In the evolving landscape of digital companies experimenting with technologies, such as machine learning (ML), one significant application that has emerged is programmatic advertising. From uncovering patterns to anticipating results, ML-based programmatic advertising can transform ad campaigns by uncovering hidden patterns, predicting audience behaviour, and delivering personalised ads at scale. As per numbers provided by Custom Market Insights (CMI), a market research and advisory company, the programmatic advertising sector was valued at two billion dollars in 2022 and is anticipated to clock $30.1 billion by 2032, at a 35% compound annual growth rate (CAGR) during 2023-32. On that note, a company deploying ML applications to allow companies develop their digital strategies, on the basis of first-party data, is Moloco. As the company was incorporated in India, on May 24, 2023, it is yet to file its filings for its first full year of operations. In a conversation with BrandWagon Online, Siddharth Jhawar, country manager – India, Moloco, talks about the company’s ways of using ML for advertising solutions, with the importance of first-party data. (Edited Excerpts)
What specific ML algorithms does Moloco utilise for its advertising solutions, and how do they help optimise ad targeting and performance?
I think the majority of our models, based on AI algorithms, are similar to ChatGPT. They are built on deep neural networks, with some of the core transformer models being alike. Obviously, the use cases are completely different. Our flagship product is a user acquisition product, which is called the demand-side platform (DSP). For example, let’s say an application wants to acquire users. They would want users who carry high value and would come to spend time and money on major transactions, among others. On that note, we will train our models on that application’s first-party data, which would result in gathering all possible data types in a compliant manner. Then, we try to establish a relationship between those attributes and the required users. Once the models learn and understand what kind of user is required, we then start finding those users on the open Internet. When I say open Internet, it refers to a combination of applications which run advertisements, and have partnerships with exchanges, with us having access to these exchanges’ data. After going through the mentioned data, we start calculating the appropriate kind of user, in terms of their worth and spending capacity, among others. Post executing those calculations, we help the application get an increased amount of users at a lower cost of acquisition and higher lifetime value.
What is your plan to leverage first-party data and user behaviour signals to enhance ad personalisation? How will it be targeting audiences, specifically in the Indian market?
We believe that a platform’s first-party data is the most rich and valuable source of information. Personally, I believe first-party data is yet to be utilised in ways that it can be fruitful, with regard to monetisation, through following data compliance without relying on third-party cookies. In a sophisticated system, we help users feed all the first-party data points of a platform into an ML model. It helps the system learn and make inferences, on the basis of these parameters, which keeps improving over time on account of operational ML. This helps with real-time decision-making, with the system implementing further developments to understand the eventual outcome. If you can build that for e-commerce platforms, using their first-party data, then the advertisements of those platforms can be better, about being relevant and well targeted. As a result, users should be able to see much better quality ads, and should not get spammed by the advertisers, with merchants on e-commerce platforms being able to get a much better return on ad spend. They should be able to get more orders, and deep funnel outcomes for their ad spend, with the company’s ad-based revenue also increasing.
How do you measure the effectiveness of its advertising campaigns? What are the parameters employed for understanding the output from a specific campaign?
We believe that the success of a marketing campaign helps us understand its effectiveness, which increases the methodology’s scientificness. For instance, one thing to look about is at higher funnel event outcomes, such as targeted reach. One should look for eventual outcomes, with no better outcome for an application other than revenue. If a marketing campaign is directly resulting in measurable and incremental revenue, then the outcome of that campaign starts becoming very scientific. I believe return on ad spend, which is the revenue generated is based on the cost of marketing, is the most outcome-orientated scientific metric, and that’s how we hold ourselves accountable. This is what we suggest to our marketing partners, which’s for every dollar being spent, they should observe how much incremental revenue they’re able to generate. The reason is considered slightly technical. As one thinks about the marketing funnel, that person starts with impressions, then moves towards engagement, instals, transactional events, and ends with revenue. Now, any database system, specifically an ML one, is as good as the amount of the volume of data one feeds into it. If you feed it a million data points, it is much easier to draw an inference, and if you feed it 100 data points, it becomes much more difficult to get inferences. That’s where our differentiators come into play because of our deep neural networks, as they compute data at parallel threads. An implication of that is one needs fewer data points to decide . That’s what a sophisticated machine learning system does in terms of optimising revenue, with few other ML companies in the industry able to optimise outcomes.
How has Moloco performed in the last couple of financial years, in terms of revenue from operations and net profit? What are its financial aspirations for FY25, with regard to growth, revenue, or any other targets that you would like to shed light on?
In the last five years, we grew 100 times, in terms of our top-end products. In our 10 year journey as a company, for the first five years, we focused on building ML models, investing in research and development, with lesser focus on monetisation. The other data point is that we are able to help companies generate over one billion dollars worth of ad revenue. That number used to be $10 million dollars five years back, and it has grown 100 times. Today, a marketer can acquire global users by using our technology. We have taken our technology and put that inside the stack of streaming companies and e-commerce companies. If a streaming company wants to run its own ad tech stack, we can help power that stack using the same kind of technology, which has turned into a monetisation product. For Moloco, we believe that this monetisation procedure is the next level of growth formula. I’ve seen success, both in India and overseas, with regard to these monetisation products, because they help e-commerce and streaming companies become more profitable by generating more ad revenue, which a company makes using its own inventory constitutes 90% of its gross margin product. A few months back, we opened a global engineering office in Bangalore, and hired a senior engineer from Google, whose name is Nikhil Rao. Now, Rao’s building up a team of engineers to make products in India that can be functional, globally, as well.
About two months back, Moloco announced a multi-year strategic partnership with Viacom18 for ad serving on JioCinema. In terms of specific advantages, what has this multi-year strategic partnership brought to both Moloco and Viacom18? Secondly, can you elaborate on how Moloco’s data-driven approach to ad serving aligns with Viacom18’s objectives in delivering targeted advertising?
For your first question, I think what appealed to us about JioCinema is how they are driven, especially during the Indian Premier League (IPL) season. Last year, JioCinema served about 32 million concurrent users for IPL, and we saw similar patterns this time also. In that sense, I believe the strategic nature of this partnership helps us ‘pressure test’ our products at a high-scale. Considering that IPL is one of the largest live digital events in the world, it becomes a platform to test our products to work seamlessly at that scale. I think seamlessness is one of the problems we’re trying to solve for them, given the magnanimity of what they’re building, to ensure that systems remain stable. In terms of the second question, it comes down to a broader point about the streaming industry. Since COVID came into the picture, globally, content streaming, be it long-form or short-form, grew significantly in terms of watch time. Today, when one thinks about the advertising market, there’s brand and there’s performance, along with hybrid modes in the middle. In India, our digital ad market is worth about nine and a half billion dollars. I think most sophisticated advertisers want some kind of driving solutions but they’re increasingly spending more money on performance solutions. So, a gaming or an e-commerce application would probably allocate 20% of their marketing budgets on branding and 80% on performance, in that order of magnitude. YouTube is able to execute this task well, and is able to get all the performance drivers when it comes to video advertising. What we want to do is enable YouTube ad-like features for streaming applications of any size, for them to be able to tap newer advertising tools and clock a higher share of revenue for them and their advertisers.