Behind the Science of Getting Up Close and Personal
- By Winston Thomas
- November 17, 2022
We’re living in a world of hyperpersonalization.
Movies like “Minority Report” and “Blade Runner” awed us with real-time personalized suggestions and how AI algorithms may know us better than ourselves. Yet, we don’t blink twice when Netflix and Amazon suggest potential movies or books, and TikTok calibrates video feeds to our changing moods and preferences.
Data availability is part of the reason for the emergence and success of hyperpersonalization. As companies find better ways to mine and learn from the vast data troves they already have, they’re getting closer to us with the data we are providing in one way or the other.
“The core part is real-time analytics,” said Dr. Dennis Leung, SAS Hong Kong’s director of customer advisory.
At the recent 5th Chief Digital and Data Officer Hong Kong Summit 2022, Leung noted that hyperpersonalization has several pillars. “So the first thing is knowing your customer. And it doesn't cost a lot. You do a lot of analysis based on the data you already have within your company.”
Next, companies need to find out when to deliver the data-driven experience. This is about getting the context right and is especially crucial for hyperpersonalization as it is done in real-time. The wrong moment can create bad results. Sometimes, people may cry foul and call it an invasion of their privacy.
It is why Leung thinks that an automated feedback loop — the third aspect of hyperpersonalization — is equally crucial “to improve on whatever you’ve done.”
In a separate interview, Leung noted that the loop should not just stop at the customer service or marketing team. Instead, it should stretch to the product development teams to improve response times. For many companies, this can be a stretch.
Extrapolating the data axis
When it comes to hyperpersonalization, data is a double-edged sword. Because of our vast troves of data, coupled with the easy accessibility of the cloud, we’re able to personalize to individual moments.
But the uncontrolled growth of unstructured data, the lack of standardized data catalogs, and the difficulty of managing data flows from trusted data sources have many questioning whether they are opening a can of worms.
So, should companies wait to get their data house in order? Leung believes there is no need to wait; companies can untangle their data mess and do hyperpersonalization at the same time.
“One reason is, if you are looking at customer data, you are using a subset,” he said. So, we are not dealing with all the data, but only a fraction — although a significant fraction.
Instead of trying to get all the data in order, Leung believed it is essential to know what data you already have. “I think the key part is knowing your data and how you consolidate your customer data first.”
Knowing your data can take time. Leung observed that today's real hyperpersonalization successes are from companies that started their customer data consolidation “five years ago.”
Leung advised starting small for those who don’t have the luxury of such a long timeline. “The first step is to get some data, do some analytics, prove hyperpersonalization is working.”
Working with what you have is where Leung feels SAS has an edge and the reason for a string of hyperpersonalization successes.
“So why are we successful? It's not about the technology. We integrate with the existing infrastructure. So, with our solution, in our case, our decisioning engine, we can deliver the content at the right time using all the intelligence we build.”
Critical automated feedback loops
Hyperpersonalization requires a lot of work. While automated data capture can do a lot of the heavy lifting, it still involves manpower to ensure that the processes are right and aligned with customer expectations.
It’s one reason why Leung sees automation feedback loops as a prerequisite for hyperpersonalization.
“When you issue an offer, whether the customer accepts it or not, you will need the feedback. So, if the context is wrong, you can learn and offer something different,” he explained.
It is why Leung believes AI intelligence plays a crucial role in hyperpersonalization. “The feedback is not just technology; it is also a business feedback. So it is really about the effectiveness of whatever you're trying to do,” he described.
However, an AI model that learns in real time can drift. If biased, wrong, or sparse data is added to the model, it degrades relevance and effectiveness.
Leung felt that companies should assume that their model will deteriorate over time and not wait for it to happen. “Why will it deteriorate? Because the customer needs and the environment keep changing. So that's where ModelOps comes in.”
He advised companies to start building multiple models. So, the AI can pick the right one and decide whether you need human intervention to improve accuracy. ModelOps automates these processes so companies can choose the model that fits the context.
Stop trying to do it alone
Hyperpersonalization has many moving parts with little wiggle room to rethink or investigate.
So, doing it all alone can be very costly and risk-prone. Instead, Leung advised companies to lean on solution providers who already work with different use cases and learn from them.
The first step is creating the right partner ecosystem. If you’re ready for hyperpersonalization but your partners are not, the initiative will only deliver on some promises and not all.
Leung feels hyperpersonalization projects should involve all stakeholders involved in the value chain, even if they are outside the company walls. This allows them to work toward better data and intelligence sharing that can mutually benefit everyone.
“You need to work with your business partner to ensure your offer can go through the right channels, and you can analyze the data they capture,” said Leung.
Equally important is getting the right solution providers to support you. Although Leung highlighted SAS’s strong track record of hyperpersonalization use cases, he stressed that companies ultimately need to see the relationship as a long-term partnership, where each company learns from the other.
Ignore governance at your own risk
In hyperpersonalization, Leung believes governance is the hardest part.
Leung notes that ethical or responsible AI is already becoming a central talking point in Europe. He urged Hong Kong companies to stay prepared when using AI for hyperpersonalization.
It is not going to be an easy journey. After all, it is difficult to get compliance, ethics, privacy, analytics, and product development teams to work together, Leung admitted.
“These people don't usually work together. So, the direction needs to come from the top. Then you must develop the right culture as these teams learn to work with each other.”
However, Leung advised Asia Pacific companies to do this now and not wait for regulations to materialize. “Don't overlook AI governance or ethics. Because when the compliance or regulatory people come in, it's no joke.”
Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].
Image credit: iStockphoto/Prostock-Studio
Winston Thomas
Winston Thomas is the editor-in-chief of CDOTrends. He likes to piece together the weird and wondering tech puzzle for readers and identify groundbreaking business models led by tech while waiting for the singularity.