Mobile Posse is a fast-growing company bridging the gap between two demanding worlds: carriers and advertisers. Mobile Posse provides next generation advertising and mobile customer relationship management (mCRM) solutions for the active home screen on mobile devices, helping advertisers, content providers and wireless carriers to reach consumers. The company offers consumer behavior insights to carriers based on their user profiles, psychographic information, behaviors, trends and feedback. This allows Tier 1 carriers in the US to offer customized content and promotions that enhance customer satisfaction. These carriers include Verizon Wireless, MetroPCS, U.S. Cellular and Cricket Communications.
To offer insights to carriers, Mobile Posse must be able to collect data that will help answer questions like:
- “Is there a certain time of day that X user often drops calls? What time of day does he want to use his phone the most?”
- “Why do users on Y handset accept the Mobile Posse solution more than those on Z handset? What causes certain handsets to have higher acceptance in terms of their interaction with the Mobile Posse service?”
- “How many people opt in the first time they see a message versus four hours later?”
With 13.5 million users on different handsets, Mobile Posse must understand details such as the length of time each message remains on each device and what logistically allows people to opt into its service. The company must process millions of transactions very quickly while supporting exponential data growth. All of that data needs to be manipulated with great flexibility, and it must be up to date so Mobile Posse can make optimal decisions regarding the relevancy of content that is presented to each user.
Hadoop supports this data collection, processing and analysis with greater flexibility, scale, and reliability than any another technology. Mobile Posse trusts Cloudera to ensure its Hadoop environment runs consistently and effectively, using Sqoop to pull data from the relational database into Cloudera’s Distribution Including Apache Hadoop (CDH). Flume is used to port data to MySQL, and Oozie is used for scheduled workflows. The company is training users to use Hue and Beeswax so they can run reports directly on CDH.
Business Challenges Before Hadoop
Mobile Posse’s analytics engine originally ran on Microsoft SQL Server, but it became too expensive from a licensing perspective. The company migrated to MySQL, which offered a more cost-efficient approach, but faced difficulties running reports while simultaneously loading data. Each night, Mobile Posse “checks in” that day’s data from 13 million phones and loads them with the next day’s advertising input. Handling multi-directional data loads is a balancing act that requires a reliable, high performance platform. If anything were to interrupt this check-in process, Mobile Posse would not be able to offer the most relevant content to users based on their most recent activity, which would directly impact subscriptions and revenues.
Because of this, analysts weren’t allowed to access the database until mid afternoon; it would interfere with data processing.When reporting was made available, all the analysts would hit the system concurrently and “lock up the database due to all the simultaneous queries,” according to Scott Totman, Mobile Posse’s CTO.
Mobile Posse’s reports are processing intensive, and the availability of those reports had become a major issue. Month-end campaign reports were especially time-consuming and inflexible.
Further, Mobile Posse’s business model required a platform that could scale horizontally (as opposed to vertically, which is where relational databases accel) – Mobile Posse needed to understand unique users across unique handsets across carriers. Anytime users wanted to look at the data in a new way or run a new analytic procedure, they had to re-index or create a new table.
When evaluating alternative solutions, a Mobile Posse architect saw Cloudera’s online videos explaining the value of Hadoop, an open-source project consisting of reliable data storage using the Hadoop Distributed File System (HDFS) and high-performance parallel data processing using a technique called MapReduce. Because Hadoop uses commodity hardware, the cost per terabyte of storage is, on average, 10 times cheaper than a traditional relational data warehouse system.
The benefits of Hadoop were clear: it offers reliable, cost-effective data storage and high performance parallel processing of multi-structured data at petabyte scale. But the quickly evolving platform and tools designed to support and enable it are complicated and can be difficult to deploy in production. Cloudera’s Distribution Including Apache Hadoop (CDH) simplifies this process by bundling the most popular projects related to Hadoop in a single, integrated package with steady and reliable releases. Mobile Posse decided to deploy CDH with support from Cloudera and was up and running in a few days. As Totman noted, “Cloudera provided the right way to deploy Hadoop the first time. They saved us several weeks of development effort.”
“Hadoop is the right solution for Mobile Posse because it’s built to scale naturally the way we needed to while maintaining flexibility,” said Totman. “And when I think of enterprise Hadoop solutions, I view Cloudera as the leader.” Through effective collection and understanding of the data Mobile Posse collects, the company has driven double digit engagement growth. Mobile Posse understands its users and doesn’t bombard them with irrelevant messages. Users opt in to send machine-generated data back to Mobile Posse and their carrier, allowing Mobile Posse to run extensive analytics on the data to continuously improve content, and enabling carriers to analyze the data for network improvements. If usage of a particular handset has changed dramatically, Mobile Posse can make adjustments on the fly. The company can identify problems that occur during a campaign and make necessary adjustments before the campaign completes. Analysts can forecast advertising performance to accurately gauge impressions and revenues. The end result: a better experience for the consumer and higher revenues for Mobile Posse.
Hadoop has also simplified and enhanced Mobile Posse’s reporting capabilities. Users can access the database 24x7, regardless of whether data loads are in process. Mobile Posse has been able to cut out seven-eighths of the time and effort required to report on campaigns at month end. And Mobile Posse’s reports are more flexible today; users can filter by audience, phone, usage pattern and more.
With CDH, Mobile Posse is positioned for continuous growth by leveraging big data to successfully serve the very demanding telecommunications and advertising markets.