Insights

Data Governance – A Better Approach to Value Creation

Data Governance – A Better Approach to Value Creation

Written by: Mike Lampa + Anna Navarette, GDM Advisors
Artwork by: Tim Main, Walker Fine Art, Denver

Governance – what is it?  If you look it up, you would find multiple variations of what it is or could be or even should be.  In most cases these definitions lend to that monolithic approach we are all so familiar (and frustrated) with.  This boil-the-ocean approach typically includes everything from data architecture, modeling & design to consumption of the data.  With these bigger efforts, Gartner has provided us insight that 60-85% of companies’ data-centric projects fail because of poor or absent data governance.  It’s no wonder why so many organizations are demanding we find a new way to implement and operationalize a successful data governance program. One that generates value.  

Great Data Minds believes a lean/agile approach to governance is the best operating model to move your data and analytic program forward.  Take a continuous improvement approach. Identify the next most pressing problem you are trying to fix or risk you are trying to mitigate, build the right thing, and build the thing right. Execute with a product approach vs. the project approach.  Governance considerations are an inherent part of the requirements for building data and analytic products. Since data continues to grow and the desire to utilize the data to enable analytics which provide intelligent information to the business continues to grow, a strong and pervasive governance mindset is a must.  The key is to avoid creating monolithic organizational constructs and implementing invasive governance processes and workflows. Data and Analytic Governance is the responsibility of everyone in the organization from data producers to data curators to data consumers.

Utilizing building blocks to define your data governance touchpoints is a must to identify where that governance will contribute to the value of the data and analytic products over their entire useful life-cycle.

Infrastructure:
In the lean-agile mindset, there is the need for ensuring the “architecture runway” stays ahead of the scheduling of delivering new data and analytic products.  What enabling technologies are needed to further the value generation from data and analytic governance across the organization? Technology to help move your data program forward. 

Policies & Procedures:
Avoid building a siloed corpus of governance policies and procedures. Instead ensure the continuous ideation, definition, development, validation, integration, deployment, monitoring, refinement and disposal of data and analytic products include the necessary governance considerations as part of that product life-cycle. These procedures are intentionally  integrated into the Agile Product Delivery core competency.

Oversight & Metrics:
Every Data and Analytic Product that goes through the Agile Product Delivery life-cycle has an intentional design around the creation of execution, usage and outcome telemetry.  This telemetry data serves the purpose of enabling the creation of governance metrics which facilitate the continuous oversight over the creation and utilization of the data products and analytics products are in accordance with their intended uses. When the metrics indicate shift or deviance from expectations, those data and analytic products will be triggered for governance review. 

Meta Data / Lineage:
Starting with critical data elements (CDE, define your data elements, come up with consistent nomenclature, document, utilize and adjust as you go.  This will create your data glossary. When possible, identify and define the lineage of your data, this will allow for opportunities to upgrade, deprecate as you become more mature in your data program.  Continue to review your data pipeline and add information to allow teams to better understand where the data is coming from, how it is being used and update this content into your data catalog.  It’s not everything, but it is the critical things. 

Data Security, Usage, Ethics:
Data security, usage and ethics are a component in your data strategy that cannot be forgotten. It is imperative that as part of your data program you understand how the data is being used, by whom and in what content.  Note who should be using things and understand where all of these PII type data exists. With the ongoing regulations in the US and abroad, knowing your data is a corporate responsibility.  This is a means of protecting your company and your clients/customers.  

Data & Analytic Outcomes:
Keep a pulse on how people are feeling about the data they are receiving and compare to the prior temperature before the program was put into place.  Determine what is trying to be solved and measure against that objective.  Other things to consider is how well the data and analytics (BI & ML) products are generating the expected business value.

Data Quality:
Incorporating defined data quality requirements for those critical data elements and implementing data quality business rules interrogations and providing visibility to the ongoing trends of data quality business rule adherence.  Overall assuring that the quality of the data is constantly being managed as part of the data pipeline.   

The Application of Governance:
Consider that you need to govern with an appropriately sized hammer, balancing the organization’s need to continuously innovate (try new things out, learn fast, pivot or preserve) against the organization’s need to protect the data and analytic products as valued assets.  Organizations have an obligation to their stakeholders to protect their interests, be in compliance with the ever changing regulatory landscape and continue to govern with appropriate rigor to ensure continuous regulatory compliance, organizational reputation protection  and ensuring the respect of the people and markets they serve. 

Is Data Mesh the next great frontier to democratize your data?  We’d like to hear from you!
Anna Navarette: anavarette@greatdataminds.com
Mike Lampa: mlampa@greatdataminds.com

Microsoft’s Data Strategy and How it Accelerates a Data Culture

Microsoft’s Data Strategy and How it Accelerates a Data Culture

Microsoft has been making significant waves in the analytics and data governance space recently with its release of Azure Synapse and Purview. Join Matt Smith, Microsoft’s CDO for the West Enterprise shares Microsoft’s Go To Market strategy to enable digital and data transformation in the enterprise. Matt will share his career and customer experiences and how they have influenced Microsoft’s customer and partner ecosystem to achieve more. 

What you will learn during this session:

• Enabling change without changing
• How platform convergence is enabling acceleration
• Democratizing AI for all workers
• Adoption of groundbreaking innovation on the front lines
• Enabling enterprises to get their arms around their data corpus.

Getting Out of the Machine Learning POC Cycle Funk

Getting Out of the Machine Learning POC Cycle Funk

Join us as we continue to educate our audience on efficient approaches for implementing AI/ML solutions.

About this Event

While the vast majority of companies are waking to the possible benefits of integrating AI applications to their products and internal business processes, few are succeeding to do this at scale and see returns on investment. Those that are, stand to achieve a severe competitive advantage.

What you will learn during this session:

What is the impact of successfully scaling AI/ML and what it is worth to the organization.

A clear understanding of what a POC cycle is.

What is key in your approach for escaping this cycle

Why is it that so many models that seemingly perform well in development then fail to perform in production?

A strategy to maximize success in moving from lab to production

Please join us and our special guest Eric Topham, CEO and Founder of The Data Analysis Bureau, one of a leading AI/ML organizations as we discuss efficiencies in AI/ML programs and provide insights on how to stay out of the dreaded POC cycle.

Art Credit: Jean Albus, Walker Fine Art Denver

Becoming a Better Caretaker through Integrated Data

Becoming a Better Caretaker through Integrated Data

Written by: Julie Burroughs, GDM Advisor
in collaboration with Mike Lampa, GDM Advisor

One of the hardest jobs in life is being a caretaker to your elderly family members.  Not only is it time consuming, the paperwork is daunting and the realization that you will soon lose them can take its toll.  

Recently, my 94 year old mother, soon to be 95, was sent to the emergency room three times within a 45 day period for the same ailment.  As in most large cities, it is the ambulance game. The one that is the closest to the patient picks them up.  If you do not indicate which hospital you would like to go to, the medics pick.  Even though it was stated at her nursing home that we prefer one hospital over the others, that request never seems to be communicated. The result, three separate visits to three different emergency room locations.  

With each trip, usually occurring in the middle of the night, I would receive the obligatory phone call asking “can you tell us what is going on with your mother”?  As part of my agreement with my brother he manages my mothers  health information. So, I call him at 3:00 AM ET to get his help. By the way, it helps when you have a sibling with medical credentials. Needless to say we are both in a haze and after a 15 minute call with the on-call ER physician we realize he is working with only information we are able to provide.  Nothing from the nursing home, nothing from the other visits to ER.  Based on this lack of shared information, my mom was subjected to the same tests and extended stays until they figured out what’s wrong.  In the end, it was the same diagnosis three times over.

The Great Data Minds team has been fortunate enough to be involved with the development of data products which are making the world a better place. This is true with one particular project we are working on currently alongside a very innovative team. This interoperability product is working to make a difference in how Patient Health Information (PHI) is shared among all of that patient’s care providers and care coordinators.  Here is the first bit of good news. AWS, Microsoft, and Google Cloud Platform (GCP) teams are investing significantly in this area, specifically around their Healthcare API’s supporting the Fast Healthcare Interoperability Resources (FHIR) standard.  With this particular team, we are working with the Google Healthcare API and the Google team that is supporting the technology.   

Phase One of the program started with a Minimum Viable Product (MVP) where we were able to connect one of their 18 Accountable Care Organization (ACO) Partners to their Interoperability Hub that is used for  intake of the PHI data.  Within 8 weeks (four sprints) we were able to prove the GCP technology and demonstrate the ability to notify a patient’s primary care provider, clinicians and care coordinator that their patient received treatment outside of their network.  With every step of the MVP, Google technical teams were available for consultation and collaboration. Google also provided funding to prove the technology.  Next step, scaling the solution out to the remaining 17 ACO partners within the network.  See Figure 1 below for an idea of the InterOperability Reference Architecture we used for this solution.

The Centers for Medicare & Medicaid (CMS) and the Office of the National Health Coordinator (ONC) requirements are holding care providers accountable to provide access to and share PHI among the appropriate providers and the patient.  This is not an easy task for organizations, and for a vertical that is still reeling from the digital transformation to Electronic Medical Records (EMR) systems. Many are running out of time to meet the new CMS InterOperability requirements.  But with the only alternative being hand coding as of two to three years ago, the work that is being done in this area by top cloud providers will make it easier for organizations to adhere to the compliance measures. 

So, what does the future hold?  Embarking upon a project that will take our B2B (provider to provider) work with Healthcare API to a B2C (provider to patient) solution giving the caretakers of the world yet a bit more help and confidence.  

War Wounds and Wins: Machine Learning Stories from the Frontline

War Wounds and Wins: Machine Learning Stories from the Frontline

Machine Learning is quickly becoming a household term and everyone wants to build and use novel ML systems, but getting started harder looks different than you’d expect. 

We have invited some of the best of the best data science professionals to share their experiences with deploying machine learning solutions.  Join us as we learn from them what to do and not to do along with hearing about their projects and the impact they had.  

Is Citizen Data Science A Good Thing?

Is Citizen Data Science A Good Thing?

Written by Mike Lampa, GDM Advisor
Image credit: Zelda Zinn, Walker Fine Art, Denver

Self-service Discovery Analytics established itself soundly in the mid-2010’s thanks to pioneers like Qlik, Tableau and Tibco Spotfire.  These vendors had a new sales model approaching the influencers and decision makers in the business and offering entry level licenses for a price that one could essentially put on a Purchase Card because that price was within their discretionary budget authority.  The User Experience of these platforms closed a huge and painful gap for the business users, but opened new concerns around data security and governance. The IT teams developing and supporting the enterprise analytics solutions were taken aback and tried to resist the movement to no avail.  Then the industry researchers published points of view on the value proposition of self-service and offered frameworks and approaches to enable the movement.  IT eventually came to the realization that it is best to embrace self-service analytics and collaborate with the business for the good of the enterprise. Once that acceptance caught on, the vendors enjoyed significant licensing revenue and account expansion.  Then those business analysts felt their next constraint. They were still dependent on IT to deliver the desired data for their self-service discovery analytics & visualization use-cases. 

On the heels of the incredible success and market disruption brought about by these self-service Discovery Analytics platforms came another wave of self-service enabling technologies; self-service data preparation / curation tools. The vendors followed the same commercial and user experience model as their self-service discovery analytics counterparts.  They enjoyed the same success and caused a similar market disruption in the data integration & transformation segment.  This time the research firms were out in front of the self-service data preparation products.  They offered adoption frameworks, product reviews, comparisons and ratings.  IT again struggled and resisted supporting this next self-service wave because it blurred the line for where and who should be building the data pipelines that extract, transform and load datasets with content considered fit for purpose by the business analysts producing and consuming those datasets. On the surface the self-service data prep platforms are functionally equivalent to the traditional IT-oriented ETL / ELT platforms. But they offer a significantly simpler user-experience. This helped reduce constraints and the dependency on IT to provide datasets. Business end-users building datasets also raised “data anarchy” concerns as these datasets proliferate across the organization.  Those organizations that intentionally embrace the value proposition of self-service simply extended the rules of engagement and governance guidelines to account for self-service data prep in addition to self-service discovery analytics.  Many of those same organizations also invested in modern data catalogs in order to automate the knowledge sharing, tracking and administration of those datasets throughout their entire life-cycle. 

Enter Self-Service Predictive Analytic Platforms and the Citizen Data Scientist.  Gartner defined the Citizen Data Scientist as “A person who creates or generates [machine learning] models that use advanced diagnostic analytics or predictive and prescriptive capabilities, but whose primary job function is outside the field of statistics and analytics.”  Glassdoor reports that Data Scientist is the number one job for the last four years. The U.S. Labor Department reports the demand for data science skills will drive a 27.9% rise in employment through 2026. There is a mounting huge demand for a short supply of qualified data scientists. Add the fact that organizations across all industries and all sizes are realizing Machine Learning (ML) and Augmented Intelligence (AI) is the new analytics table stakes. Increased demand, shortage of supply and mounting pressure to develop and operationalize ML models before the competition does.  A perfect storm! There is early evidence of software platforms coming out targeting the data savvy Business Analyst / Citizen Data Scientist persona once again.  The platforms provide a low-code or no-code user-experience offering pre-built ML models and algorithms targeted for pre-conceived classes of predictive analytics use-cases.  The stakes (rewards and risks) are orders of magnitude higher in ML and AI as compared to the descriptive / diagnostic analytics.   We need to scale but we need to keep our organizations out of trouble. How do we compensate for the unique expertise of a qualified data scientist to ensure we are using proven methods, selecting the right models for the right problems, using the right data to feed the models for training and validation purposes, ensuring the data being used and the outcomes being produced are accurate, ethical, legal and unbiased?  Start with leveraging and extending the self-service framework to include continuous innovation, exploration and vetting of new predictive analytics use-cases and hypotheses. Quickly determine whether to pivot to a different use-case or persevere.  In the event the decision is to persevere, it’s time to engage the qualified data scientist to perform a comprehensive Model Risk Management review prior to adding the use-case to the Analytics portfolio for continuous integration / deployment and operationalization release.   The value proposition here is we have the business domain expertise mashed up with the data science expertise.  We’re grooming the business analysts to be more self-sufficient even in ML / AI based analytics.  This helps scale the ideation of promising new ML / AI Analytic use-cases conserving the demand for the data scientist to focus on the mentoring of the business analysts and developing the next generation of more complex models to be published to the library of pre-built ML models.

Yes, it’s a good idea to extend self-service into the predictive, machine learning analytics space.  Extend the rules of engagement and governance frameworks to include [machine learning] Model Risk Management. Pair up the Citizen Data Scientists with qualified Data Scientists. Establish a continuous innovation / exploration discipline to incubate and validate new analytics-enabled use-cases. Then watch your competitors navigate your wake in the rear-view mirror.