Big Data has been the buzzword for the last couple of years. The potential insight that can be gained from huge unstructured data sets processed through cloud-based open source analytics tools will fundamentally change the way enterprises do business.
Or so says the hype. In the real world, though, companies aren’t tumbling over each other in a rush to start Big Data initiatives. In a recent online poll of IT World Canada readers, half said their companies weren’t buying into the Big Data hype. Interestingly, though, a quarter said they were on the fence until they saw a relevant business case.
That’s how many companies are drawing benefits from analytics, regardless of the size of the data set, right now: by applying the technology to specific, real-world business use cases, some of them problems as old as business itself.
When fans think of the enterprise that is a National Hockey League team, they see the operation in terms of the product on the ice: 25 athletes in the rough-and-tumble of the world’s fastest sport. They don’t often spare a thought for the infrastructure that puts that team on the ice, or spectators in arena seats.
Most NHL franchises and venues have a similar staffing structure, according to Bill Nowicki, vice-president of ticketing for the Carolina Hurricanes and PNC Arena.
“The Carolina Hurricanes run at about 150 full-time employees, which probably doesn’t sound like a lot,” Nowicki says. “But, in all honesty, we’re run by our part-timers.”
“We run probably somewhere in the neighbourhood of 1,500 part-time employees, so it’s almost 10 times our full-time staff, and that’s what really keeps the building ticking for us.”
Where franchises tend to differ is in the staffing of ticketing operations: group sales, corporate sales, single event sales, all will be run differently in a year-round, high-profile venue like Madison Square Garden or Toronto’s Air Canada Centre than at a single-use venue like Raleigh’s PNC.
“Everyone has their own philosophy,” Nowicki says.
The ticketing operations group for the Hurricanes boasts many years of experience in the sports marketing game. When the NHL releases its schedule in the summer before the hockey season, ticketing staff would pore over the schedule, knowing from gut-level experience which games would sell themselves, and which would require some promotional muscle.
“When we got that schedule, I could identify a handful of games where, no matter what we did, we’d struggle on those games,” Nowicki says.
The ‘Canes and PNC started applying a more scientific approach after changing the venue’s ticketing system in 2004. The old Ticket master system was event-centric; it could identify who went to an event, but not what other events a ticketholder had attended. The new database was customer-centric, allowing staff to track attendance histories of individual ticketholders.
PNC brought in Ideas Revenue Solutions, now a subsidiary of SAS Institute, to determine what thy could make of the historical data. “We started to look at what information was important with regards to optimizing our pricing” Nowicki says.
The eye-opener for Nowicki was that the ticketing group was sometimes throwing too much promotion at weaker games, cannibalizing traffic from stronger performing games.
“We were so concerned that that game needed all our help, we put all of our attention to it, and really, it became the deal-of-the-day type of game, where people are coming into that Tuesday night game with Nashville going, ‘I know I’m gonna get a discount, what am I gonna get?’ as opposed to those with the expectation of paying our standard price,” Nowicki says.
Mining that data has allowed the ticketing group to draw up profiles of their various planholders, and rate prospects against those profiles. The Hurricanes offer season tickets, half-season tickets, and three different 12-game group plans. The key was determining what data fields were relevant.
“How far out you live from the arena, is that important? If we’ve got fans who live more than 50 miles out, is that indicative of whether they’d by a plan or ticket?” says Nowicki. “If your median income is $50,000 or less, but you live within 30 miles of the building, are you right for a certain plan? Are you always going to be a casual ticket buyer? That was the second portion of what the analytics started to do.”
Monster Worldwide is a household name in online recruiting, working in 55 countries. The company more or less invented the online recruiting industry in 1998, and Jean-Paul Isson likes to boast of its pioneering online commerce history — it was the 254th dot-com company, he says.
With 400 employees worldwide, Monster posted $900 million in revenues last year. But it faces the same expensive challenge that most services companies face: customer retention. As the company’s global vice-president of business ntelligence and predictive analytics, Isson has his team spending a lot of time focused on the problem of customer churn.
“As in any business, retention is key, so we provide the company with some insight, a predictive model of what companies we should retain, and why,” Isson says.
Isson’s team built a retention model to target customers who are a risk to leave Monster, and how to keep them in the fold.
“We score the entire universe of our customers, and provide that score of churn risk to our sales force into the CRM system,” Isson says. The analysis provides a likelihood of churn. If a company’s score is 0.5 per cent, for example, they’re not likely to churn. To simplify, customers are scored high-, medium- and low-risk for or the sales staff.
“We did that for the U.S. and Canada, and the result of that was really very, very impressive,” Isson says. Retention increased 15 per cent in North America. Sales and marketing also had double-digit increases in productivity because they knew what at-risk customers to go after for early renewals and to address risk factors.
A 15 per cent retention increase on a $900-million top line has a huge bottom-line impact, Isson says. “That retention model has really shown the power and the value of analytics to the business,” he says.
Isson’s team also turns its predictive analysis tools on customer content. An optimization tool benchmarks the performance of customer job listings to determine how they stack up against competitors listings, and how they can improve listings to match the benchmark. If, for example, company is looking for a data scientist with particular education and experience criteria and is getting 300 views and 10 applications, while similar postings get 450 views and 50 applications, special customer service team will work with premium customers to determine what features and formating will improve performance.
“The customers love it because there is no sales pitch behind it,” Isson says. “It’s all about customer satisfaction.”
In the end, that’s a retention tool, as well. And retention is an opportunity to upsell. Isson says the retention program has increased wallet share by 27 per cent.
Vancouver’s police department is responsible for a metropolitan area with a population of about 2.3 million people and a major Pacific coast port that serves as a gateway for narcotics traffic from Asia and Vladivostock, Russia. Its 1,500 sworn officers are supported by 400 civilian employees, 25 of them crime analysts who pore over data from police departments province-wide to detect criminal and geographic trends, and tease out leads and profiles that help solve crimes.
It’s a field that’s gaining traction in a traditional conservative profession. And it has driven results, says Special Constable Ryan Prox, Vancouver Police Department’s analytics standards advisor and an adjuct professor of criminology at Simon Fraser University.
For example, VPD brought an FBI profiler in to help with a series of methodical sex assaults on children. The violence of the assaults was escalating; the profiler worried there would soon be a body count. An 18-month investigation targeted 561 sex offenders who fit some elements of the profile, but got police no closer to the culprit. The analytics team managed to isolate a suspect within seven weeks. The DNA match was conclusive.
VPD’s intelligence led policing model is based on the CompStat (short for “comparative statistics”) accountability process first developed by the New York Police Department and replicated at major police departments across North America. While the CompStat model used in the U.S. can be a highly confrontational process, Prox refers to VPD’s implementation as “CompStat Lite,” a more collaborative process to help units pool resources and develop interdiction strategies.
It’s a recent development. VPD’s shift to intelligence-led policing five years ago was driven primarily by two factors: the impending Vancouver Winter Olympics in 2010, and the fallout from the case of serial murderer Robert Pickton, charged in the murder of 26 Vancouver-area women a decade ago.
One missing element in the Pickton investigation that allowed his to escape identification and capture was an inability to communicate data across jurisdictions. A common front end for records management, Versadex, was mandated in the overhaul, and the billions of province-wide records are pulled into a single data warehouse. Desktop-level Access databases were eliminated. The VPD addressed a culture of “coveting of information,” Prox says.
“It breeds a lot of transparency,” Prox says. “The technology forces people to share.”
All the police departments within the province can access the data warehouse through a Citrix virtual infrastructure. Data can be distilled down to the minutiae that can be the missing piece in solving a crime.
The analytic success has also bred greater acceptance of the role of technology in the investigative process. Before, the attitude was, “What’s this wizardry stuff,” Prox says. Now, if the primary consideration fr staffing an investigative team is who the lead detective will be, the secondary one is which analyst will work with the team.
“Those two components can make or break an investigation,” Prox says.
DETERMINING BEST PRACTICES
The 18 staff at the national headquarters of Big Brothers Big Sisters of Canada support about 1,200 staff at 118 offices of the organization, which pairs at-risk youth with mentors. Those staff, in turn, support anywhere from 10,000 to 20,000 volunteers paired with 40,000 kids.
Malcolm McAuley, Dynamics system manager for the charity, says the organization pulls data from its Microsoft-based customer relationship management system into an analytics system to understand trends in service.
“That information helps fed our marketing initiatives,” he says.
Many of the key performance indicators on Big Brothers Big Sisters scorecard are built around the heart of the organization: the volunteers who mentor youths, and the youths they mentor. The organization measures how long matches last, the percentage that last beyond six months, and the reasons matches fall apart.
He also measures how efficient the offices are at the intake process; how long does it take from application to matching a volunteer with a youth.
“Some of our agencies process hundreds of applications some only handfuls,” he says.
A soon-to-be-released report based on the analysis of the CRM data will show that agencies that focus on the elements of customer service have better results in terms of returning volunteers, McAuley says. That will allow the organization to focus develop best practices that can be shared across offices.
Big Data Opens the Door for Prescriptive Analytics
Making customer-level decisions that balance risk and profit just keeps getting harder. And when you think you have it right, turning them into actions can be even trickier. You also need to consider the factors that make smart decisions difficult. Big data. Regulations. Customers who want an offer, fast, or else you’re going to lose them. No doubt some of these challenges sound familiar. And this is where prescriptive analytics represents the next step in the analytic journey.