MapR is the data hub for DataSong’s marketing effectiveness business, enabling the company to serve its clients better through increased efficiency, scale, flexibility, and performance.
DataSong, a marketing analytics agency based in San Francisco, provides marketing effectiveness solutions to leading omni-channel retailers, including Williams-Sonoma and Neiman Marcus. Retailers use DataSong’s comprehensive tools to measure, plan, and execute marketing more effectively, thereby growing revenue and improving marketing ROI. .
DataSong is focused on helping marketers determine the most cost-effective
marketing tactics and channels. “Many marketers use a top-down
approach. Our methodology is bottoms up. We use granular user-specific
data such as web logs and email provider logs. We analyze multiple factors,
not just the last click. It’s much more accurate,” explains DataSong CEO
“We look at every customer you’ve ever had. It’s not an average; it’s very specific to that retailer,” says Wallace. “To provide this for our customers, traditional database technology would not scale. We turned to ApacheTM Hadoop® because it’s cost-effective to scale and you can get answers back fast.
DataSong evaluated Hadoop alternatives and chose MapR. At DataSong,
the MapR Distribution for Hadoop runs on a cluster made up of 14 Dell
Dual Hex Core nodes with 64 GB of RAM.
DataSong receives regular data feeds of customer marketing information from their clients such as web logs, core metrics, email provider logs, financial transactions, etc. “When data comes in, we put it right into Hadoop. We do not do data manipulation outside of Hadoop,” Wallace says. “Our teams know Java so the majority of what we do is in Cascading. It’s been an efective tool that helps us to keep things common across everything we’re doing.”
Streamlining Workflow with NFS
MapR support of the NFS file system was a major selling point. “We needed a way to get data in and out very efciently. There wasn’t a good solution from Cloudera or Hortonworks,” Wallace says.
“NFS reduces friction in the organization. We had Java coders who had to put stuf in Hive. And NFS allowed non-developers to come in and do things. It would have been an impediment to our productivity without it,” says Wallace. “Being able to have our people be self-sufcient and get the data where it needs it to be, instead of spending hours of training on Hadoop, makes life a lot easier for everyone.”
Data Hub for their Business
DataSong uses MapR to run the central data hub for their business. “Our business is all about data. If we don’t have the data, we can’t do anything,” says Wallace. “As soon as data comes in, it goes into Hadoop. We don’t know if we’ll use the data right away but we are able to keep it as long as we want to.”
“MapR is our data store for everything,” he says. “It’s our processing engine, the main file system, the utility, the hub of everything for our business.”
Increasing Operational Effectiveness
A major benefit is the ease of use and how that unlocks the power of Hadoop across the organization. “In the past, analysts didn’t even know where data lived. Our MapR solution is making data accessible and usable to the rest of the organization,” he says.
From a management standpoint, MapR was more cost-efective than Cloudera. “Whether managing an installation, ongoing maintenance, or performance tuning, it’s very easy to manage and keep it going,” he says. “I do not have to spend time managing or training on technology, so I can spend my time focused on our customers.”
Receiving Valuable Support from MapR Community
Wallace has also been impressed with the support he’s gotten from MapR and the broader community. “There is a strong community using MapR. Whenever I have a question, there’s almost always someone who responds and explains how they solved the same problem. It’s a great community to share questions and answers,” he says.