Getting Ready for Real-Time Decisioning
There is no end in sight to the growing amounts of data in today’s digital world. By 2020, analysts project there will be 6 billion smartphones and 50 billion telematics devices as an ever-increasing number of people access the internet and devices go online to communicate in a connected world.
And companies are struggling to keep up. Whether they are tracking hits on websites, selling products online, engaging with customers or trying to prevent fraud, there’s just too much data to manage effectively — and subsequently, companies are becoming data rich and knowledge poor.
Whether you go back 10, 20 or 30 years, there has always been more data than data platforms could handle. What’s different today is that companies need to understand the temporal elements of business intelligence in order to achieve success. Today’s customers expect to receive personalized experiences and quick responses or they’ll go elsewhere. To meet these changing customer dynamics, companies need to use data in real time. In the past, they weren’t able to. But now they can…and in the future, they must.
Farm Data; Don’t Mine It
A lot of opportunity results from all that available data, but businesses struggle to figure out how to manage it.
Data mining, the concept of examining data for patterns and predictive analytics, has been around for a while. But the better approach to managing data is farming it, not mining it. The distinction is that mining depletes while farming is sustainable. This is important because data is a resource that companies can reuse, cultivate and reinvigorate. If companies are not getting a continuous return on one of their key assets, then they are at a competitive disadvantage.
Rather than simply collecting a lot of data into a data lake, and hoarding it there, companies should ask how can they convert that data into knowledge and then make it actionable. Part of today’s digital transformation involves being able to decide the next best action instantly.
For instance, when the insight gained from data farming is moved from a data lake to an edge database, algorithmic decisioning engines can apply the insight in real time to make a recommendation, prevent fraud, make a payment, or route the visitor in the best possible way.
Rise of Real-Time Decisioning
Understanding the temporal elements of business intelligence is essential to creating a strategy for extracting value from data. Actionable, “in-the-moment” insights drive better decisions that can greatly improve the customer experience, increase operational efficiency and support new business models.
For example, one of the most important aspects of AdTech is an advertiser’s ability to find the right audience at the right time. As a visitor arrives at a digital property, publishers and advertisers exchange and process massive amounts of audience segmentation data. All of this data must be stored and analyzed in such a way that it is accessible within milliseconds for decision-making. The right decision has to be made in the blink of an eye or the audience will go elsewhere.
Traditionally, it has been a slow process to filter through vast amounts of data to find a predictive behavior that can then find its way into a real-time decision engine to help make smart decisions. Trying to find patterns in petabytes of data is like looking for a needle in a haystack. Often, companies seek actionable data found in discrete parts of their business, and it’s difficult to break down those silos to bring those components together.
“Transalytics” — or the convergence of analytics and transaction systems — has been one approach used to overcome this problem. But many problems arise from trying to do real-time analytics within the same system that manages traditional transactional data.
Another method is to combine the “moment-to-moment” touches from streaming data with the more static data from historical transactions. By operating on both these datasets, a company gains closed-loop, real-time business insights.
Defining and Designing Real-Time
There are some factors businesses need to consider to determine whether investing in real-time analytics will return value. First, they need to decide how to define “real time.” There is a wide range of what people would consider real time. Is what they seek actually real time or just faster than what they have now?
Consider an e-commerce company that handles a high volume of transactions. To maximize its revenues, it will need real-time analytics so it can receive those insights that enable it to quickly deliver the right offer or price in the fastest time possible, creating a truly friction-free digital commerce shopping experience.
On the other hand, if you’re tracking down illicit money laundering, you would gather banking wire transfer data from the past five to 10 years and look for patterns. After months of applying various statistical tests, you’d be able to produce suspicious information to pass along to law enforcement. Some intelligence still takes weeks and months to harvest from the data pool. There is still a place for that, but it’s not real time and not the way of the future.
Another concern companies need to think about is where in their business they could benefit from real-time analytics. Operational uses cases usually are a great fit because those are about quickly making a decision based on certain information.
A great example is e-commerce companies doing fraud prevention for online transactions. Such companies need to execute thorough, complex risk calculations in less than a second in order to avoid cart abandonment and, ultimately, loss of money to fraud. Another example is risk management in capital markets. Brokerage companies need to complete risk calculations in real time during market hours, and don’t have the luxury of performing the analysis overnight.
Every time we use the internet, the receiving website has only milliseconds to decide the next best action — to show you an ad, resume your last session, collect your data, approve your transaction, or perform a number of other actions.
About the author: John Dillon is the CEO of Aerospike, a developer of an enterprise-grade, non-relational database. He has more than 30 years of experience building high-growth technology companies, and a passion for empowering developers creating next-gen cloud applications. He previously held CEO roles at Engine Yard, Hyperion Solutions, Navis and Salesforce.com. Earlier in his career, he was an engineer at Electronic Data Systems and held sales management roles at Oracle as the relational database management company grew from $100M to $1B.