As technology has evolved, so too have the shopping habits of the modern customer. With an endless slew of reviews, images and video demonstrations constantly on hand, many of today’s consumers no longer casually graze on a diet of store window offerings and impulse buys. Instead, armed to the teeth with keywords and search engines, they open their browsers and join the hunt for the perfect catch, that premier purchase: the one product with an ideal price and incomparable features.

Immersed in the tangle of (often contradictory) opinions and product advice online, even the most knowledgeable customers may wander into unhelpful territory. As marketers, the goal is to guide buyers as they peruse the untamed jungles of cyberspace. It’s key to provide the tools they need to find their path. Singularly powerful among those tools is machine learning, a concept that top brands tailor to automatically connect customers with the messaging they want to see, at the moment they want to see it.

Marketers who earn real results using machine learning and automated ads keep in mind a few rules of thumb when building and deploying digital campaigns.

Prioritize Growth

The specifics of what machine learning algorithms are optimized to complete can mean the difference between breakaway success and bankruptcy. Concentrating not on immediate ROI, but long-term results, is key; for example, by using custom intent audiences to draw in YouTube viewers searching financial terms on Google, e-investment firm Betterment recouped their ad spend six times over, while also achieving a 245% increase in brand-related Google searches. Taking a comprehensive approach to machine learning will generally lead to better outcomes than a strategy based around short-term efficiency. Another example: a machine learning algorithm that prioritizes action around completed purchases—those made online, in-store, or over the phone—may be more likely to deliver actionable ads than an algorithm optimized only for online information requests.

Attract High-Value Customers

It’s tempting to assume that customers with expensive tastes are those driving sales long-term, but in many cases, this just isn’t true. Often, customers may make modest purchases, but do so consistently over time. In all things (and especially purchasing habits) the pareto principle holds. Most marketers are aware that the overwhelming bulk of sales, 80%, are generated by only 20% of customers; these are the customers that automated solutions should be tailored to satisfy.  Acquiring high-value buyers and maximizing their customer lifetime value (CLV) should take precedence when allocating resources, including machine learning.

Don’t Neglect Creative

Many would argue that the internet is already overstuffed with automated marketing; in this case, the ads that succeed will be those capable of eliciting the most human response. Forging connections out of raw emotion has always been the hallmark of effective media. Fortunately, we now have machine learning tools that harness creativity to develop meaningful marketing. Google’s responsive search ads function allows creators to enter numerous headlines and descriptions, which are then tested to determine which combination performs best. Similarly, using Youtube’s Director Mix tool, marketers can harness machine learning to deploy a variety of video, voice-overs, stills and other ad types in patterns customized to suit specific viewer bases.

Advertisers who managed to craft a strong automated ad strategy started by applying these basic principles. Implementing strong creative to ensure that high-value customers are satisfied, while also branching out via a growth-oriented approach creates a solid foundation from which to establish campaigns that fulfill the potential of machine learning.