Predictive ML in the World of Privacy

Author: Ivaylo Bahtchevanov

Navigating the New Digital Landscape

In our digital world, privacy is always top of mind for consumers, vendors, and regulators. The digital landscape is constantly changing with respect to consumer data as policies like GDPR enforce limited tracking, and even more so with the Apple iOS 14 update implementing ATT, which enforces new data sharing policies. Third-party tracking dropped from 70% to as low as 10%, giving digital marketers significantly less data to work with.

Marketers and advertisers everywhere are scrambling to navigate the new rules, which have had a severe impact – on average conversions are down significantly while CAC is up 30-50% in the past year alone, largely due to the latest privacy changes. They are facing an uphill battle given the cookie deprecation and limited availability of customer data. Meanwhile brands are pulling back on their digital spend due to constrained budgets, rising inflation, and the market moving away from the growth-at-all-costs model. It is also now harder to identify your ideal target customer given limited data points on their previous purchases and their online activity. With the cost of advertising to users significantly higher, brands are shifting resources to find new ways of engaging users.

With the new changes in place, it’s become crucial to lean into channels that will enable you to own the data collection initiatives and the relationships with your users. Online businesses need to rely primarily on first-party data for customer insights in order to identify target customers, and then personalize and enhance the user journey. It is more important than ever to engage with your customers in ways that uniquely resonate with their preferences and needs. 

Another key to operating in this new environment is constant experimentation – what works well for one set of users likely won’t work for another set. The ability to run many experiments quickly to identify the most optimal strategy becomes an increasingly valuable competitive advantage in the new landscape.

Kumo.ai: Powering your DTC Strategy 

Kumo.ai brings a new approach to performing ML and enables ecommerce businesses to power their growth and GTM strategies. Using state-of-the-art graph learning, Kumo learns directly from raw data, maximizes signals from the natural relational structures, and makes highly accurate predictions about future customer behavior. 

More signal from less data: Kumo needs significantly less data than traditional ML approaches to understand how customers will behave in the future. With limited information, you can quickly identify which users will bring the most value to your service so you can focus on enabling them. The platform also makes it possible to overcome the cold-start problem – when a new user joins your service, Kumo leverages the full context of the enterprise graph to make predictions without having historical data on the given user’s past behavior.

Personalize user journey: Kumo’s modeling approach is uniquely designed to identify preferences and affinities which can be used to curate the user experience and provide highly personalized outreach, serving as an extension of your core platform. You can optimize marketing and retargeting campaigns through personalized messaging that resonates with each user’s needs in a timely manner (i.e. discount code sent to users who purchased a complementary item).

Rapid experimentation powered by accurate predictions. Kumo offers a powerful abstraction called the predictive query, a no-code language that gives you the flexibility to specify any number of ML tasks at hand. By defining multiple queries, you can rapidly generate predictions and test many hypotheses on how customers will respond, with no additional overhead. 

If you’re interested in trying out the product for your marketing or growth teams, reach out to us directly.