AI and machine learning technologies have slowly been coming of age for many years, quietly revolutionizing industries the world over, including adland.
Marketers face a huge challenge as we transition to a privacy-first, cookieless era – how do brands advertise online when they don’t have access to insights around what consumers want?
In this Q&A, Tal Shaked, Chief Machine Learning Fellow at Moloco, gives his take on the power of ML to iteratively make educated guesses, based on data points, as to what ads will work for which customers.
1. Can you briefly discuss the relationship between machine learning (ML) and marketing? How are they interrelated?
Tal: In the past, marketers created ads that appealed to a broad category of consumers – consider the ads seen on billboards, magazines, and broadcast TV during high-profile events such as the F1 Singapore Grand Prix or the SEA Games. Advertisers had limited ways to target specific groups of people and it was difficult to measure the impact of these advertisements.
Today, the combination of ML and digital advertising has changed the way advertisers can customize and enable real-time targeting of advertisements, and are far more effective in reaching the right target audience at the right time. Moreover, advertisers now have insights into users who engage with their ads and carry out further actions like app installations or purchases.
Such user segments hold immense value for advertisers, who are willing to allocate significant resources and investments to acquire and interact with them. It is this process of optimizing an advertisement’s reach that is known as performance advertising.
Operational ML is the best technology to maximize value across users, publishers, and advertisers.
2. How can companies or brands utilize ML to enhance or improve marketing/advertising efforts?
Tal: Operational ML has proven to be the best technology for leveraging first-party data to understand and gain better signals from the user for performance advertising. By using the advertiser’s own, unique first-party data, which has been collected via their direct relationship and includes specific signals on user behavior within the app, we can shorten the time to learn and train models, enabling marketers to more efficiently achieve the first return on investment in days or weeks.
In practice, there are many more objectives that advertisers optimize for, and they often change those objectives based on seasonal trends, holiday promotions, end-of-quarter pushes, or other business objectives.
3. What are the challenges that companies face when it comes to using ML in marketing?
Tal: The mobile advertising ecosystem consists of complex interactions across users, publishers, and advertisers. Developing apps, especially popular games such as Candy Crush Saga, require significant resources and publishers need to develop ways to monetize these apps. Often developers do so by advertising their apps to maximize user installations. Subsequently, they generate revenue from those users via in-app purchases and/or ads integrated within the app. Users get value from these apps and might also stumble upon other apps of interest through in-app advertising, further enhancing their experience.
There are many other players in this ecosystem that either provide services on behalf of advertisers (like Moloco) or provide complementary services on behalf of publishers to maximize the revenue they generate from the ads shown in apps. Another challenge that companies face is accounting for real-life world events such as holidays and sports which impact human behavior – it is therefore important to provide controls and capabilities to adapt campaigns accordingly.
4. Related to this, how can companies navigate the growing pains in operationalizing ML models? Also, can you briefly discuss the importance of creating models with purpose and specific use cases from the outset?
Tal: There are many nuances to operationalizing ML models – it is not a silver bullet that alone enables amazing product experiences. Companies have to build models, often from the ground up, for specific use cases, and then integrate these models into their products to create customer value.
An example of this would be our work done with Singapore-based live streaming application, Bigo Live. We partnered with them to identify valuable, active users that would support their global expansion. Moloco brought to the table access to global inventories, enabling Bigo Live to expand to over 20 geographies.
5. What trends do you see happening when it comes to ML-powered digital marketing in the Asia-Pacific region?
Tal: Asia Pacific leads the smartphone adoption and use – consider the island state of Singapore, where individuals spent more than 5 hours per day on mobile apps. And while many organizations recognize the importance of mobile marketing strategy that prioritizes user acquisition and revenue generation, we’ve seen advertisers struggle with identifying creative development and optimization, as well as budget allocation and execution.
We’ve also seen an increase in Asia Pacific and Southeast Asian businesses expanding globally and entering new markets. An example would be our work with Singapore-based live streaming platform Bigo Live.
Lastly, the Asia-Pacific region is also experiencing a growing emphasis on ethical AI and data privacy. With the implementation of regulations such as GDPR and local data protection laws, businesses are increasingly focusing on transparent data usage and ensuring that ML-powered marketing strategies adhere to stringent privacy standards.
