Order Out of Chaos

February 26, 2010

By Bill Virgin

TECH_iphonefull
Inrix app
Inrix’s iPhone App uses predictive analytics software to
forecast traffic information to help commuters better plan their trips.

Predictive analytics: You might not know the term but it’s a
good bet you’ve run across or been subjected to it.

The recommendations for other books, CDs or movies you might
like when you place an order with Amazon.com? Those are predictive analytics at
work. The teller who suggests an investment product the bank is pushing, while
you’re making a deposit? Predictive analytics again. The approval on your
mortgage or credit card application, and the interest rate you got? That’s
credit scoring with predictive analytics behind it.

A growing number of Seattle-area companies not only know
what predictive analytics are, but they’re also using these data analysis tools
at the core of their business propositions or writing the software used by
others.

Examples abound. Kirkland-based Inrix Inc. employs
predictive analytics to provide traffic information in 130 American and
Canadian markets, as well as 14 countries in Europe. Says company spokesman Jim
Bak: “We can tell people with a high degree of certainty not only what traffic
looks like now, but what it looks like in the next hour, in 15-minute
increments.” For iPhone applications, Inrix offers those 15-minute predictions
for up to six hours in advance, and can do more general forecasts for the
coming year.

Farecast, now owned by Microsoft, applies predictive
analytics to tell travelers the optimum time to buy an airline ticket. Varolii
Corp. in 2008 unveiled a predictive analytics tool to help customers
“understand, analyze and strategically target more effective customer outreach
for collections, customer service and loyalty programs.” Bellevue’s DS-IQ Inc.
runs predictive analytics to help retailers design in-store digital marketing
campaigns.

ValueAppeal uses predictive analytics tools to tell
customers whether their property assessment is too high. Insightful Corp., a
Seattle-based developer of data analytic tools, was a publicly traded company
until being acquired by Tibco Software Inc. in 2008.

The Department of Energy’s Pacific Northwest National
Laboratory at Richland has a “technosocial predictive analytics initiative” to
study whether such an analysis can be applied to the effects of global climate
change on homeland security and defense, or guiding counter-terrorism efforts.

If all this activity sounds like data mining, it is-but it’s
also much more. “The difference is data mining is about taking mass volumes of
data and looking for trends,” says Craig Chapman, Inrix’s chief technology
officer. “Predictive analytics is taking that data and creating models from it
so we can predict” what will happen with a borrower’s likelihood to repay, what
a retail customer might buy based on past purchases or, in Inrix’s case, how
bad the traffic jam is.

“Predictive analytics takes a lot of the techniques you find
in statistics, data mining and game theory to analyze current and historical
facts, and make predictions about the future,” adds Paul Freed, managing
partner at Seattle-based executive recruitment firm Herd Freed Hartz, which has
handled multiple requests to find candidates for predictive analytics
positions.

“Data mining does an excellent job of organizing data to get
an idea of what happened in the past and what’s happening now,” but predictive
analytics is much more sophisticated than simply extending those historical
trend lines into the future.

Inrix, for example, takes road-sensor data, weather, event
schedules (in the other Washington, you’d want to know whether Congress was in
session, for example), information on road repairs, school schedules, and GPS
data from delivery truck fleets and taxis, and other iPhone and Google
applications.

That’s a lot of data to wade through, some more valuable,
others more misleading. “A car sitting at a stoplight doesn’t mean a car stuck
in traffic,” Bak explains. A road sensor can be out of commission or give bad
readings.

The trick is to translate that wealth of data into a
meaningful and accurate prediction. Anyone can tell you where traffic snarls
typically occur. Inrix hopes to tell you how much worse than normal those snarls
are going to be.

ValueAppeal starts with property value assessments from
county governments, then combines that data with recent sales information of
comparable property. Beyond that, it incorporates non-statistical criteria that
can help determine whether or not a property is overassessed. Does the home
have a view? A view of what? How good a view? What’s the distance of those
comparable-property sales from the home in question?

Coming up with mathematical formulations to predict traffic,
airline fares or whether a property-value appeal is likely to succeed is just
the start. “We consider it a learning algorithm,” says Charlie Walsh, founder
and chief executive of ValueAppeal. “The more we do it, the better we get at
it.”

“We try different features; we might find a feature doesn’t
have that much effect on the model,” adds Inrix’s Chapman.

“Predictive analytics has always been behind the scenes,”
Freed notes, but it wasn’t until five years or so ago that he picked up on the
notion of predictive analytics as its own industry with companies built around
it.

It took the convergence of huge data sets plus powerful
mathematics to make sense of them, to create that industry, he adds, and the
application of those concepts by local companies such as aQuantive (now also
owned by Microsoft) in analyzing consumer behavior online. “Terabytes of data
were being collected,” Freed says. “The problem was we had these large data
sets, but there wasn’t money to be made” off the data itself.

The real value was in the “actionable insights to be grabbed
from the data” to be gained from the predictions of consumer behavior. Says
Freed, “One insight can make a company.”

Now that businesses have grasped the concept and have the
tools to make it work, predictive analytics is a hot field, with Seattle in the
middle of it. Existing firms will use predictive analytics in fields like
health care. New companies are coming.

“Seattle’s uniquely in a great spot for predictive analytics
startups,” Freed says. “You have big pools of talent” from companies including
Microsoft, Google, Amazon, Yahoo and Washington Mutual. “We have a good history of successful
companies [that have created] a positive spiral. When you have successful
companies, they attract bright people, who attract new companies. … When I’m
doing a nationwide search, the sell is all the other companies in the space.”

Predictive analytics is “red hot in recruiting,” Freed adds,
which makes it a rare bright spot in the current economy. That’s good news for
those with the skills and the educational background companies are looking for.
Says Freed, “If you have a predictive analytics Ph.D. or a machine learning
background, you can write your own ticket.”