Artificial intelligence (AI) is growing in importance in the digital advertising world. According to Forbes, in 2019, the majority of marketers either planned to or used AI in their audience targeting (81%) and in their audience segmentation (80%). And for good reason: AI-supported media allows advertisers to more easily determine whom to talk to, when to talk to them and what to say. Not to mention its ability to help messaging rapidly adapt as consumers’ behavior changes—which has been a particularly important feature this year.
It’s no secret that COVID-19 has turned consumer behavior on its head. According to our recent report Consumer sentiment during COVID-19, consumers’ shopping priorities have changed; they’re buying more non-perishable groceries, household cleaning supplies and frozen foods. And preferences for methods of shopping have also shifted, with home delivery, take-out and curbside pickup taking off in popularity over recent months. Consumers have also been far less brand loyal, opting to snag whatever’s budget friendly and available for stocking up.
The same research shows a clear trend toward digital. More than 40% of consumers who ordered groceries online plan to do so again, because it’s easier and they’re pleased with the timing of order processing (see graphic below).
Across Epsilon’s portfolio, we saw an increase in online sales of nearly 35%, but even that couldn’t combat overall revenue decreases. The retail industry as a whole was down 16% in April, and we’ve seen 95% drops in the airline travel industry. Restaurants have been forced to only seat consumers outside and relay more heavily on takeout/delivery options. Movie theaters have been closing, delaying release dates or forcing entertainment studios to release films on demand or through streaming services.
This sudden disruption also changed the way consumers interacted with their favorite brands, making it extremely difficult for brands to know whom to even talk to.
Combine unpredictable consumer behavior patterns with tightening, increasingly scrutinized advertising budgets and you’ll see a natural fit for AI. Backed by data-driven insights, brands can use optimized AI to deliver more relevant messages at the right place and the right time. This results in more valuable interactions with consumers and less waste of media budget.
While AI has always been able to help marketers adjust to behavior changes, it’s not as easy as it sounds when facing a pandemic. To see meaningful gains, it’s critical to work with a partner with a powerful platform who knows through experience how to adjust modeling parameters to account for recent COVID-driven consumer behavior changes. This results in optimized AI, the components of which we’ll examine below.
How optimized AI works
An optimized AI platform should power machine-learning models that are designed to react to real-time data and make decisions about the best time to deliver messages—and when to hold back. Models should be customized for each client at every stage, from audience selection to verifying ad quality. To see the most results, each client should have a model dedicated for them.
Campaign bidding becomes more intelligent and selective as models gather data, reducing wasted impressions and maximizing ROI. With optimized AI, hundreds of intelligent decisions can be made for each individual based on actions they take with your brand each day.
Read more: Email’s growth potential
Knowing just how unpredictable and rapid consumer behavior changes are these days, it’s more important than ever to be able to make quick, effective decisions based on real-time data. Relying on historical data is certainly a component in putting out successful messaging, but right now the majority of your focus should be on that real-time data interpretation.
A timely approach
As the impact of COVID-19 began to set in, the team behind Epsilon’s CORE AI platform was digesting data from the models. We were able to quickly identify a major challenge: The rapid changes in consumer behavior would make pre-COVID data less relevant, potentially resulting in diminishing targeting accuracy and a lower average of consumers converting.
With early recognition of what was happening, we were able to optimize our clients’ models to prioritize consumers’ recent behavior, increasing the importance of recent site visitation and purchases while the models were valuing consumers and learning more about them.
How machine learning fits in with your loyalty program: Epsilon's practical application of machine learning
To be clear, these models aren’t ignoring historical data (cutting up to 10 years of profile data would be a mistake); they’re simply prioritizing behavior of recently converting customers. This allows the models to adjust faster than ever and keeps our clients delivering the highest standard of personalized and relevant messaging.
Our updated models resulted in a 25% increase in client return on as spend (ROAS). Testing of the model updates began with a major retailer and was eventually deployed across all of Epsilon Digital Media portfolio once deemed successful. Needless to say, this was a win for brands during these trying times. We’re currently evaluating the models for post-COVID use (hoping that opportunity comes soon).
Are your models accounting for real-time behavior changes?
In this new normal, it’s all about quick decisions based on accurate, real-time data. To see if your AI is optimized, ask your partners how regularly they are reacting to change.
With the increased pressure to justify budgets and demonstrate performance, every second counts.