EA Principles Series: Treat Data as an Enterprise Asset
EA Principles Series: Part 3
This is part 3 of a seven-part series about our Enterprise Architecture principles at Chick-fil-A.
The importance of data
We all know that data is critically important to organizations. If we want to get the best results possible — the best possible customer service experience, the best job experience for our team members, the best operating efficiencies as a business—we are going to need to collect and manage data really well, so we have to develop a culture that sees data as an asset.
The cultural part of this is very important. We need good implementation and analytics practices, but it is just as important that any person in our organization approaches data with the right mindset when it comes across their laptop screen or mobile device.
How do we develop a culture that ensures data is accurate, well-understood, and decision-making-worthy? That we produce and share new data permutations and transformations that are correct and useful to the organization? That we do not hoard our precious data assets to ourselves? That we default to sharing from our transactional systems to our analytics ecosystem?
Principle: Treat Data as an Enterprise Asset
Here is our verbatim.
Data can and should be a prized enterprise asset and potential differentiator at Chick-fil-A. Data must be cleaned, managed, catalogued, shared and secured carefully and intentionally. Data owners should think of their data like a product. Owners should consider potential internal (or even external) customers, their needs and requirements, and the overall quality and usability of the product. Legal and privacy policies should be well known and adhered to as well.
Teams should practice “democratization of data,” making intentional plans to share their data with the rest of the organization. When making data available for downstream consumers, teams should leverage strategies like API-First, or provide offline access via the Analytics ecosystem or a non-transactional datastore (Redshift, etc). Teams should not allow direct, non-API access to their transaction systems from outside of their software stack.
Why? Availability of a business domain’s data is a key aspect of enabling velocity and momentum for gleaning information and business value from analytics. Whenever two teams need to work together to access the required data, velocity will slow to the speed of the slowest team and will not help the next team that needs the same data.
Commentary and Tradeoffs
What are the desired outcomes of this principle?
Outcome 1: Create an organizational mindset around data as a product and potential differentiator
Have you heard the term “AUM?” Assets Under Management. It is generally used in reference to financial assets that are managed by an investment or hedge fund. Presumably, these financial assets are being actively managed towards maximum yield for their owners.
It might be dumb, but perhaps we could talk about Data Under Management (DUM) in much the same way. Presumably, the data that we have should be actively managed towards maximum yield or utility for our organization.
To do so, several things have to happen, much in the same way they do in the financial world.
- We must understand what assets we have — we can’t manage what we do not know about, and we cannot understand what we cannot find or read about. There is no google search result for our enterprise data (thankfully!) so we have to develop methods to catalog the data assets we have, what they mean, and how they are useful internally. We do this in a number of ways that serve technical audiences (think Hive metadata catalog) and business audiences (business glossary, registry of datasets).
- We must understand the macro environment — how does a dataset relate to the larger business? In what context was it created and where did it come from? In what contexts does it make sense to be used? We need to be able to easily answer this question and enable analysts to do the same. This leads us to investments in data lineage and other systems to help build a shared organizational context. We also need to make sure we consider how data will be used in the future when we are creating it: are we capturing the right fields, the history that might be needed (though sometimes that can be created downstream), etc? It is easy to develop an application that creates / consumes data that meets the requirements of a stakeholder, but we must go further and consider the needs of the enterprise.
- We must generate returns on the data assets we have — just storing data is not super useful. It can even be a liability (security risk). We have to put our data to work. For it to be put to work, it has to be clean and well-managed. Only then can we begin to report against it, build algorithms and models against it, and share it with others to generate business value: our desired return.
We encourage teams to think of their data as an asset and to anticipate its organizational value. Not all data assets are equal, so we have started our journey with things closest to our operating model and strategic investments, and worked outwards from there. This principle is aspirational and still something we are growing towards at Chick-fil-A.
Outcome 2: Ensure data security and privacy
We have a lot of data at Chick-fil-A. It is therefore critical to ensure that it stays secure in its transactional systems and secure within our analytics ecosystem. In addition, we have to ensure the data we have is used legally and ethically, which is of the highest importance to our organization. While completely open access to data would be very exciting from an innovation standpoint, this outcome brings some balance to the equation and requires that we have appropriate and flexible access control models that restrict access to a lot of our data while still empowering people to explore datasets that are less sensitive or that are related to their work responsibilities.
Outcome 3: Encourage democratization of data
Finally, we want to encourage the sharing of data. Data that is locked up is not useful to the enterprise. All teams are pushed to share their data to our Enterprise Data Lake, and to create Enterprise APIs or Business Events as it makes sense. This breaks silos down and unlocks the ability to become an agile, nimble, and responsive business. We’ll share more about that in our Composability principle.
Outcome 4: Achieve Data Quality
Inspired by AirBnb’s “Midas Promise”, we seek to advocate for the following:
- Accuracy — Ensure data is validated for accuracy with as much automation built into pipelines as is viable.
- Consistency — “Certified data and metrics represent the single source of truth for key business concepts across all teams and stakeholders.” That is verbatim from AirBnb, and we have nothing to add to it.
- Timeliness — One of our internal analysts likes to say, “the longer the data latency, the more use cases fall off the radar.” We believe this to be true. We consistently seek to advocate for more real-time sharing of data and less batch (though batch certainly still has its places). We need to notify consumers when data is delayed in its processing or may be stale via automated processes and “incident management.” Much of this is still aspirational for us, but we are moving in that direction.
- Cost Efficiency — We are a little lighter on this as a lot of our goals are about solving growth challenges, not maximum cost optimization. That said, we tag our AWS resources and have dashboarding capabilities to help us understand the costs associated with our pipelines and data storage / transformation routines.
- Usability — Data should be well-labeled and supported by documentation. We do this through our Data Lake catalog where we track metadata completeness and other attributes like quality scores.
- Availability — AirBnb’s take is that certification is mandatory for important organizational data. We agree, but are not to a point where we have reached internal agreement on certification criteria or where we can get organizational buy-in to pursue this, so this remains aspirational as well.
Conclusion
This principle is a great example of something we shared before in our EA Matrix post: we clearly have and govern a perspective on “data as an asset,” but we do not own and steward this area alone. We do not own all the enterprise data.
Thankfully we have Data Architects and Analytics Engineers and Platform Owners in our shared platforms practice at Chick-fil-A who help us with these outcomes. We also have an Enterprise Data and Analytics department that helps develop analytics practices within lines of business and helps steward some of our data products and portfolios. We are not in this alone, and that is good since what we hope to see is not a mega-architecture practice that has to be involved in everything to make sure its on the rails, but rather a culture of treating data as an asset.
In conclusion, data is a powerful asset that can lead to transformation, both in its silos for functional purposes and when it is unlocked and shared with the enterprise. An organizational mindset that treats data as an asset will ensure we share our data well, but also protect it and manage it carefully to ensure we get the best possible returns from it.
Other posts in this series:
