Email is one of the most effective bank marketing channels available. Most banks send emails with little regard to optimization – they create an email, then send it. Other banks, like ourselves previously, pour over countless amounts of data to optimize open rates, clicks, and conversions. Now, it is about personalization and using artificial intelligence (AI) to guide the effort. Using AI for bank email marketing can get you a 2% to 13%+ lift, or improvement, depending on the technique. In this article, we give you our top five lessons to inspire your bank to adopt AI in marketing or leverage our data to improve your effort.

The T-Shirt Problem

If you ever need to order t-shirts for a group, there is a well-defined statistical formula that looks something like the data below. The data set is fantastic as there are tens of thousands of orders placed each year. Despite all the statistics and data, anyone that has ever done this knows that if you use the formula below to order, you will likely be off on about 15% of your order. That is the average margin of error when it comes to ordering t-shirts.

Maybe your group is too small, say below 100 people. Perhaps it is your cross-fit group, and they skew larger, or maybe it is younger than the average group, and they skew smaller. One person will purchase more than one shirt. Statistically, those people will be medium or larges, and your data will further be altered. Combine all that with an aging inventory, loss, theft, and damage, and you have likely under or over-ordered by 15%, costing your group approximately $10 per shirt plus the opportunity cost of not having a size available to sell.

If you do this once, you never do it again. You learn to have everyone in the group pre-order whenever possible.

All the above holds for email marketing. You can use all the statistics you want, but you are likely going to be off by 15% or more. However, instead of incurring a $10 cost or missing out on a $10 profit, you are likely missing out on a $300 profit for a retail bank customer and hundreds of thousands for some commercial customers.

Luckily, there is a better way.

Using Artificial Intelligence (AI) For Email Marketing

AI can look at a large data set, in our case, a sampling of about 500,000 emails over the past 90 days, and look for clues to what prompts customers to open, click and, in some cases, convert. AI makes predictions, tests the hypotheses, improves the approach, and retests down to the individual level. What machine learning has done for our emails would have taken a team o five analysts more than 100 hours to compile. AI does it in seconds for a fraction of the costs and likely better.

Because of its powerful and evolving capabilities, AI can do a number of things at scale efficiently to include drafting copy, writing subject lines, creating graphics, and improving delivery. Once your bank understands and uses these tools, you will never go back.

We have tested various systems over the past quarter, and below are our top five lessons.

Lesson 1: Fridays Are Good Days

Before AI, we had rough statistics that showed Thursdays, Tuesdays, Wednesdays, Monday, Friday, Saturday, and Sundays were our best days to email customers, in that order. The flaw in our approach was that we looked at data in aggregate and as a snapshot in time. We would send emails at 6 am, 10 am, or 12 pm and then look at our open rates at the end of the day.

The problem with that approach is, like t-shirts, it generalizes your customer.

By customizing delivery times based on individual behavior, Fridays are actually a pretty good day to deliver business-to-business (B2B) emails. It turns out that many customers are looking either for distractions on Fridays or are trying to clear their inboxes before the weekend.

This finding alone radically changed our approach and jumped our marketing lift. We now send an email out and let the AI-driven email application deliver it at the best time according to each customer’s behavior. It turns out that Thursdays are, by far, the best day for email delivery, followed by Friday.

Lesson 2: Emails Get Open After Lunch (Usually)

Prior to AI, our data showed that the best time to send an email was six in the morning. However, if you customize the delivery time, you can improve all your metrics, particularly your email open rates. It turns out that if you deliver an email closer to lunchtime, you have a better chance of getting it open and acted upon. Business customers tend to open more emails after lunch than they do in the morning.

As the data below shows, 10 am on Thursdays and Tuesdays is the second-best time to deliver an email. Monday, Wednesday, and Friday mornings tend to be not such great times.

The other interesting insight that comes out of this data is how long customers work at email opening on Thursdays. Email open and click rates remain above average (the gold line) well into the night at a constant rate.

Lesson 3: Demographics Can Hurt Your Bank

If you sit around your bank talking about winning more millennials or Gen-Zers, you are not only wasting your time, but much of your efforts are likely counterproductive. Grouping customers by age, occupation, income, gender, or ethnicity tends to overgeneralize to the point of ineffectiveness and can be insulting at worst. Assume all Gen-Zs are technologically proficient or active on social media, and you have already missed your message to about 49% of your audience. Further to this point, while only 11% of Baby Boomers are more active than your average Gen-Z customer on their smartphone and social media, that 11% generate more profit for banks than the 49% in aggregate.

Stereotype your customers at your peril.

The far better approach is to market on intent and then let customers, or potential customers, self-select. Through the power of AI, you can individually target customers using specific colors, graphics, subject lines, and word choices. For any given intent, your customers are likely to break down into something like the following segmentation below.

AI For Bank Email Marketing

Here, a majority of customers are not going to care about your message, while your “loyalist” will care about almost everything you send them. Using this data, you can either custom craft a message or let AI alter your message and delivery to increase engagement among each subsection.

With either method, you have a quantifiable approach for improvement.

You can now stratify your audience along an almost infinite amount of dimensions. Those that care about loans, those that care about deposits, saving for retirement, mortgages, treasury management, community involvement, sustainability, and whatever else you want to create engagement scores for. You want to develop more homeowners as customers or customers that care about payments, no problem, as AI will test a path to success.

Lesson 4: Using AI for Bank Email Marketing Can Improve Your Financial Model

Craft a message for the engagement model like the one discussed above, and AI can predict your performance based on each customer’s probability of past engagement. Take the email data, and AI can do the same for mobile SMS, organic search, paid search, and digital advertising. Add these together and you not only have a solid basis for setting and allocating your marketing resources, but you will now have a basis for estimating the demand for any given product.

For example, based on the engagement model above that was created around a lending product, we know with some level of accuracy that an email about a new lending product will be opened by 21% of our target customers, clicked on to learn more by 9% and almost 2.2% will take us up on the product within the next 90 days.

The kicker here is that our models are new. The more you use, build, and train your model, the more powerful they become. As such, not only does your accuracy increase, but your effectiveness can double. That 2% conversion rate can quickly go to 4%. By using AI for bank email marketing, if you can increase the effectiveness of a single campaign by 2%, that one campaign will likely more than pay for the AI effort.

Lesson 5: Word Choice Matters and You’re Not Smart Enough to Figure It Out

To be fair, AI isn’t smart enough either to figure out the exact message and word choice that works the best for each customer, but it is rapidly improving.

The wording for lesson five above was AI-generated. You can render your own opinion on its quality.

Some of the copy within this article was AI-generated, and the subject line plus the title on our newsletter (that likely led you to this article) was AI-generated. While there are various subject lines you could have seen, probably you either saw – “5 Lessons We Learned Using AI for Bank Email Marketing,” which was generated by our best subject line writing human, or you likely saw “ How AI is Changing the Game for Bank Email” that was the lead AI-generated line in our testing.

We will let you know if humans or machines wrote the subject line better. Regardless, AI-generated copy to match each user is likely the biggest area of growth and improvement for bank marketing over the next two years.

Putting This Into Action

We will highlight more AI bank marketing data, case studies, and specific applications in the coming articles. We will show how AI can help write your bank’s press releases, advertising copy, design graphics, financial reports, and even produce videos.

We will show you the how, and the reasons why, you no longer have to create a marketing campaign. You tell the application that you want to send an email or post a display ad. The AI will not only make the content but will choose the product, promotion, channel, and time. AI will keep improving your campaigns until it reaches a level of optimization.

AI is quickly changing bank marketing, and email is just the tip of the proverbial iceberg.

Tags: , , , , , Published: 07/11/22 by Chris Nichols