Why being a data-driven company isn't enough

In the race to digitalize big data, business use cases are too often neglected. Here are ways to make big data and AI projects work for your business.

Business Data Analysis Concept. Marketing Strategy, Analytics with People Characters Analyzing Financial Statistics Data Charts on Computer. Vector illustration

Image: invincible_bulldog, Getty Images/iStockphoto

Companies might be better off in their artificial intelligence (AI) and big data business projects if the goal wasn't so much on being data-driven but on being business-driven.

I'm thinking about a thought-provoking article in the Harvard Business Review from February 2019 by Randy Bean and Thomas H. Davenport titled Companies Are Failing in Their Efforts to Become Data-Driven. The authors say that companies are failing to become data-driven, though they are aggressively pursuing AI and big data. The article also said 77% of executives believe that business adoption of AI and big data is a major challenge. Even if companies are data driven, adoption can be daunting.

Here is a case in point: An IT department of a major financial company I am working with is digitalizing and indexing all of its documentation. The project manager told me the project's goal is to get rid of all the paper and to digitalize it in a central, searchable repository of data that applications and systems across the company could link into.

From an IT standpoint, it is easy to see the logic; however, I am struck that no one at the company identified the business processes that could be optimized by this repository of data that was being digitalized, or thought about the analytics and business insights the company could derive from this data, which would become searchable. 

In short, there is no strategic plan or business direction aside from digitalizing the data. The product is truly "data driven," since it is producing data, but it isn't business driven, and the company isn't positioning itself to see any impactful business results for many months. 

SEE: Data analytics: A guide for business leaders (free PDF) (TechRepublic) 

Many organizations are going through a similar struggle. An inherent challenge with many of these technologies is that you just can't plug them in like a traditional transactional system and start cranking out invoices, inventory reports, and purchase orders. Instead, most AI, analytics, and big data projects go through a series of iterative processing and testing until there is a consensus among IT, data scientists, and business users that the results these projects yield are "true." Along the way, some of these projects make it and some don't, and the risk is that the project work you're doing turns into a data-driven exercise that isn't business driven.

So how do you stay on point and avoid the trap of turning into a data-driven organization instead of a business-driven company? Follow these three tips.

1. Identify your business cases before buying technology

This is very important. If your company can't visualize a measurable revenue, cost reduction, work environment, or customer satisfaction benefit from the analytics, IoT, or digitalization, you shouldn't be spending your budget.

2. Get out of pilot mode

For the past few years, analytics, big data, and AI projects were granted a mulligan when it came to producing measurable business results because they were new technologies that were being run in pilot experimental modes with the understanding that the projects might or might not work. This honeymoon is over. Management expects big data projects to produce tangible business results just like transactional data systems.  

3. Communicate project status and methodology

Even if big data projects are now considered to be in a mature mode and are expected to produce results, this doesn't change the fact that they are more difficult to implement than transactional data systems. 

Big data analytics and AI use algorithms that have to be continually refined until they reach at least 95% accuracy before being placed in production. Consequently, there is an iterative approach to big data projects and testing until you reach an acceptable level of accuracy. This repetition can give management the impression that a big data project isn't being run well because of the continual modifications and retesting. Because of this, it's essential for CIOs and big data leaders to explain in plain English the differences between transactional and big data testing methods and projects so management understands the process. 

Remember your goals

Big data projects should always be business-driven from inception. It is not enough to collect, curate, and process data for data's sake, expecting that the business use cases will simply follow. Keep business goals in mind, and you'll keep focus on what's really important.

Also see