Mobile advertising company InMobi grabbed headlines earlier this year when the Indian unicorn declared, after a decade in business, it turned operationally profitable in 2016. Set up in 2006, the Indian data-driven company provided much cheer to the Indian startup ecosystem plagued by takeovers. In an earlier interview with Analytics India Magazine, Srikanth Sundarrajan, the Principal Architect with Inmobi, revealed how big data and analytics formed the backbone of the mobile advertising platform.
Here, we detail how artificial intelligence, specifically machine learning played a key role in this data-centric organization, hailed as one of the best Indian unicorns. Essentially, InMobi solutions address the app developer’s needs across the entire app lifecycle, from app discovery and distribution to user acquisition, monetization, engagement, and retention, delivering real value throughout the app lifecycle, according to the the book apponomics.
The ad-tech platform that captures and uses real-time mobile analytics from about 1.5 billion unique mobile devices to identify insights like the best time of day to serve an ad, or the best format to achieve a rendered ad (not simply a served ad). In 2014, Abhishek Bapna, then Product Manager at InMobi, revealed the core of search had moved from keyword to social networks. But in an app ecosystem, neither of them are really useful to understand user engagement and life-time value. What is most important is the ability to understand their app behavior?
Human cognition in algorithms
Bapna explained how algorithms are a manifestation of how human beings think and operate. He revealed how InMobi, spent a lot of time trying to understand how human cognition works. Citing some key points, he shared how the app ecosystem works differently from the web ecosystem and what acts as a trigger in the app world. So, how does one drive user acquisition or user engagement and give long term value in apps?
- The ability to understand users in their ad behaviour is most important in the app ecosystem and when it comes to ad behaviour, it is not always driven by intent.
- To understand why users download and behave in a certain way and for that you have to go from decision to downloading to first action and uncover what triggers those nuances that leads to actions.
- Big differentiator from the web world where everything was HTML-based and the whole web was seamless thing but in the app ecosystem, there are all barriers to taking actions because it is far more interactive and engaging.
- At InMobi, understanding algorithms are a manifestation of how humans think and we spend a lot of time to understand how human cognition works. Our algorithms are trying to mimic how human cognition behaves.
Why machine learning in marketing trumps traditional marketing
In a recent article by Preetham V V is VP Data Sciences, Deep Learning at InMobi, machine learning, enables brands to have a deliver a targeted, coherent approach in “engaging consumers with a consistent voice, tailored to individuals across omnichannel end-points.” As compared to traditional marketing, machine learning can be leveraged to improve channel efficacy by a minimum of 80 percent and unbounded maximums if done right. Even simple deep learning and machine learning techniques on KPIs like click-through-rates, conversion rates and bid-to-win ratios can bump the efficacy by 50 percent to 200 percent. Preetham further added,that machine learning models thrive on data and dimensions about the user go beyond demography. Dimensions used are users marketing channel affinity, past attributes, view-throughs, the frequency of visit per channel, average opportunity to see, etc. can all be computed through other machine learning techniques.
Can data science enhance creatives?
According to Rajiv Bhat, Senior Vice President of Data Sciences and Marketplace at InMobi, who is responsible for all things related to machine learning, when it comes to devising advertising campaigns, analyzing consumer behaviour should come first. In his article, Bhat explains, how data should trump creative. He cites that marketers should deploy “techniques from data science, statistics and artificial intelligence to analyze structured and unstructured data, discover patterns and relationships to make predictions about future outcomes and events.”
Driving a data-driven strategy allows companies to take advantage of real-time engagement opportunities and cash in on breaking trends. “Creatives can have different half-lives; understanding the longevity of creatives and making that part of the creative strategy is key,” he shared.
Try deep learning using MATLAB