Automation relies on data. When this data is unverified, outdated, biased, taken out of context, or even maliciously tampered with, problems start to arise. Yann Lepant, Accenture’s MD of Technology on the challenge
Autonomous data-driven decision-making is now the lifeblood of many business operations.
But with it comes a new vulnerability: inaccurate or manipulated information compromises the insight that so many businesses, institutions and even governments now depend on to make critical decisions.
Forget fake news, fake data is the latest threat to businesses.
What is Fake Data and Why Does it Matter?
The sheer rate at which organisations are now able to collect data has far outstripped the ability of humans to keep pace. Combining analytics and autonomous decision-making, however, can crunch these large quantities of data in a way a human never could.
Done correctly, it paves the way for a much more intelligent enterprise and has the potential to make both employees and processes more productive. But, there’s one key caveat: automation relies on data. When this data is unverified, outdated, biased, taken out of context, or even maliciously tampered with, problems start to arise.
This “fake data” is a very real challenge. Accenture’s Technology Vision 2018 report found that 79 percent of organisations today are basing their most critical systems and decisions on data, yet many have not invested in the capabilities to verify the truth within it.
There have already been a handful of high-profile businesses to fall victim to fake data. Amazon saw how a reliance on data could incite malicious users when its product reviews became subject to manipulation.
Third-party sellers were paying people to submit fake reviews, artificially inflating their product and seller ratings to manipulate the site’s algorithms. Amazon had to invest significant time and resource into tackling this issue. The solution was to give weight to verified reviews from customers who had purchased the product via the site and ban those who received discounted items outside of it.
Deliberate manipulation is, however, rare. Organisations are far more likely to overlook data that has simply gone past its sell-by date, but could still come with a hefty price tag. United Airlines realised that its seating demand forecasts were based on decades-old assumptions about flying habits. The result? Inaccurate pricing models were found to contribute to $1 billion a year in missed revenue.
How Can Organisations Defend Against Fake Data?
Organisations must be vigilant on three fronts when it comes to data: provenance, context and integrity:
Tackling the problem begins with provenance. Business must take steps to ensure the accuracy or truthfulness of the data driving decisions. There needs to be a process in place that verifies the history of data from its origin throughout its life cycle.
For example, if you buy data from a third-party source, it is the responsibility of the business to verify the data before it is used for its own purposes. Or if a business collects its own data, it must check to make sure it is not inadvertently incentivising data manipulation. For example, consumers might trick algorithms unknowingly while trying to protect their privacy online.
Secondly, businesses must be thorough in making sure data is only applied within the context that it was captured. The United Airlines example is a case in point of data being used in the wrong context and time. Something that is true in one context may be misleading in another. Also, Data has a lifespan and if used outside of that window could lead to the wrong decision. Data collected in the past may not necessarily be reflective of current trends and behaviours.
Thirdly, companies need to maintain the integrity of their data by ensuring it is held securely. Cybersecurity is evidently not a new issue, but the threat is becoming more sophisticated. Hackers today are not just copying data, sometimes they are deliberately changing it.
If this scenario was to take place on banking or medical systems, total chaos would ensue. Businesses need to make sure they’re vigilant about how sources might be manipulated, providing traceability so it can be proven at any stage. Here, new technologies like blockchain could help manage transactional history. But, it won’t act as the quick fix for data’s provenance or context, like many hope it will be.
Hiring a chief data officer or experienced data scientist can help on all three fronts. But companies can make a start now by adapting existing roles across cybersecurity and data analyst teams to build a data intelligence practice.
Looking to the Future
Left unchecked, fake data could become an existential threat for business. Decision-making will become questionable at best and corrupted at worst.
As AI starts to take more real-time life-altering decisions, in the form of driverless cars and medical diagnoses, this issue will move beyond the boardroom and into our everyday lives. There is a huge prize to be won for those organisations that can harness the considerable power of data and trustworthy autonomous decision-making – soon no business will be able to grow and compete on a global scale without this – but only if they take steps to ensure their data is beyond reproach.