on our industry with a window into data, followed by our Business of Digital: Report on Metrics for more insights into the Canadian consumer and the media audience marketplace.