As AI-branded technologies continue to proliferate (ChatGPT, anyone?), machine learning has become a hot topic in the marketing automation space. Companies can gather vast quantities of data about their customers, and finding shortcuts for using that intel is alluring. But it takes careful consideration to shine a light on where there’s real substance and where some AI promises don’t hold up.
Cutting through the buzzwords to find out if AI has true functional value for your brand can be tricky. Here’s how three of the most frequent AI claims fall short, and how to achieve better results by personalizing campaigns.
Sending messages at the ideal time can improve open and click-through rates. STO has been around for some time, and AI models purport to make it effortless. But machine learning isn’t the only thing that’s been evolving; advanced privacy technology now muddies the data so much that STO automation can become meaningless.
Reality: Quality content overcomes questionable data
Open rates, which AI typically relies on for STO, have become unreliable. Apple, other ESPs, and corporate security screens now have privacy settings that interfere with open-rate tracking. STO features that depend on open rates pull in false metrics that appear legitimate, making send-time choices based on bad data.
Developing workarounds is time-consuming and resource-intensive. And even if STO features could accurately track open rates, it would be difficult to predict how many data points are needed in order to truly optimize the send time for each individual.
In reality, the exact moment your message arrives is less critical than the actual content you’re sending. Your first priority should be crafting a message that’s engaging, relevant, and useful to the reader. Consider this example: you’re a subscription box service, and you’re releasing a limited edition add-on that’s sure to sell out quickly. The timing of your email does matter: customers need to get the message in time to take action. But if your subject line doesn’t clearly convey the value and urgency, people are unlikely to open the email, whether they receive it at 6:04 or 7:13.
Smart solution: Personalize send time based on location with Customer.io
Using Customer.io’s Timezone Match allows you to choose the ideal time of day for the specific message you’re sending, and ensure customers receive it at the right time in their location. It uses first-party location data to automatically add or subtract hours from campaign send times, so if you want your email to arrive at lunchtime, it hits each person’s inbox when they open their lunchbox, wherever they are in the world.
Once you’ve developed content that performs well, you can set up A/B tests to zero in on the send times that work best for your specific audience. Make sure to measure success with meaningful metrics; you can set custom conversion goals to track customer actions relevant to your campaign aims.
Lead and brand loyalty scoring uses a prospect’s characteristics and behaviors to predict how likely the person is to convert, helping marketers target pitches. AI promises to automatically generate accurate lead scores. The problem? Algorithms can’t understand as much as marketers can about their business.
Reality: Meaningful lead scoring requires nuance and visibility
The attributes and behaviors that predict customer conversion are specific to each company (and, often, each campaign). Let’s say you’re marketing a freemium fintech app. Customer A and Customer B both download the app and connect their bank accounts. Customer A goes on to create a savings goal, while Customer B sets up a budget.
By observing customer behavior, you’ve learned that setting a savings goal indicates the customer is more likely to convert to a paid plan. So you’d give Customer A a higher lead score.
How would AI score those two customers? The answer is impossible to know. AI works in a black box; it will spit out a lead score, but you won’t know why. And the algorithm lacks the insight your team has about the factors that drive successful outcomes for your audience.
Relying on a black box to decide what actions are significant, with no visibility into how those determinations are made (much less any control over them!) is a recipe for uncertainty and unpredictable outcomes.
Smart solution: Automate meaningful lead scoring with Customer.io
Using easy-to–build automations, you can create lead scores that reflect your unique customer journey. Simply create a lead score attribute, and use the Create/Update Person action in campaign workflows to increase or decrease the lead score based on behaviors you’ve designated as noteworthy indicators.
Returning to the fintech app scenario above, you could create an event-triggered campaign to increase a customer’s lead score attribute when they create their first savings goal, then move them into a segment that triggers a conversion campaign. If the customer fails to click through after a certain number of messages, you could reduce their lead score using a true/false branch. Here’s how to put this approach into action.
Messaging customers in the channels they prefer can reduce unsubscribe rates and increase engagement. The idea of using AI to automatically direct messages to the best channel based on customer behavior might sound appealing. The problem is that AI doesn’t care about context: it will zero in on the channel regardless of how appropriate it is for your message and goals.
Reality: Holistic strategy should drive channel selection
While user preferences are important, in isolation, they aren’t enough to guide your decisions. The customer’s engagement and lifecycle stage are crucial factors in choosing the right messaging channel.
Imagine you have a productivity app and you create a campaign to promote a new time-management feature. Customer A and Customer B have both engaged most with in-app messages. An AI solution would use that channel for both customers when you send your campaign.
But that approach misses important context: Customer B has been inactive for 90 days and is at risk of churning. They won’t be enticed by a new feature if they’re not even opening the app; a smart marketer would connect through email or SMS.
Another problem with AI solutions is the difficulty of judging data significance. How many data points, across how many channels, are required to know the best channel for sending a particular communication to a specific individual? And how can you be sure that AI is controlling for additional factors, like the time sent? The answer is simple: you can’t.
Smart solution: Leverage customer preferences wisely with Customer.io
Customizing your automations for channel preference provides greater flexibility and drives better results than out-of-the-box AI options. Start by considering the goals of the campaign, the content, and the call to action. That perspective will point to the ideal channel for reaching the individual and persuading them to take the next step. Ask yourself, “When I’m ready to send the message, where is the customer in their journey?”
For example, you can build segments of users who engage with SMS but not email, and vice versa. Once customers are segmented, you can include or exclude them from campaigns at the trigger level or even at the message level. You can also use branching within a single campaign to decide which messages they should receive. Here’s a recipe for automating messaging channels aligned with your campaign goals.
At Customer.io, we’ve found that automating personalization delivers the greatest value to customers. Machine learning makes a lot of enticing promises, but in the marketing world, it can’t compete with the powerful combo of good data and insightful marketers who know their customers and businesses.
Sure, AI can be useful in certain situations (and pretty funny in others). But if it doesn’t deliver practical value for your business, it becomes a drain on resources you could be using to personalize messaging relevant to your brand and audience.