Veracity of data for finance: Step-by-step

Share this post:


Financial decisions require confidence in the veracity of data on which they are based. The challenge is that by 2015, 80 percent of the world’s data will be uncertain.

Data uncertainty arises from many sources and in many ways: as inconsistencies, ambiguities, technical errors, time delays, assumptions, and even sarcasm and outright deception.

The challenge is how to deal with uncertainty when analyzing big data. How do we represent uncertain data, reduce the uncertainty of data, and reason about that data in a way that allows us to make sound business decisions?

New approaches are emerging to account for uncertainty in data at a giant scale.

veracity-of-data_1Step 1. Establish the base

Every organization has data it can trust. This base data comes from reliable sources and provides the foundation for making sense of less certain forms of data. A key step in managing uncertain data is establishing and providing a unified view of the base so that each department has access to a single, trusted source of data.




veracity-of-data_2Step 2. Gather more data

An instrumented and interconnected world offers an abundance of new data. Most of this new data is low veracity: it comes from uncertain sources, such as social media, blogs or video. When handled correctly, this data provides context and increases the value of our base data. Start by examining as much data as possible from a variety of sources.



Step 3. Combine the data

Integrating related data helps maximize context and reduce uncertainty. For example, mobile location data in complex environments such as cities is notoriously imprecise. By combining GPS data with recent social updates and accelerometer data—and correlating that with weather data or past purchases—CFOs can gain a better understanding of where our customers are and how likely they are to buy. That information, in turn, can be used to make better immediate-term financial forecasts and decisions.



Step 4. Extract insights

Analyze and extract insights from all available data. Tools such as natural language processing, customer predictive analytics and social media sentiment each go beyond the basics to extract meaning, reveal insights and build context awareness. Be clear about the questions you want to answer so you can employ the right set of analytics for the job.




Step 5. Build models and scale to the organization

This includes building models that incorporate uncertain data. Mathematical approaches such as stochastic reasoning or Monte Carlo simulations take uncertainty into account to yield predictive recommendations. To foster a culture that values data-driven decisions, make these models understandable and accessible across the organization.




More stories

How blockchain-based payment is changing the cannabis industry

Legal cannabis is a $7-billion business in the U.S.—and it’s almost all done in cash. That’s in the process of changing. The mental picture of people stuffing piles of $20 bills into their mattresses isn’t far off from the truth for some cannabis business owners. The drug may be legal in some states, but it’s still […]

Continue reading

The (next) big one: Despite warnings, new massive security breaches are only a matter of time

Security experts these days can often feel like they’re shouting into the wind. Amidst the rush by businesses and consumers to embrace new technologies, warnings about security vulnerabilities fall by the wayside. Who can be bothered with such trifles, when there is so much digital awesomeness all around us? The Internet of Things hack last […]

Continue reading

Halftime helper: Will the locker room of the future be cognitive?

The average NFL halftime lasts only 12 minutes.That means coaches have a mere 720 seconds to digest thirty minutes of football, analyze the performance of dozens of players, assess the health of the injured, rehydrate their team, and make real-time adjustments to a game plan that took weeks to develop. Oh, and if they have […]

Continue reading