AI is becoming more prevalent across digital aspects of business, but not every customer is thrilled with the new technology. In its fourth annual “Creepy or Cool” survey, RichRelevance found that only 32 percent of respondents felt OK about AI. The vast majority, 81 percent, believe organizations have an obligation to tell customers when and how AI is being put to use.
That means any company hoping to benefit from AI’s capabilities and retain customers needs to be transparent about the technology fueling its platforms. But companies shouldn’t fear AI. The biggest opportunity most organizations have is prioritizing their data to find the right balance of customization.
There’s a sweet spot between overgeneralization and overpersonalization: It’s somewhere between blasting out 10,000 copies of the same email to 100 different targets and attaching a picture of the prospect’s house in an email. Companies have loads of data on their customers, and that will only increase. However, when that customer information is used incorrectly, it can be off-putting or downright unnerving.
Hitting the right balance
Prospects and customers are more than data points on a spreadsheet. It’s exciting to have rows of helpful information that can help engage people, but like a first date when one person has clearly Facebook-stalked too much, it can be overwhelming. Look at the audience critically to ensure you don’t flex an AI strategy that might cross the line for some readers.
The whole team should discuss what level of detail will be comfortable for readers. How would employees prefer their information to be used? What’s impressive to see in a targeted email? And which data points, when used, feel like they cross the line?
Just as with any other process, form a hypothesis and test it; isolate variables and qualify the responses received. Which audience members responded favorably? Which buyer persona do they most resemble? Are there any patterns within those buyer personas? These insights build and begin to inform future experiments, leading to better-informed strategies around which data to use as well as how and when to use it.
It might sound complicated, but there are a few ways to ensure that the process achieves the right balance. Here are three steps you can take to use AI for your emails in a way that benefits your company and your customers alike.
1. Organize and connect your data.
Data silos can kill an AI strategy before it gets off the ground. The entire point of leveraging machine learning is to uncover connections between seemingly disparate dots. If you silo those dots away from one another, your AI can’t use them to form links. And without them, the chance to collect real insights diminishes.
According to an Evergage study, 98 percent of marketers agree that when implemented correctly, personalization helps customer relationships grow stronger, and nearly 88 percent report that their customers expect this type of customized experience. But almost half of marketers have four or more systems storing their customer information. These companies are limiting the insights they can derive to meet consumer expectations.
Ensuring that these siloed data sources are brought together will be a task for a technical role at a company, or perhaps for an outsourced hire. When my company, Sapper Consulting, was younger, we didn’t see the need for a full-time technical role. We leveraged our network and talent-sourcing resources like Upwork to tackle projects as they came up.
2. Test and measure everything.
Every aspect of a good email message should be tested and tracked. This includes subject lines, preview text, hooks, calls to action, body paragraphs, titles, industries, regional content, purchasing habits, content downloads, webinar visits — you get it. The more data points you’re able to test, the more future decisions those findings will impact. If all goes well, effective content patterns will emerge. But don’t stop testing even when some tests yield positive results.
The silver bullet subject line that got an 80 percent open rate will eventually lose its luster, so it’s important to have other effective strategies on deck. Allocate the budget and resources needed to hit KPIs, but never give up R&D. This testing can occur across the entire organization, from developers and product teams to front-line customer support teams. Measuring is partially dependent on the right software, but analytical minds are what transform spreadsheets into action — think data scientists.
Spotify excels at this type of iterative testing of new content types. Its curated Discover Weekly playlists are largely driven by an algorithm created by Echo Nest, a music intelligence company Spotify acquired in 2014. While Discover Weekly might not be perfect, it learns more about customer preferences from songs that get skipped and songs that get “liked” and added to a user’s personal library. This is one of the clearest examples of real-time small-batch testing using consumer feedback.
3. Listen to your customers.
If customers don’t like something, they’ll let you know. It’s not hard to imagine ending up on a Buzzfeed list of creepiest marketing emails if you leverage the wrong data points. If there’s a clear pattern of disliking the use of a home address in solicitation emails, don’t use home addresses.
“Personal” means different things to different people; true personalization means being aware of and catering to the preferences of individual recipients. Once personalization begins to work against the best interests of your customers, it’s a detriment to your marketing strategy.
According to a Periscope by McKinsey study in Europe and the U.S., about 40 percent of consumers reported that the messages brands send them are only sometimes personalized. Of the Americans surveyed, 31 percent said the emails they receive are typically relevant to them, while 23 percent reported seldom or never receiving relevant messages. Brands clearly have work to do, so listen to consumers when they give both positive and negative feedback.
Using AI is quickly becoming the norm across industries, but consumers expect brands to be transparent about how, when and why this technology is incorporated. The three steps above are the keys to wowing customers with AI-driven personalization instead of creeping them out.