With technology advances making data storage boundless, companies have additional pressure to collect, analyze and deploy data.

But eagerness has left companies with large stories of unused, uncatalogued information.

Here's why this happens: More than half of organizations' data is unused, according to a survey of 1,300 business leaders by data analytics and processing firm Splunk. Yet, 81% of respondents say data is valuable to their overall success.

Part of the disconnect with unused data — which Splunk refers to as "dark data" — is businesses are unaware of what data they store, or, they know what they have but do not know how to use it. 

The lack of visibility into data stores cramps eagerness to apply artificial intelligence and machine learning. While in the early stages of adoption, the application of AI and ML requires data organization.

This calls for businesses to rethink their approach data. Consumer credit reporting company Experian starts at the end. 

The company is beginning with the customer and consumer, Shri Santhanam, executive VP and general manager of global analytics and AI, told CIO Dive. Working backwards helps the company "pull back and understand really what the ML and AI capabilities to realize that outcome are."

Ultimately, it will help Experian understand what data outcomes require, whether that's improving financial inclusion or building new customer-oriented services. 

"It's less of a technical or an analytical exercise in collecting that data," Santhanam said, which can lead to "enormous amounts of investment in people, time, data without clarity on the outcomes, which we want to get."

Overhauling data

Applying artificial intelligence and automating insight generation is a far cry from traditional data analytics. 

"Traditional analytics is math," said Michele Goetz, principal analyst at market research firm Forrester. Training algorithms for analysis is the complete opposite.

ML capabilities in an AI system are best at classification, interpretation and inference around patterns, said Goetz, in an interview with CIO Dive. Just like we train "ourselves to have domain knowledge and become experts, algorithms are being trained to be specialists."

The reason to train those systems is because a company has a purpose in mind, whether that is meeting a customer need or finding a cancer diagnosis. 

Handing an algorithm raw data will prove fruitless. It could come back with disparate conclusions or concepts. 

Instead, effective ML applications always start with "what we want to do," Goetz said. 

Looking at the end goal of a project first might sound like a simple concept, but it can prove difficult to execute. 

Companies are not always set up with the end in mind, Santhanam said. Stakeholders might provide different project pieces, but do not hold end-to-end consumer and business impact. 

Technology can also prove to be a limiting factor. Since the arrival of CIO Barry Libenson in 2015, Experian has worked to reevaluate how products were built and software developed. 

Libenson's goal was to create a stack that had more synergy, reusability and a more flexibly delivery model for clients, he told CIO Dive in an interview earlier this year. The modernization effort is grounded on a "build anywhere, deploy anywhere" strategy, freeing customers and software developers from concern about where they want technology delivered. 

The infrastructure investments are the foundation for Experian's data strategy. The effort shows foresight and a commitment from leadership to actually enable an end-to-end innovation cycle, said Santhanam.

Experian's data-oriented roadmap

It's early days at Experian for Santhanam, who joined the company in June. 

His efforts will focus on building a foundation around data, analytics and AI while simultaneously working backwards from key outcomes. In addition to working with the CIO, he will partner with stakeholders across the business. 

Santhanam said success in his first year will show: 

• Evidence of tangible consumer impact.

• Clear evidence of market success for Experian and its customers.

• Meaningful progress on building foundations and kick-starting an agenda for rolling ML and AI into the business.

His data-oriented efforts can apply to products like Experian Boost. The product allows consumers to introduce data to build out credit files and improve their score by showing regular payments on services including utilities or phone bills, according to the company's annual report. The goal is in part to bring more consumers into the credit system.  

Experian Boost is a ripe application for AI-driven data, instead of traditional analytics. 

Building the product required "first truly understanding the consumer problem that we wanted to solve," Santhanam said. Looking at the issues of financial inclusion, Experian came back and asked what data and analytics were required. 

The business realized "part of the data to help consumers achieve better outcomes resided with the consumer themselves, he said. Rather than "going into an exercise of trying to collect that data, we created a model where we said 'we'll allow the consumer to help themselves' and then pulled back from that and created the analytics."