When business intelligence startup Domo recently asked 301 marketers about their data needs, 83 percent of respondents stressed the importance of real-time marketing data, and only 37 percent said they could actually access the data they need in real time.
But how are those participants defining real time? Domo doesn’t know for sure. Based on the company’s experience with customers and prospects, Domo’s chief marketing officer Heather Zynczak says it’s likely they equate the concept of real time to how often they need the information, rather than thinking of real time as the moment data is generated.
There could be an ocean of difference between those definitions. Understanding that diverging interpretations of real time exist can spur marketers, their line of business colleagues and vendors to ask each other questions that ultimately get them on the same boat when discussing how fast data should flow and be acted on.
“Real time is kind of an odd term in a way,” says Jim Riesenbach, chief executive officer of the retail analytics company iInside/WirelessWERX. “It really depends on the context … Real time is only relevant to the extent it’s actionable within a meaningful and somewhat immediate time frame for the client.”
The Different Speeds of Real Time
Of course, there’s a pesky semantical snag to acknowledge: Real time data doesn’t technically exist. “You cannot be here and there exactly at the same time,” says Yves de Montcheuil, vice president of marketing at Talend, a business process integration and application vendor. “You can get close and with some programming languages, especially in the military or medical systems, you can get super close to real time, as in within a millisecond, but it will never be real real time anyway.”
Still, on the continuum of marketers’ perceptions about real time, sitting at one extreme is the real-time bidding form of programmatic ad buying. That’s the process in which advertisers bid in real time for the opportunity to show an online ad to a particular visitor based on that person’s profile.
DataXu is one company that helps brands and agencies participate in programmatic buying. Its vice president of corporate and business development, David Shapiro, explained at the Teradata 2013 Partners Conference & Expo in October that DataXu has “about 50 milliseconds to make a decision of whether we want to bid [to show an ad] and how much to pay. Close to a million times every second our system is getting asked, ‘Do you want to buy an ad?’ across all the exchanges we’re connected to.”
A close step down from the fast data delivery involved in real-time bidding are the initiatives that trigger slower but nonetheless swift actions. For instance, Mindjet, which sells software that helps teams develop ideas, collaborate and execute on those innovations, uses live data to quickly tweak some website content for individual visitors based their browsing behavior.
Likewise, when the Houston-area outlets of pizza delivery chain of Papa John’s wants to put an offer in front of local consumers just moments after they’ve tweeted something that signals an intent to buy pizza, it relies on real-time information from HipLogiq, which helps clients target and engage customers over social media. (An 8-month stretch of a Papa John’s Twitter campaign that launched in late 2012 averaged a 52 percent conversation rate.)
Business Needs, Technology Costs and Human Resources
The concept of real time can take a different hue when the task is less urgent. Even then, the same organization can have different timeliness requirements for different applications. Domo, for instance, has customers who say they need some data at one interval of time and other data at different intervals. Maybe they want to see new sales leads coming in by the minute, Zynczak says, but want daily updates on the number of new social media followers.
Zynczak herself likes to check leads hourly that come in through Domo’s targeted social advertising campaigns, while she has a member of her team check them every 15 minutes. “To have an hour [without the data] could cost us up to $100,000,” Zynczak says. Taken together, the combination of 60- and 15-minute intervals are as close to real time as necessary for her business, she says.
Riesenbach says for a retailer, real time could mean getting data this day, rather than this instant. He says iInside’s sales team and executives try to determine what clients really mean when they bring up real time. “We constantly hear people say, ‘I need this in real time and I need this immediately,’ and what we are able to do is point to the cost-benefit model for each thing,” he says. “Obviously the more immediate the data, the higher the cost.”
Moving data more quickly might require better data feeds or additional hardware like sensors or server capacity, for instance. “Our job is to figure out how do we constantly scale and make our data faster and more relevant at the lowest possible cost, and we’re always trying to do that, but no matter how you cut it, there’s always going to be a cost-benefit analysis to speed,” Riesenbach says.
The costs aren’t always tied to machines: Companies may also need to invest in human resources savvy enough to handle the open spigot of information in real time. “Do you have somebody who can put in place a model that can say, ‘OK, if I connect this dot and that dot, what do I get?’” says Mehdi Daoudi, chief executive officer of Catchpoint, an IT analytics company that sells Web performance and application monitoring tools. “So the cost is not only in technology, but also it’s a human cost to be able to understand what all these data points mean.”
Avoiding Bias with “This Just In” Data
Sometimes it’s best to use algorithms that blend historical data with the real-time data. “If you’re only looking at what’s happening in real time, you are subject to many anomalies that can happen that can throw your data out of whack on a singular basis but are not really tied to any predictable patterns,” says Riesenbach.
IInside, for instance, works with grocers that want to keep the wait times at their registers to two minutes or less. The analytics firm doesn’t just consider current queue times. Riesenbach says, “The important piece is when we aggregate the data, we’re looking at not only what’s happening in real time, but we’re integrating that with the historical data to provide projections that allow the client to staff in real time appropriately, have the correct number of registers open at any given time and even predict from time of day, day of week, when they need to have that kind of staffing.”
Likewise, some marketing managers like to see data coming in real time but purposely wait to act on it. Matt Belitsky, senior manager of marketing intelligence at Mindjet, prefers that his team wait at least 48 hours to confirm a market trend versus an anomaly from real time data. Belitsky uses an audience discovery platform from Tealium to get live data about what visitors are doing on Mindjet’s domains, like how they’re interacting with a newly published whitepaper.
Belitsky says he’s been skeptical of some results, like when data showed a number of customers were interested in reading posts related to the construction industry. “We said something must be broken because there are so many people that are absorbing this content, where is this coming from?” he says. It turned out to be legitimate—construction had become one of Mindjet’s popular verticals.
But Mindjet is a global business, and Belitsky has found there can be anomalies in terms of the popularity of content because it’s read in various time zones. Thus, the 48-hour hold on decision making. “Before we change our strategy, we need statistical confidence in our data,” he says.
“Imagine you saw that [a content topic] was really popular and you said, ‘We’re going to create a whole bunch of new content around this for the whole globe,’ and then you realize it was actually just popular in Germany,” he says. “If I can see [data] in real time, I appreciate that, but I just don’t make hasty decisions in the moment.”