Photo: @addie2354 via Twenty20
Are machine learning and AI the path to enhanced personalization?
Two recent surveys from Evergage and Arm Treasure Data have found that marketers are increasingly hoping that artificial intelligence (AI) can take personalization to another level.
The Evergage survey of 314 marketers across industries and countries (largely U.S. based) found 68 percent use a rule-based approach to personalization, compared to 51 percent using a triggered messages/notifications approach and 40 percent using a machine learning/algorithmic approach.
The percentage of companies using a machine learning/algorithmic approach to personalization, however, has climbed significantly from only 26 percent in 2018. Usage of the two other approaches for such purposes remained stable versus the prior year.
Moreover, of those marketers not currently using machine learning/algorithmic personalization, 42 percent plan to begin using it within the next year.
The survey found only 30 percent of those that use machine learning/algorithmic personalization are “very or extremely satisfied” with their personalization programs, but that was better than the 19 percent score for rule-based targeting and 16 percent for triggered messages/notifications.
The most popular locations for personalized experiences were e-mail content, 73 percent; home pages, 49 percent; landing pages, 41 percent; interior pages, 36 percent; and online ads, 30 percent.
A recent survey of 200 marketing executives from Arm Treasure Data with Forbes Insights found only 24 percent applying AI on a significant scale to deliver personalized experiences. Six in 10, however, were starting to adopt the technology to personalize at scale and 57 percent see AI as essential to executing their personalization strategy.
Similar to other recent reports, both studies show personalization can significantly advance customer relationships, even though execution remains a struggle.
The greatest obstacles preventing companies from making personalization a bigger priority is a lack of personnel, cited by 46 percent of survey respondents, according to Evergage. That was followed by a lack of budget, 43 percent; lack of knowledge/skills, 38 percent; lack of organizational alignment, 32 percent; and access to data, 30 percent.
Arm Treasure found that customer personalization is still not a top corporate priority. The study stated, “Technical challenges include addressing data quality issues and data silos, while a lack of effective change management limits companies from achieving greater success.”
Discussion Questions
DISCUSSION QUESTIONS: What are the advantages and shortcomings of using machine learning/algorithmic approaches for personalization? Will AI help retailers and brands overcome obstacles faced in executing personalization?
Poll
BrainTrust
Mark Ryski
Founder, CEO & Author, HeadCount Corporation
Dave Nixon
Retail Solutions Executive, Teradata
Despite the headlines that laud ML/AI as making everything better, the reality is more complicated. The challenge with using machine learning is that it is only as good as the data it has and the assumptions that are made that inform the algorithms that drive the results. In my experience applying ML, we’ve discovered that it’s a process of trial and error. For many retailers/brands, this is all still pretty new and there’s a lot of learning that still needs to happen. Ultimately ML/AI will help retailers and brands improve personalization but it will take time and getting clean, reliable data will be critical to success.
AI, and more specifically, Artificial GENERAL Intelligence, AGI, is going to transform society more thoroughly than did the Industrial Revolution. See: “Possible Minds: Twenty-Five Ways of Looking at AI,” John Brockman – editor (Author).
Yes, Mark, and I would like to add this bit. My thoughts regarding AI and ML lean toward the personalization of the experience more so than a personalization of product offerings. Knowing ahead of time what the consumer may be looking for requires much from the data and habits of the customer while directing the customer through the process based on their answers to questions might better propel them through the shopping process all the way to buy! For my 2 cents.
Kind of a trick question. Yes, AI and ML could help with personalization efforts, but it costs money to accomplish this. It implies very localized in-store assortments, and our data tells us retailers are still enamored with the efficiencies of standardized assortment.
And when it comes to email marketing, I’m just happy that the very large DIY chain I buy things from has stopped sending me ads for snowblowers every winter (I haven’t lived up north for 16 years now). This is not “rocket science” or even “AI science.” It’s basic zip code analysis.
Personalization has a long way to go, and will come with costs as well as benefits.
AI is a natural for marketing. With machine learning retailers can stop using blanket daily emails (!) and start highlighting products a given customer might really want. We haven’t even skimmed the surface on this.
AI and ML can help create the personalized experience that us mortal humans cannot. The capability of storing customer data and using the data effectively through AI and ML are powerful when used correctly. Comparing one customer’s data to the data of thousands of similar customers can help predict what the customer will want/buy next. Personalization is personal and, if used incorrectly, can alienate the customer. Even Amazon’s not perfect, but they have found a good balance that makes the customer feel great about doing business with them.
Machine learning and algorithmic approaches are a huge benefit because they enable genuine personalization across a huge customer base.
The potential downside is that technology-based personalization often struggles to connect with and motivate consumers. In addition to getting good, reliable data, it is crucial that personalization algorithms be built based on customer insights and human behavior considerations because the objective is to affect customer behavior. The goal is not simply to modify the experience for each customer, it is optimizing the experience for each customer to make finding and purchasing what they want seem effortless.
The real magic happens when personalization connects on an emotional level with customers, and strategically designed machine learning algorithms can help achieve that objective at scale.
Large volumes of data, that can be trusted and analyzed with confidence are the keys to making “machine-driven” decisions work.
Right now it is more hype than reality but we will soon achieve “enlightenment” where it becomes relevant.
And for resource-strapped organizations, this workflow is an iterative process that takes discipline and time. Build the models and algorithms, test them, challenge them, tune them and redeploy. This isn’t set it and forget it.
The issue with achieving success for retailers currently is that they are still struggling to get the right data in the right place and extract the right intelligence from it to even begin to act on it (for now).
But soon, we will see the results that ML and soon thereafter AI has been hyped up to deliver.
Consumers are ready for next-level AI because they have been hearing about it forever. Currently, personalization is limited to your name at the top of an email blast, and a repeat of whatever you looked at recently online over and over and over.
The retail industry has been talking about personalization forever – it’s interesting that it is still not a top corporate priority. Talk is cheap, but a lack of budget, knowledge, skills and access to data are strong reasons not to move forward.
Purpose-led machine learning, AI and advanced analytics could certainly help drive enhanced personalization. However, there isn’t a proverbial switch companies could turn on to become a personalization powerhouse. There are far more cultural and change management structures that need to be in place before retailers could truly benefit from driving personalization strategies.
It’s very much a crawl/walk/run strategy and several companies get it right. Consumers enjoy the occasional personalized email, an Instagram ad post, and even a text as long as it’s not creepy or overwhelming and they have the ability to opt in or out.
Sephora is an example of a company that has a customer-obsessed strategy. They spend the time to know their consumers by providing the products and services that meet their needs. Amazon and Google are continually leveraging ML/AI. Bottom line, these technologies are in their infancy and evolving.
Machine learning and crowdsourcing/community building are the only scalable paths to personalization and both take considerable startup investment.
Programs are only as effective as programmers. A drill press is a great tool, but it’s no substitute for a wrench when you want to tighten a bolt. Proponents often seem to be claiming that AI/ML are autonomous, totally rational technologies that somehow can save us from our all-too-human biases when the truth is that they are the creation of our biases. Traditional personalization tools fail because we insist on cramming distinct individual consumers into generic boxes, i.e., “all vegan, female Millennials believe X.” What AI/ML do is allow us to process our mistaken assumptions faster. The problem isn’t the tool, it’s the craftsman.
People are experiencing both the advantages and shortcomings of using machine learning and AI for personalization in retail. The advantages are that people can be very clever in developing the algorithms to be deployed in personalization efforts. They can build out models that segment shoppers into sets of a few and then customize promotions that dramatically increase the potential for success. Shortcomings start with the fact people are fickle. Their behavior changes as frequently as the weather and perhaps more. What works today may not work tomorrow and almost certainly won’t work in a few years.
Up until now, the shortcomings of personalization efforts have overshadowed the advantages. Recent, and I mean within the past year, advances in technology are finally giving us the right tools to create the algorithms that not only work to understand and predict the fickleness of the shopper but create the right incentives to address it.
Given the requirement for creating and testing AI algorithms, personalization is likely to be more successful at defined target market levels rather than at the individual level. Getting to individual level personalization is likely going to take some time.
Hope isn’t a good strategy here: “marketers are increasingly hoping that artificial intelligence (AI) can take personalization to another level.” Are AI (rules or ML) solutions for marketing personalization worth exploring? Absolutely. Is it a good idea to explore this in isolation from a corporate or enterprise plan for leveraging data, AI and advanced analytics? No way – there’s no evidence to show that companies are winning in that way. Personalization is one use case of many. Companies should prioritize and have a plan to test new methods relative to the current state.
The application of ML and AI still has a long way to go. I expect a few of the larger retailers will continue to make investments in it, and once they post in their quarterly results that it fueled their results the laggards will quickly try to follow suit. I love the idea of it, don’t get me wrong, but there’s a few more trends that are further in line here and I expect those to get most of the attention in the short-term given they have a more practical (and visible) result. An example would be using AI for safety-stock analysis for in-store inventory. That application can show an almost immediate tangible results. AI becomes more palatable when the words thrown around are tied to cost (at this point), not experiences (just yet).
You can view personalization through the lens of a maturity model with triggers at the basic/foundational level, rules based as advanced and AI as distinctive.
Personalizing customer communications and recommendations based on customer context is key to optimizing revenues. The problem is there is too much data for a human to analyze and synthesize for each customer.
Taking the expertise from the best associates and incorporating these processes into a system that is triggered by rules and/or algorithms to optimize the experience (just like the Amazon experience) is the concept of AI.
Obviously, it is critical that the inputs are accurate information to ensure accurate communications and recommendations. Garbage in, garbage out. That old phrase is as appropriate today as it ever was. Retailers need to be aware of the potential mistakes that AI can make based on historical data and biases and monitor the results of AI to help catch faulty logic.
Humans are not perfect and neither is AI.
We are seeing real, measurable results leveraging AI for personalization. There are some good tools from various vendors that utilize true machine learning by gaining better promotional personalization as more and more trusted data is consumed by these tools.
Based upon the surveys mentioned in this article, as well as global studies on our own, my belief is that the overriding obstacle is not budget nor skills, but the capturing of the right data, both internal and external, and the trust level of that data.
AI/ML’s number one value add for personalization is scalability. It allows for a machine to figure out what kind of messaging, options, customer path, and incentives to offer a customer based on a myriad of parameters. What makes it better than outright rule-based delivery of personalization is that it also has the ability to adjust its rules based on changes in circumstances and customers- allowing for scalability of the maintenance and support of personalization across wide and diverse audiences. That said, it’s still very early in the game and AI/ML has a long road ahead. Retailers have to invest heavily in the right people and tech to shift their financials and see results.
Good article Tom! To my knowledge, the customer experience in every field is the key factor for the success of a company. Innovations in technologies can make customer space more realistic and competitive. Nowadays artificial intelligence indeed has changed the scenario of customer care departments. Companies can go with tools like CSAT.AI, Salesforce Einstein for better customer experience.