Data•May 11, 2020
Data May Be the New Oil, But Few Know How to Refine It: Data Strategy Series Part 1
No one argues about the value of data in this new century. Both people and machines are generating enormous amounts of data. In 2016 an IBM Watson analysis estimated we were creating 2.5 quintillion bytes of data every single day; said differently, 90% of all the data in the world was created in the last two days! With the abundance of data, the natural and right question many leaders are asking is, “How can I use it to my advantage?” The natural answer is artificial intelligence (AI) and machine learning (ML), and there is no longer any doubt that AI/ML and augmented reality solutions have the potential to change the way we work and interact with systems and each other based on this data.
the big data opportunity
Unlike previous waves of speculation (mass automation predicted by Time in 1961, the remarkably sophisticated HAL 9000 of 2001: A Space Odyssey, or even Life predicting in 1970 that “in from three to eight years, we will have a machine with the general intelligence of an average human being”), today no speculation is needed to begin to understand the impact AI/ML will have. Indeed, it’s already changing the world. AI/ML systems are all around us, including facial recognition to unlock your phone, object/weapon detection in X-ray machines, and fraud detection on credit card transactions. Every person interacts with dozens if not hundreds of AI-powered systems every day. What’s more, the impact of these systems is measured in the hundreds of billions of dollars across most sectors.
the unfortunate data reality
Yet in our experience, very few companies realize these benefits. There is a disconnect between what’s technologically viable in the real world and what we see companies successfully implementing with their data strategy. We aren’t alone in this observation:
VentureBeat AI reports 87% of data science projects never make it into production.
NewVantage reports 77% of businesses say adoption of big data and AI initiatives continue to represent a challenge.
Gartner says 80% of analytics insights will not deliver business outcomes through 2022 and 80% of AI projects will “remain alchemy, run by wizards whose talents will not scale in the organization.”
Worse, these companies end up incurring enormous costs for consulting firms, implementation projects, complicated systems, and storage of unused data that doesn’t produce any value. A study by IDC in 2012 estimated that as much as 99.5% of collected data never gets used or analyzed. It’s doubtful that number has changed much in the last 8 years—in fact, it’s probably gotten worse.
Why are these efforts falling so short? Why can’t companies tap into the enormous value that lies in all their data?
We typically see a combination of four issues when working with our clients:
A rush to “do something” (without understanding why).
Poor data quality/governance.
Immature data engineering practices.
In this blog post, we’ll cover the first mistake and some of our recommendations. We’ll follow up this article with separate blog posts on the other three.
mistake #1: don’t rush to just “do something”
Albert Einstein said that “a perfection of means and confusion of aims seems to be our main problem.” Nothing describes the current state of big data initiatives in our industry better. Consultants (like us!) and industry pundits tell us we need to be doing it. We have created thousands of tools, developed patterns and best practices, and spawned dozens of conferences to tell us how to do it.
Yet few seem to start with what they trying to accomplish or why. Tom Davenport, a senior advisor at Deloitte Analytics, captured this perfectly in CIO, “The biggest problem in the analysis process is having no idea what you are looking for in the data.”
As a result, we typically see “enterprise” efforts that try to create enterprise taxonomies, capture every piece of data from every system, or rework the entire data architecture across the enterprise. These massive efforts rarely succeed and are usually the source of the statistics outlined above.
a few recommendations
Instead, we’d recommend starting with a more directed approach:
Identify one or two key business issues to address.
Determine what data needs to be gathered, analyzed, or exposed to address those issues.
Build just enough to address that business concern. While it’s great to have a good enterprise data architecture, we recommend implementing only the thinnest slice of that necessary to achieve your goal. Otherwise you might be wasting resources or delaying value.
Measure improvements and adjust based on what you learn (more on this in the future posts).
For example, the marketing department at a large retail client wanted to increase sales through personalized marketing and customized customer journeys. This is a rather large goal, so Credera divided the larger goal into small consumable pieces of work. One simple place to start in customer segmentation utilizing customer data is by grouping customers by their sales and targeting segments differently through email campaigns. Development-wise, this task is rather simple if you have customer and transaction data in the same location.
We were able to quickly enable this segmentation through the use of data. The marketing department was able to test campaigns on different segments of their customers and improve email campaign performance.
take note of constraints
While that may sound simple, it often is not. The other three issues above often come into play. For example, the lack of data or inconsistent data will often limit how effective you can be in the first iteration. Don’t be afraid to scale down your efforts to get some type of value. Take note of these constraints and then address them in future iterations. We’ll cover some of these strategies in the next blog post.