By Andrew Miller
Bank marketers are used to change and perhaps no change is more consequential than machine learning’s impact on our ability to attract new customers online. The rise of machine learning comes at a perfect time as marketers are frequently asked to “do more with less” and our access to data exceeds our human capacity to consume, analyze and react in real time. Luckily, computers are great at processing lots of data and don’t need naps.
It’s safe to assume that most bank marketers don’t moonlight as programmers and many of us do not have access to teams of developers. So how does one gain the advantages of machine learning without having to write any code?
First, let’s establish a few basic terms to make sure we’re on the same page. Then we will explore the current “no code” machine learning tools that help us reach more prospective customers, more efficiently.
How is machine learning different from artificial intelligence?
The two terms are often used together but actually have very different meanings.
Machine learning is a process for training computers to learn from large data sets without programming every possible outcome in advance. Artificial intelligence describes technologies, often built on machine learning models, that enable a computer to simulate human behavior.
In most marketing contexts, ML tools analyze historical data to predict future outcomes. Behind all the “black box” metaphors and buzzwords is simply a method of making more educated guesses about what is most likely to happen given historical patterns. Not what will happen.
Machine learning tools for bank marketers
One does not have to look far to see machine learning at work. Many ad platforms, content creation tools, and analytics platforms already incorporate advanced machine learning algorithms to improve performance and efficiency.
Here are some of the advantages you can gain with machine learning today:
Set optimal cost-per-click bids based on campaign goals. Google Ads, Facebook Ads, Microsoft Advertising and many other pay-per-click ad platforms use machine learning algorithms to optimize bids in each auction as it happens. Their algorithms take many signals into account, including many that are unavailable for manual bid adjustments.
Advertisers can select campaign goals such as “maximize conversions,” “maximize conversion value” and “target cost per action ” that are powered by machine learning algorithms that choose the right bid for each keyword, searcher, and conversion goal to maximize the likelihood of achieving your marketing goals.
Target the best audiences and placements for online display ads. Advertisers no longer need to select specific sites for their display ads to appear on. Machine learning-enabled targeting in Google Ads, Microsoft Advertising and Facebook Ads automatically combine the best site placements, ad units and bids based on individual web users’ likelihood to convert on your site.
As an example, the ad platforms are getting better at distinguishing between a user looking for a consumer banking product versus a commercial banking product who may both be using the same search keyword. This improves the likelihood of showing them an appropriate ad based on their intent.
Marketers can experiment with new campaign types such as Performance Max (Google Ads) and Automated Ads (Facebook Ads) to extend their campaigns’ reach even further and optimize performance across multiple devices and channels without having to create separate campaigns for each. For example, a single Performance Max campaign can show ads on Google search, the Google Display Network, Google Maps, Gmail, YouTube and Android devices. Previously, each channel would have required the management of separate campaigns with separate budgets and audience targets.
Create the perfect ad for every prospective customer in real time. Google Ads offers machine learning tools for search ad copy and display ad creative as well. Responsive Search Ads and Responsive Display Ads allow advertisers to test thousands of creative versions simultaneously. Instead of creating complete ads, advertisers upload several variations of each element—headlines, descriptions, images and calls to action. Google automatically rotates the ads and optimizes for the combinations most likely to convert for each visitor.
Bank marketers face a challenge if regulatory messaging is required to appear within an ad. Advertisers can “pin” certain elements of an ad which ensures they show up for users, but Google will mark the ad as being of lower quality because there are pinned assets and fewer testing opportunities.
Determine the value of each marketing channel with attribution. We are all dealing with less data as a result of privacy regulations and limits on capturing personally identifiable information. This can limit our ability to understand the full value of each marketing channel and use website analytics to measure outcomes such as account openings and financing applications.
Marketers that use Google Analytics 360 or Google Analytics 4 can use a Data-Driven Attribution model that helps fill in the data collection gaps by modeling predicted behaviors based on samples of visitors that are trackable. DDA takes advantage of machine learning to look at 50+ recent touchpoints in a customer’s journey to assign a value to each channel in reports.
Unlock insights buried in mountains of data. Analytics tools and ad platforms create more data than a human can possibly analyze and act on. Machine learning tools can identify trends and anomalies in large data sets faster and with greater precision.
For example, it may not be obvious if your bank website suddenly gets 300 percent more traffic from a neighboring town when your minor league baseball sponsorship is activated. Google Analytics Insights will surface the +300 percent anomaly and allow you to act on it, perhaps by creating a new webpage to greet and convert the new fans by promoting a new account offer.
Practical considerations for incorporating machine learning tools
Machine learning is a buzzworthy topic. But it’s not a panacea for all of your marketing woes. Marketers should consider these scenarios when augmenting their work with ML tools:
- Machine learning requires a lot of data to train the algorithms. Few banks have enough structured marketing data for training and testing their own ML tools, but advertising and analytics platforms can aggregate data to generate recommendations and insights across industries.
- Machine learning makes educated guesses based on historical data and cannot be expected to navigate unfamiliar situations such as changes in local regulations, offline events or strategic shifts.
- Humans are still required to provide oversight and monitor the computer-generated insights and recommendations to determine how to react—or whether to react at all.
- There likely won’t be one ML tool that does everything well. Marketers will have to piece together a suite of specialized tools because of the myriad combinations of platforms, data structures and strategies.
How to prepare for the future of marketing, augmented by machine learning
Fewer humans will be needed as machines are trained to handle repetitive or data-intensive tasks. But there are many areas where human marketers have a distinct advantage. We are still better at articulating objectives, developing strategies and brainstorming new creative ideas. We will also need qualified humans to provide oversight and guardrails for machine learning tools to know when, and how, to put them to work.
We can anticipate further consolidation of machine learning tools to accomplish broader tasks, but humans and machines will accomplish more together than either could alone.
Andrew Miller is co-founder and VP for strategy at Workshop Digital, a digital marketing agency in Richmond, Virginia.