AI for e-mail? Marketers enthused, but not without concerns
MarketingCharts staff
Through a special arrangement, presented here for discussion is a summary of articles from MarketingCharts, which provides up-to-the-minute data and research to marketers.
Virtually all marketers (96 percent) are confident that machine learning can personalize e-mail content down to an individual’s specific interests and improve the customer experience, according to a survey of more than 400 U.S. marketers from The Relevancy Group, sponsored by OneSpot.
Beyond personalizing e-mail content, the application of machine learning can eliminate some manual tasks that occupy a significant portion of marketers’ time, per the report. Those executional tasks include content selection, HTML coding and proof testing of messages, which marketers say collectively take up almost one-quarter of their work week.
Senders at enterprise companies and those at retail/e-commerce companies appear to have the most to gain, according to the study.
Were they able to free up time by introducing machine learning, seven in 10 respondents say they would use that saved time for program planning, expansion and strategy, about half would put more focus on data analysis, and more than four in 10 on subject line optimization.
But the use of machine learning in e-mail marketing doesn’t come without some challenges. Seven in 10 are concerned about the implementation and training involved, and the same ratio are concerned about switching or adding a platform to their marketing tech stack.
These concerns seem well-founded, considering that marketers who have already adopted machine learning in their e-mail programs spent four months on average on implementation.
Meanwhile, two-thirds are concerned about giving up some editorial control to the technology. That’s a shared concern held by many different constituents when it comes to artificial intelligence: a recent survey from The Integer Group found that only one-fifth of shoppers said they would be comfortable letting AI proactively order items they haven’t previously purchased but might like.
Nonetheless, The Relevancy Group believes that concerns surrounding machine learning in e-mail — including trusting its effectiveness, an issue for four in 10 marketers — will fade as marketers gain more comfort and data from using the technologies.
Discussion Questions
DISCUSSION QUESTIONS: Does machine learning offer more benefits than drawbacks in personalizing e-mail content? What aspects do you find most promising and most concerning?
Poll
BrainTrust
Gib Bassett
Director, Solutions Marketing with Alteryx
Tony Orlando
Owner, Tony O's Supermarket and Catering
Most retailers can’t even distinguish between males and females in their email campaigns. Let’s not run before we can walk …
Agreed that a most retailers are behind. And at the same time there’s an opportunity to skip the small steps in the middle to go directly with a more advanced solution. It’s limiting to think that incremental improvements are the only steps possible. With the exponential growth of technology, greater leaps are not only possible, but necessary.
Consumers want highly personalized email. Once they get it through AI, will they still open it? AI is not a 100 percent solution, and if marketers abuse the privilege of being able to put messages into consumers’ inboxes, consumers will simply unsubscribe.
The most costly part of any initiative is always the mistakes, and personalizing email content with anything much beyond transaction data is a minefield of error. As AI infers and deduces these inexact sciences can easily put the relationship and brand equity at risk. Some marketers, such as political campaigns, that send a barrage of emails consider “close enough” to be “good enough.” But commerce seldom has the same latitude in protecting or building the consumer/brand relationship.
I would reframe the question around how analytics can support better personalization, and machine learning is one method to support the use case. I would also say that the drawbacks such as staffing and learning should be mitigated by having a strategy that includes how analytics supports both marketing and the overall retail or consumer goods company in pursuit of better CX. Questions like this I think are too tactical in nature. Companies should step back and evaluate their maturity with regards to analytics, the competencies of their people, and their business objectives. The drive to improve marketing and create more personalized email doesn’t have to slow down, but it should be considered within a broader construct in my opinion. Too many companies have failed or struggle as a result.
Email as a personalization channel has its limitations, as everyone is overwhelmed with so many emails as it stands. A bit too much personalization or frequency of emails may result in consumers tuning out, or unsubscribing from the content. Email has unfortunately evolved into the new snail mail and consumes time from our busy day.
In our mobile and digital-first world, democratizing the personalization and the level of interaction should be in the hands of the consumer, and the preferred channels should be native apps, messaging or text messages. That way, the pop-up messaging is well received and the consumer can click for more details if they wish to engage further.
Marketers are always looking for a quick fix to address their quest for reach and frequency. The new world requires marketers to pay attention, listen and learn who their customers and shoppers are — what they expect, what they aspire to and what inspires them. That won’t happen with a click of a button. Knowledge, timing and relevancy are the required elements. If machine learning is used for those attributes then there is valued potential. If all you’re going to do is use technology to optimize your subject line and blast away — you’ll fail.
I love truly good AI and machine learning experiences. When a consumer can go to a brand and have a “conversation,” so to speak, that helps them make a buying decision at that moment — happy consumer.
Now let’s look at this as it relates to email marketing. Yes, email marketing would have more benefit than drawback if it were personalized and AI can help with that but, it still relies on where the consumer’s head is at the time. An age-old problem of email marketing: what to say to get an open.
I am hoping to see a lot more email campaigns that send consumers back to brand sites encouraging them to have a “conversation” online where it can be interactive, in the now and can lead to a purchase. Perhaps the email itself need not be personalized to what the consumer might want to buy but rather as a followup to the last interaction. “How did you like that red dress you bought last month? Can you afford a new pair of shoes to go with it? It would make a whole new outfit. Click the link.” For my 2 cents.
AI technology is great for email content that is standard. Some good email marketing taps into current or local trends/humor which I believe may be difficult for AI to capture. It also depends on how a brand’s email marketing is required to capture the essence of the brand personality. That will be a bit tougher to do as it is hard even for some humans to translate into copy.
First, let’s have a discussion about how much email goes unopened in the first place. Personalization only matters if the consumer actually reads the email, and there is plenty of evidence that most messages are pretty much unwanted right off the bat. If machine learning is used to identify better times and places to send email within a customer’s lifecycle, then sure, I see lots of benefits. But using high-end optimization to personalize something that consumers find annoying in general seems a lot like “paving the cowpaths” rather than doing something that benefits both consumers and retailers.
Remember, supposedly consumers only wanted a “faster horse” – so when they say they want more personalized emails, that’s pretty much the same thing. Of course they want them to be less annoying and personalization can help. But what would be better is more tailored engagement overall, within which email can play a more intelligent role.
Machine learning and AI, as presented in this article, focus on one aspect of email marketing — making the content more personalized and therefore more accessible and consumable by customers. That, of course, assumes that your customer opens the email in the first place! Where machine learning may bring the most benefit is in looking at analytics data to help create more interesting and relevant subject lines that cause higher open rates. If your customer doesn’t open it, no amount of personalization will generate a sale! Most retailers are still stuck in a mode of moving from “spray and pray” email blasts to controlling the frequency and relevance of the content to each customer segment. The level of advancement being discussed here may be wonderful, but most are not in a position to take advantage of it yet.
Machines have their place for sure, but success for many businesses still comes from a beating heart with a smile — a person who actually listens to your concerns. I can give you a million examples on how to make a sale and build loyalty through the personal touch, but this is an internet-driven site, and I fully appreciate what our online social media program can do to bring in the customers. However, once they come in, how they are greeted and treated is what builds loyalty. That is much stronger than robots, or AI, and small business folks who know how to engage with their customers will win the loyalty game the old-fashioned way, which is one customer at a time.
Combining online personalizing with the old-fashioned style of service would be a win/win, and that is how the future of our business will continue to stay relevant. Have a great weekend everyone.
Great article and cool research! The point about “two-thirds are concerned about giving up some editorial control to the technology” sparked my interest.
Our agency runs email campaigns for e-commerce companies, and this topic of automation comes up a lot. Our philosophy is, “Automate science, humanize art.” Because as a protocol, there isn’t an unlimited amount of automation that can be set up for email marketing. It’s a finite thing.
Machine learning is amazing, but it’s also key to make your email strategy much better with real, curious, focused human labor. That’s one way brands can keep a sense of control in an otherwise chaotic marketplace.
ML/AI offer far more benefits, and actually, I have a hard time thinking of ANY drawbacks for marketers. I’m not saying this as an employee of a company that is helping retailers around the globe with these technologies as we speak, but from a user standpoint. The right user interface provides the most intuitive use of AI that I have seen. There is seamless linkage across business functions, marketing, supply chain, merchandising, etc., where colleagues see the effects of actions in other business areas and can proactively make intelligent decisions based upon those actions. Also, ML leverages the best of humans and the best of machines and it gets “smarter” the more data you feed it. This is great stuff, and I’m convinced that surveys like the one in the article are based upon the user’s/manager’s fears of the unknown, rather than practical experience.
Marketers with large email lists who run a lot of campaigns will benefit most. It takes a lot of training data to figure out which buyers respond to which offer and which creative at which time.
Optimizing content, time/day, offer, and the like for micro-segments is a big chore. It’s easier to optimize around open rates, which average 20-25% for retail, but harder to optimize around click-through rate, which is 3-4% for retail. Optimizing around purchase rates is harder still when the action taken is measured in double-digit basis points.
Using AI will lighten the workload for marketers, but it will take most brands a lot longer then they anticipate to yield a result that’s predictive of anything.