Businesses should begin to consider “Data-Optimized” Networks. We’re seeing an industry shift in data and analytics from “big data” to “good data,” led by artificial intelligence (AI) pioneer Andrew Ng. His evangelizing of concepts like “data-centric” and “data ops” rightly point out that 99% of machine learning (ML) research focuses on models, and only 1% on actual data. He also has examples that show where focusing energy on improving data quality – even with smaller data sets – has greater improvements than focusing on improving models.
This exposes a deeper problem in the partnership between business and data domains of knowledge in organizations: Business leaders will say their organization and culture is ‘data-driven,’ pointing to hundreds of millions of dollars of investment in expensive software systems accompanied by painful, time-consuming, and complex organizational change management to undergo “digital transformation.”
Prioritizing Reliable Data Frameworks: The Myth of Data-Driven
However, if you talk to the data scientists and business intelligence analysts, they will resoundingly say that most of the data isn’t clean or reliable; it’s biased –and that’s if it can even be found or accessed at all. Instead of using data to find insight, they spend 80% of their time or more cleaning data.
If asked to make changes that might address this, quite often business leaders will say they’ve been promised this would be solved many times, but each has failed, and so they stopped believing in it. The dissonance between these two disciplines is neither productive nor harmonious, and the problem for the organization remains unsolved.
Why a Data-Optimized Network?
For many business teams “pretty good” data or “fairly clean” data might allow them to show their executive leadership how they use data to show strong indicators of directionality. Many business teams don’t want to, or don’t have budgets to, bring on data scientists. However, this status quo comes at cost: Are executive leaders truly getting the maximum ROI for their investment? Are they missing out on opportunities that they aren’t even aware of? How can we prove or disprove that the measures being used are even the most important measures if we aren’t willing to evaluate it? And thus being “data-driven” turns into a myth—it’s a story of people lacking scientific rigor to challenge existing held beliefs.
Why are we looking for incremental improvements in a pool of “data swamps” that were promised to be “data lakes” that would solve this problem? What the industry needs is a reliable framework, a playbook that step-by-step and end-to-end describes how to create, curate, and disseminate pristine data. This playbook needs to integrate UX design, data science, software engineering, product management, and business strategy domains of knowledge into a holistic approach that guarantees an ROI for executive leadership investments.
What is a Data-Optimized Network?
At Clarkston, we call this framework the “Data-Optimized” network, and we are helping clients remove themselves from trying to clean up data swamps and instead get the ROI they were hoping for.
Here’s how we describe it:
You can be data-driven – but if data isn’t clean, it’s very difficult to use effectively.
You can be data-centric – but if only data scientists have meaningful data, it’s siloed.
You need to be data-optimized. This means:
- Data is high-quality, interconnected, and scalable.
- Data is a first-class citizen.
- Data governance is the foundation.
- IT & operational process decisions are optimized to maximize the quality, dimensionality, and volume of data in the most efficient manner.
- Every single touchpoint of your critical data is in a unified data model.
- The granularity of data unlocks value to maximize efficiencies.
Our experienced Data + Analytics consulting team would love to hear your thoughts and challenges on your journey with data and become your partner in becoming a data-optimized organization.
Subscribe to Clarkston's Insights
Contributions from Ron Itelman