Refining the enterprise data strategy to deliver financial impact

There are five critical but often overlooked areas that need to be an integral part of building a robust enterprise data strategy—a data strategy that can deliver a tangible financial impact for an enterprise. This presumes that due diligence has been done on the fundamental aspects of building a data strategy, such as business goals alignment, data management, data governance, data privacy, and data security, across people, processes, and technology enablers.

To deliver tangible financial impact, an enterprise data strategy must consider five critical areas:

#1 Follow the money.

When building a data strategy, you need to follow the money i.e. link it to a clearly articulated financial outcome. This is especially important in today’s business environment when your entire business model is continuously challenged by competitors who are agile and nimble, and where the goalposts are shifting constantly.  Data strategy should ultimately be about improving your bottom line and top line, anything less will become quickly redundant and unnecessary overhead. The data strategy needs to be built on key financial levers spanning your entire organization’s value chain. Each of these levers then needs to generate an ROI, that can be quantified, monitored, and measured. These financial levers should fuel and guide the execution of your data strategy. 

#2 Beware of the missing bullet holes.

There is a fascinating World War II story about Abraham Wald, an eminent mathematician. The US military was losing a large number of fighter planes over enemy territory in Europe.  The proposed solution was to armor the planes on the fuselage, as the available data from the returning planes showed that the maximum bullet holes were on the fuselage, and there were very few on the engines. Wald however suggested armoring the engines instead. His insight was based on asking “Where were the missing bullet holes?’ He reasoned that the planes that got hit on the engines were not coming back, and could not be counted in the available dataset. This phenomenon is called survivorship bias, a form of selection bias that ignores data from failures. When crafting a data strategy, it is vital to actively look out for any missing data that needs to be captured to complete the full picture. 

#3 The bigger truths, like the bigger fish, are always deep below the ocean surface.

To find the truth, you will need to look beyond the surface level. The surface is where the waves lash, and there is maximum noise on datasets. The deeper you go (i.e. cleanse and normalize data from all available sources), the better your chances of uncovering the truth. Noise in data is essentially unnecessary data, that has no meaningful use, and you can expense a large number of resources in extracting, interpreting, and transforming this data—only to find the insights generated are unreliable. This has serious implications for the execution and outcome of a data strategy that is meant to ensure qualitative data availability at all costs. Organizations must make data noise reduction a high priority at the start of the data transformation journey and implement a continuous feedback mechanism to filter out the noise from the data for reliable and trustworthy outcomes.

# 4 Quantity matters, but quality matters more.

Deep learning algorithms require a large amount of data to make sense of inputs and bring about acceptable outputs. This is mostly true if you want to detect a cat from a mouse, or a happy from a sad face. However, if you want to detect cancer cells from an MRI, PET, or CT scan, it requires relatively lesser but higher quality data. Quality of data is a critical area that will drive consistent and successful outcomes for a data strategy. Machine Learning pioneer Andrew Ng argues that focusing on the quality of data fueling AI systems will be critical to unlocking its full potential. He talks about “data-centric AI”, the discipline of systematically engineering the data needed to build a successful AI system. Make quality of data a top priority for any data strategy rollout.

#5 Data has an expiry date.

Like anything in life, if you get your timing wrong, you will miss out on potential opportunities. Building and executing a data strategy needs to be time-bound. This means that you do not have unlimited time to unleash the full potential of your data. The staler or older a data set becomes, the more it starts to create a negative counter effect on the results you will get from harnessing it. If your business decisions depend on data, then it is imperative that you have access to fresh data, and where possible live data, that is telling a story that is relevant for your business today. Historical data has a purpose, but live and real-time data is what will offer the business cutting-edge insight, as the future starts to look less and less like the past.