Opening Cross: The Power of Predictive Analytics to Simplify Content Complexity

Aside from predictive text, which can be both a blessing and a curse, and the musings of dodgy telephone psychics (well, how did you think we find our stories?), predictive technologies—and in particular, predictive analytics—are one of the really exciting areas currently being explored by the market data industry.

Predictive analytics use available data to make an educated guess about what will happen next. For example, based on how prices have responded to certain events—say, earnings announcements or US employment figures—in the past, how can we reasonably assume they will react to current and future events if we know certain parameters of those events?

Until now, analytics have largely looked backwards rather than forwards, to better understand activity at one specific moment, and to place that in the context of past activity—to spot the proverbial needle in the haystack, or “find the signal in the noise across a variety of data sources from structured sources... to unstructured sources, like news and social media,” says Asif Alam, global head of machine-readable news at Thomson Reuters, in the Analytics report accompanying this week’s IMD.

Mostly, these were confined to desktop displays, finding colorful graphical ways of representing and correlating price movements with trading volumes and other factors, while black-box calculation engines crunched numbers in similar ways to produce machine-readable data inputs for algorithmic trading engines.

But the nature of analytics is changing. “Next-generation analytics applications provide insight into all of a firm’s data to help traders get at the specific opportunities that are available in the market as they occur,” says Ben Plummer, chief marketing officer and senior vice president of strategic alliances at Datawatch. “Portfolio and risk managers must have access to the same information as traders, at the same time, to understand the full context of a firm’s opportunity and risk profile.”

But next-generation analytics aren’t just about incorporating more data: they also need to incorporate new types of data that traders and investors may never have considered as inputs before—including social media and news “buzz,” as monitored by MarketPsych for its predictive tools, as well as geographical and sensor-based data, among others. And they need to go further towards predicting the impact and the “edge” to be gained from specific circumstances. For example, EidoSearch, which recently appointed data industry veteran entrepreneur Jeff Parker chairman, allows users to search current and historical price movements that follow a selected pattern, and shows what happened to each after the selected timeframe, giving users an idea of what they can expect from current patterns.

In the Analytics report, Adam Honoré cites Kensho—which uses high-speed parallelized algorithms and visual tools to help users predict the impact of global events on stock prices—as an exciting analytics provider. Kensho last week announced that it is using Nasdaq OMX’s FinQloud dedicated cloud computing environment to host its algorithms and the data it requires to run its analysis. According to a survey from tick database provider OneMarketData, interest in cloud services is increasing among the capital markets industry—a good thing, since the analytics envisioned for the future would otherwise require prohibitively expensive data infrastructures.

Recognizing the need to make data easier for developers to consume, Edgar Online has enlisted API developer Mashery to help open up its financial and disclosures content to new audiences, without requiring them to sign up for, capture and manage onerous amounts of data.

But if there’s one thing any kind of analytic must do, it must make things simpler. With data, strategies and computations becoming more complex, analytics must be the balancing factor that take this increased complexity and distil it into actionable insight. Without suitable analytics to make sense of them, many of the new datasets have no value.