How to use AI and machine learning to personalize and optimize campaigns
Marketers can leverage this technology, but first they need to centralize their data.
AI is revolutionizing how marketers engage customers. Beyond how a chatbot like ChatGPT might change the way customers search, AI and machine learning models can also equip marketers with the power to personalize and optimize their messages to customers.
Automation and optimization for personalized messages
“[Automation and optimization] are two broad areas that marketers leverage machine learning for,” said Alex Holub, head of machine learning at customer data platform (CDP) company mParticle, at The MarTech Conference. Holub’s AI startup Vidora was acquired by mParticle in 2022.
First, marketers can use this technology to automate a process like the generation of emails or the scheduling of when those emails go out to customers,” Holub said.
Secondly, AI and machine learning can be used to determine the best time to send the message or the best message that can be sent. This kind of optimization draws on large customer datasets automatically, instead of having data teams sift through the data and ask questions themselves.
From heuristics to optimization
Holub described a fast fashion company he worked with that replaced an older heuristic method in their email campaigns with a new machine learning optimization strategy that generated 90% more revenue.
The brand was sending weekly emails to millions of engaged customers, so they wanted to be able to pick the best product to promote to each user. The solution used personalization and automation to deliver these messages at scale.
“Prior to leveraging machine learning, they were leveraging heuristics — so they had analysts go in, look at their data and try to determine for different segments of users who should receive which promotion for which product,” Holub explained.
Using the heuristic approach, data scientists looked at past purchases to determine what messages to send. The machine learning approach could not only analyze more data quicker, but it could determine the right data to look at.
“The great thing about machine learning is that it will figure out what behaviors are the best behaviors to use in order to determine who should receive which email,” he said.
Centralizing and activating data
In order to implement AI and machine learning into a company’s messaging strategy, marketers must first make sure their customer data is centralized.
“Getting all their data in one location in a high-quality manner, that’s typically a huge challenge for businesses,” said Holub. This is why CDP technology like mParticle, Oracle and others goes hand-in-hand with AI.
Dig deeper: Oracle launches industry-specific AI models for its Unity CDP
When customer data is centralized within an organization, the next biggest challenge is for the business to be able to activate that data through the right channels and messages to customers.
“The second challenge is activating the outputs of machine learning,” said Holub. “So, if you build a machine learning model and you’re saying, ‘Hey, I should engage these particular folks with this message,’ but you’re not able to activate that machine learning model, you’re not able to activate those particular messages.”
He added, “So, typically, it’s almost always the input and the output of the machine learning that are the biggest challenges.”
Register for The MarTech Conference here.
Our new report, “Customer Data Platforms: A Marketer’s Guide” is now available for free download.
Get MarTech! Daily. Free. In your inbox.