This is the second article of the two-part series of conversations with Aditya Birla’s Group Data and Analytics cell (GDNA) — the analytics and AI arm of India’s mightiest multinational company which has strong roots in cement, aluminium and copper, carbon black and other industries like fashion and textiles, chemicals, telecom and financial services. In our previous article, we looked at how 2017 was an inflection point for the multi-billion dollar corporation when it adopted an inside-out approach to digital transformation — modernising the core system and architecting their business for change by tapping into data, artificial intelligence and digital analytics.
Now, we turn the spotlight on Bengaluru-based 50-strong digital analytics team — that is building a future-ready digital enterprise. The Aditya Birla Group operates across multiple sectors and amid its diversity — it’s the GDNA cell that is providing a clear route to value creation by building strategic digital assets which are delivering significantly higher returns.
The GDNA team helmed by Deep Thomas, Chief Data & Analytics Officer follows a centralised operating model — a common talent pool leveraged across group companies that works across multiple business segments, both B2B and B2C. By developing existing capabilities as well as strengthening core capabilities, the prolific analytics team is helping the parent group to succeed in the next decade and putting the parent group in the growth mode.
The team has built a strong data analytics platform which can support day-to-day analytics decisions. The platform “Nucleus” built on AWS cloud is governed by a strict security and governance framework and delivers more agility to stakeholders across the group. The build vs buy decision proved to be a game-changer with the in-built platform giving business heads the security and comfort of their own on-premise data centre.
Analytics India Magazine spoke to Purnesh Gali who’s leading analytics initiatives at Hindalco to understand why the GDNA team eschewed solutions from big data vendors like Cloudera and Hortonworks to build the complete analytics ecosystem from scratch. “This was 2017 and the concept of the cloud was not very well-understood. A couple of years back when the industry talked mostly about on-premise solutions, the GDNA team knew the future was in the cloud. That’s how we decided it’s best to build a hybrid cloud infrastructure. Now we are thinking of taking a multi-cloud approach,” shared Gali.
Building A Strong Business Case For AI and Digital Analytics
So, how does one actually make a case for AI and digital analytics at a conglomerate with a strong brand identity in both B2B and B2C businesses? Hindalco Industries and UltraTech Cement occupy a significant share in the conglomerate’s overall portfolio and that’s where the GDNA team started. “The process started by identifying top priorities for these two business groups, understanding the problems and requirements from business heads and then work on specific proof of concept (POC) projects that can yield high-return and high-growth opportunities,” said Nitin Sareen, SVP – Head of Analytics Solutions & Delivery.
Sareen shared a few advanced analytics use cases around improving quality and assurance in ABG’s manufacturing and asset-heavy industries. “At Hindalco, we started with the quality of finished goods. The company produces a variety of products like extrusions, rolled products, foil, etc which if they don’t meet the quality standard will come back as a defect. For these products we looked at surface defects and breakage” said Sareen.
In heavy industries, product quality is measured in terms of defects and the group is already seeing a 4-5% reduction in defects. “In the context of manufacturing and metals industry, optimising and reducing these defects can result in significant improvement and these numbers are big enough to move the needle,” said Gali, who’s leading the metals initiatives.
The GDNA team is working very closely with the Chief Digital Officer at Hindalco Industries in their entire journey of digitisation, driving analytics adoption and building capability. “We started this journey about one year ago and our goal is to make every person on the shop floor operating the machine use the data to make his/her day-to-day decisions,” added Gali.
Another big-ticket project is freight optimisation for UltraTech Cement. “The idea was how to reduce logistics cost for finished goods which in this case is cement. Cement travel in terms of volume-weighted — the longer the route the more you pay. The two factors resulting in an increase in cost are the overall distance and the volume of the unit. The team worked on identifying the PTPA – per tonne per kilometre rate better known as freight rate to bring down the costs,” shared Sareen. In addition to finding the best freight rate, the team optimally derived the best network so that the overall volume-weighted distances are minimised without hampering the network distribution.
“The team is deeply engaged with UltraTech cement to deliver high impact results across the front-end and back-end processes to optimize various business KPIs including market share, logistics cost and operational efficiency” shared Bhargab Dutta, Practice Head- Manufacturing and Supply Chain Analytics, Aditya Birla Group.
Sayani Nag, Head of Analytics – Retail Value Chain at Aditya Birla Group tells us it’s not always about inventing the coolest technologies. “While AI and cognitive technologies are disruptive forces, it is also important to have a deep understanding of business to create meaningful transformation,” she added. That’s why at ABG, digital analytics is supporting business strategy in two distinct ways — pursuing new business models based on AI-driven insights and secondly, making core business processes more data-driven with innovative solutions around price forecasting, sales forecasting, customer churn and promotions, logistics, predictive maintenance and automating margin decisions.
Developing A Platform With Products and Analytical Assets
Modularising AI assets for reusability has become the go-to approach for highly diverse multinationals. At ABG, the analytics team is building a pool of algorithmic assets for each business sector in the group to drive substantial growth. The team has adopted a platform approach to solutions by modularising core components that can be deployed across domains with minor customisations.
AIM spoke to Naveen Xavier, VP & Head Data & Analytics Products who gave us a lowdown on how the Products team is putting the group into the growth mode by building cutting-edge products that can be applied to the domain and scaled enterprise-wide. “The idea is to develop a framework for building solutions that can be extended to any business domain without changing its existing functionality. This process of building business agnostic capabilities starts by writing software that scales well and can be reused in different configurations to derive new outcomes,” said Xavier.
In a span of six months, the team has implemented GPU and CPU based solutions around facial recognition, computer vision and NLU. The Products team built a computer vision platform VEDA — Video Enabled Decision and Alerts “in precisely six months from the ground up” currently deployed in four plants at Hindalco and customised for the entire group. “Within a year, VEDA will be scaled to 300 manufacturing plants of Aditya Birla Group. From a manufacturing standpoint, safety is a very big concern and we want to ensure our workers and contract workers associated with the group should not face any safety-related issues,” said Xavier. VEDA delivers intangible amount of cost benefits by monitoring fire, oil spillage, crowd formation, people movement in unsafe zones and whether workers are wearing safety gear on the factory floors. The same solution is utilised for monitoring other assets, especially in remote locations. For example, drones feed from mining sites are analysed for volumetric analysis and in the case of new plant construction to track construction progress.
Up next, Xavier tells us about the facial recognition solution which can map emotions and recommend the next best action to sales representatives in 275 Pantaloons stores across India. In fashion retail stores, business heads have access to the customer’s transaction history but they lack contextual data on consumer behaviour and what actually goes into the making of purchase decisions. “There are multiple instances where customers come into the stores, try out products but don’t make any real purchase. This data is not available, so we are looking at how we can process video feeds from retail stores to capture customers walking in, which products are tried out, which part of the store is most visited,” told Xavier. The tool also performs emotion detection by recording the expression of customers, to better understand whether they are happy with the services.
The next product — an NLP-based solution implemented in call centre contextualises conversation and assists call centre agents with the next best action and respond effectively. The solution is one of the best ways to optimise Average Handling Time (AHT), improve customer satisfaction and response measurement. “The solution can be applied to analyse social media feedback, service feedback and product quality,” shared Xavier.
Driving Business Transformation With Bleeding Edge Tech
One of the strongest elements of success has been pursuing leading-edge technology at ABG. “It’s not just about ensuring a fast implementation and direct impact on both top and bottom line, we also have to ensure scalability, resilience and performance of our platforms,” said Manigandan Venkataraman, Lead Product Engineering at Aditya Birla Group.
For example, the team leverages cutting-edge AI techniques like deep learning, Bayesian machine learning, reinforcement learning, real-time anomaly detection, topological data analysis, etc. to increase shop floor productivity, ensure worker safety and deliver quality improvements, manage dynamic pricing and improve S&OP through demand forecasting which affects both top-line and bottom-line profitability, shared Sandesh Nandagawe, Practice Leader for collaborations with internal group functions and optimization expert.
For specific use cases, they use the best of breed open source technologies like R, Python and AI frameworks like TensorFlow. They are also experimenting with AI hardware like Nvidia DGX-1 to achieve new results and build new experience layers on top of existing products.
Open Innovation
In an ecosystem economy, you can’t just drive transformation alone. This is especially true for transnational giants who need to tap into a network of partners, industries, startups and academia to bring the outside-in view, co-innovate and invest in continuous re-skilling of talent. ABG setup BizLabs an in-house accelerator to catalyse a startup ecosystem that can develop solutions in the area of FinTech and Industry 4.0.
ABG is driving innovation internally and partnering with the ecosystem including academia and startups through hackathons and innovation challenges to accelerate the AI adoption journey and capture growth opportunities. GDNA cell is actively partnering with 10-15 AI/ ML startups that bring unique capabilities and solutions to the table. “We are also collaborating with cross-industry players with extensive domain knowledge that can be used to drive additional value. Our group is partnering with academic institutions across India, the US and Israel to explore new opportunities in emerging technologies,” said Shankar Sivaramakrishnan, Practice Lead, AVP – Group Data & Analytics at Aditya Birla Management Corporation.
Managing Capability Building
While India does have the richest pool of technical talent, data savvy talent is not easy to find, especially people who have analytics and domain knowledge. In the first post, we discussed how attracting the best talent with strong domain expertise across finance, consulting, e-commerce, heavy industries and retail proved to be a recipe for success at ABG.
At ABG, the expectations are more specialized and demanding. The team follows a very selective approach to hiring — on-boarding bilinguals who pack data science and domain expertise as well. “We aren’t just looking for a data scientist who has done a bunch of online courses but looking at domain expertise, especially manufacturing since 80-90% of our group business is that. So that narrows down the kind of resource we are looking for and there is a very small list out there in the market in terms of hires,” opines Gali. The team has a rich talent base both on the data science and data engineering side.
ABG is also one of the few mega giants offering professionals the opportunity to pick multi-sectoral knowledge. “We are getting people who are good with data science concepts and also enabling them to take up domain knowledge. I believe this is something very few companies can offer,” added Gali. The company conducts regular hackathons — MindSparks and MindFire internally and MindBlaze for external participants to work on specific business problems and foster a culture of innovation.
In addition to this, there are regular internal workshops aimed at capability building. “For example, we had folks from GE who have been working on advanced concepts like topological data modelling conduct a workshop internally. We also have regular workshops with AWS and Azure where they talk about efficient production,” shared Gali.
Wrapping Up
Leading multinationals embrace digital transformation by creating a centralised AI group to implement AI at scale. However, that’s just one part of the puzzle. To be a market leader, conglomerates need to combine advanced technologies like AI/ ML along with domain knowledge, build effective collaboration and partnership management capabilities with group businesses/ subsidiaries to respond effectively to changing market dynamics. In this initiative, ABG succeeded by putting together its Group Data and Analytics Cell under one leadership structure that has the agility and speed to deliver impact to group businesses and deliver fast-paced innovation.
