Advertising being the backbone of businesses throughout the world today, is incidentally also the biggest source of data generation. Most businesses need action oriented insights out of their mobile data. Whilst in the past, data was lacking or messy; today, a lot of work is based on self-learning systems and frameworks for capturing and analyzing data. Since, smartphones generate 10 times more data than desktops, the need for insights that lead to action in real time have increased dramatically.
Headquartered in Bangalore (India), InMobi, the largest discovery platform in the world of mobile advertising, stands tall with a backbone of Big Data and Analytics; and analytics is undoubtedly, contributing profusely towards its success.
Analytics at InMobi:
In a chitchat with InMobi, we got to know that InMobi is as much a data driven organisation as it is data hungry. Everyone from managers to engineers looks at metrics in every possible way, in every decision that is made.
Srikanth Sundarrajan, the Principal Architect with Inmobi, tells us, “Relevance is one key factor where data plays a very important role. We use a lot of models to predict how our users are going to respond to a particular proposition. Another area we use analytics is, the feedback system to optimize the performance of various functions.”
InMobi works with ‘hundreds of thousands’ of events every second. They use analytics to identify how to create value for consumers, which propositions should be served to generate highest value for the consumer in real time. Further, all this is inferred and answered while learning about the customer through every interaction, thus, they work with analyzing data in real time.
The length & breadth of data analytics:
Srikanth highlights that, on a typical day, their user interaction data ends up at 20Tb of data and in excess of about 10billion events. Further, this data is maintained in a large ‘Hadoop warehouse’. They read up to 200Tb of data from system to compute various things, which could vary from aggregation to provide reporting view, models, historical data trends that they want to pull out, and any other data.
We decided to scoop further and found InMobi uses a combination of analytics software to manage the massive data. Primarily, they use Hadoop HDFS for storage, YARN for resource management, Apache ODE for process orchestration, Apache Falcon for data management system, Apache Hive for query execution engine, Apache LENS for unification of supporting systems, Apache STORM for streaming data platform, Apache SPARK for next gen data processing and primitive and Apache PIG for ETL pipelines. Apart from this they also use many other Apache projects, droids, rediff, cassandra, etc.
On probing about their data science department and big data teams, we got to know that even though they are two different teams, they collaborate very closely. Data science has different sets of engineers & scientists, specialized and well versed with big data engineering and deliver on the analytics front.
Moreover, employees at InMobi take a lot of pride in their thought leadership and excellence in data engineering. Srikanth says, “While we consume and utilise some of the open source ‘stacks’ heavily, we have a lot of in-house management expertise. In fact, we also have our own home-grown projects which we have incubated and taken to Apache and made into top noche projects, Apache Falcon and Apache Lens in particular are examples of that.”
The employees attribute most of their success to their leadership. No wonder, InMobi cradles an employee friendly environment in a beautiful workplace which inspires, eases and motivates the employees to put in their best!
A Sneak peek at the recruitment process:
This motivated us to dig in on their hiring process, to understand what it takes to become an “In-Mobian” and here are the facts, as fresh as bread straight out of oven!
The HR specifies that, “a candidate typically has 4-5 meetings with different sets of interviewers. It’s usually the hiring manager who goes first. Each interviewer looks at a different competency or skill in the interview. The HR Business Partner and recruiter also have specific hiring factors that they evaluate each candidate against. In 10 – 15 days, once these meetings are completed, the team debriefs; and a hiring decision is taken by the hiring manager. The hiring manager and the recruiter then engage the candidate in walking through the role, its nuances and the offer.”
Moreover, we also found that a typical interview at InMobi commences with general questions, before moving on to case-lets and situations that the candidate has encountered and how they’ve dealt with it in the past. The intent is to gauge the truth in the candidate’s story. Additionally, for case-lets that are future oriented or hypothetical, the intent is to understand the rationale, thinking and mindset. Moreover, it also gives insights on the behavioral feel of the candidate and if they are the right fit for the team”, the HR concludes. If you are aspiring to join InMobi, we sincerely hope this was a useful insight!
However, we couldn’t leave before understanding one last question, after installing data and analytics in its DNA, what is the road ahead for InMobi? To which, Srikanth, told us, “InMobi has already realized that real time streaming system is the need of the hour. Thus, they are investing and focusing on predictive analytics for smaller patches of data so that they can compute and respond faster!” Fair enough and we bid goodbye.
Try deep learning using MATLAB