Introduction of affordable data plans and smartphones has given rise to an insurmountable amount of data. As a consequence of the same, modern technologies such as machine learning are increasingly being used to provide consumers with a more personalized experience. Machine learning especially is playing an active role in various areas of digital marketing to provide users with precise content recommendations. In an interview with DataQuest, Mr Naren Nachiappan, Co-Founder & Managing Director-India, Jivox highlights the role played by machine learning in digital marketing, some trends in the space in the years to come, and also about Jivox has to offer for consumers.
Role played by machine learning in the digital marketing space
Machine learning is a critical technology used in a variety of different areas of digital marketing including creative optimization, media optimization and the generation of product and content recommendations. The area of product and content recommendations is particularly critical and is instrumental in generating significant lift in digital marketing ROI.
Modern recommendation engines use a hybrid of collaborative filtering and content based methods to increase relevance, through a combination of personalization strategies, such as predictive product recommendations with environmental and date/time based messages.
Collaborative filtering recommendations automatically predict and serve products based on preferences of other consumers within a large consumer base. It also uses Behavioral Clustering to identify clusters based on a person’s past behavior, as well as similar historical activities by other people.
Example: A male consumer that is a sports enthusiast from Mumbai and viewed products on the Sony PlayStation site is put into a cluster, a microsegmentation of shoppers
Content based recommendations automatically predict and serve products based on a consumer’s interest and similar products available from the brand. It uses Product Clustering, which are tags, categories, pricing, and similar attributes to identify and recommend additional items with similar properties.
Example: A female consumer is looking for climbing gear on the Flipkart website. Her shopping behavior feeds into the Personalization Hub and the Recommendation Engine predicts and serves ads showing categories of products she has searched or clicked on.
Skilled professionals and Machine Learning
The applications of ML are numerous. Industries including automotive with self learning cars, automated online technical support, fraud detection across a range of industries, security and a number of other applications of video surveillance technologies, etc. are some of the emerging areas with ML powered applications. With the rapid rate at which these applications have emerged, there is a worldwide deficit in the number of experienced and skilled ML professionals needed to fulfill these requirements.
While we do not have enough skilled professionals at the present time in Bangalore (or elsewhere in the world), we do have access to extraordinary skilled and motivated college graduates from top tier computer science programs including BITS, NIT and IIT campuses across the country. Our strategy is to recruit fresh graduates from these colleges and train them in the applications of ML technology. That has worked extremely well for Jivox, as we have found that the technical foundation in Computer Science algorithms and computation theory of graduates from these institutions is very sound and enables them to become productive very quickly. A significant portion of our Engineering staff is now composed of recent graduates from these institutions, and that will be our strategy going forward to fulfill our staffing requirements.
Some trends in this space in future
A 360 degree view of the customer journey across paid and owned media, including email, website and all forms of advertising is increasingly becoming an essential part of marketing strategy and investment optimization.
In the very near future, de-siloing of the media and creative worlds, and the complete integration of media optimization along with creative optimization will further enhance the ROI of digital advertising campaigns.
Finally, we expect more advanced ML technologies to be applied to campaign and strategy optimization, and we also expect user generated ML technologies to be integrated into the platform.
How Jivox is using emerging technologies like Big Data integration and Dynamic Creative Optimization (DCO) apart from Machine learning
Big Data is a core foundational technology in the Jivox IQ platform. Our platform delivers billions of impressions weekly and for each impression it gathers dozens of events including clicks, interactions with the ad unit, video events, a variety of mouse events, etc, generating petabytes of data. In addition to ad generated data, the platform also aggregates data from websites via Jivox pixels, integrates with partner data DMPs for 3rd party data and also integrates first party data from brands. All of this data is aggregated in the Jivox platform using current generation Big Data management technologies. Redis is part of the core infrastructure and allows the platform to perform rapid lookups in the big data store. This enables real-time personalization, real-time product recommendations, and optimization across the entire data stream.
With the rise of programmatic media, we have also seen the emergence of machine generated creatives, enabling the virtually instantaneous generation of thousands of creative variations tailored to specific micro-segmented audiences.
Ensuring that the right creative variation is selected for each impression delivered is a challenge given the real-time requirements and the scale at which the technology needs to operate – matching thousands of possible variations across potentially millions of microsegments, with the underlying variables defining the segments changing in real-time. This is the essence of DCO (Dynamic Creative Optimization), and it is a core feature of the Jivox IQ platform. The Jivox IQ platform goes a step further by integrating the creative choice with the media choice by providing signals to the media bidding algorithms indicating the relative value of the impression from a creative perspective. That allows for a unified media and creative optimization, which is an industry first, and provides an unprecedented degree of optimization.