Using Business Insights to Complement Automation and Optimize Search Strategy

Sign up for the Street Fight Daily newsletter to get local and digital marketing insights every morning.

Search engines continue to release new tools and capabilities that allow marketers to use more automation than ever before. Options like auction-time bidding and responsive search ads have taken on optimization tasks that were once more manual. 

With automation leveling the playing field, marketers need to find new ways to use their proprietary business intelligence to keep an edge over their competitors in search. Marketers can add immense value to automation, maximizing its ability to drive strong performance.

Providing Supply-Side Information through Inventory Data

Machine learning-based algorithms are great at understanding the demand dynamics for a given product or category, but they don’t know about the supply side of the equation. Marketers, on the other hand, are experts on their own products’ sell-through rates and the outlook for future availability. The importance of this type of business data became especially clear during the pandemic, when increased online demand across categories left many retailers with fulfillment issues and supply chain challenges. 

Brands can use inventory data in a variety of ways to improve performance in search. At the most basic level, they can determine which products are selling quickly enough through unpaid channels to not need ad spend support and pause paid advertising for those items. More advanced marketers can also incorporate inventory data into bidding, creating structures that prioritize SKUs based on their current or future projected availability. This human intelligence sets up a system that enables automated bidding platforms to maximize demand capture within the bounds of a business’ supply-side constraints.

Defining a North Star for Bidding Algorithms

Creating a structure that informs bidders is just one way that humans and businesses set an algorithm up for success. Another critical piece of the puzzle is defining the right conversion metric. A machine learning-based system that trains itself is only valuable if it’s training against what the marketer or business truly cares about.

Brands can set a bidding algorithm up for success by carefully and clearly defining a north star for performance. This requires consideration of all actions that matter to a business that a search ad could impact. Too many marketers still use performance targets that focus purely on online revenue. At the very least, any omni-channel retailer should incorporate offline/store visit metrics into their conversion goal. 

Beyond that, marketers should look at what other actions are valuable to their business. Do phone calls drive purchases of higher-ticket items? What is the lifetime value for a customer who signs up for promotional emails? And what about new customers – is there a willingness to sacrifice some performance efficiency when the order comes from a new customer versus an existing one? Defining the appropriate success metrics is necessary in order for machine learning to help brands reach their desired outcomes.

Injecting Business Data into Ad Copy

Ad copy is another area where automation continues to play a larger role in search. Earlier this year, Google announced that responsive search ads would be its default ad copy type in the Google Ads UI, signaling its intent to continue leaning into automation. Like bidding, with ad copy automation, businesses need to give the algorithms the information they need to train against. What is the brand’s value proposition to users, and how can that be communicated in different ways through ad copy? Differentiators should extend beyond promotions and discounts (though those options should certainly be included) to consider what the brand offers to a customer at different points in their shopping experience.

Once differentiators are defined, businesses can tailor copy based on where they perceive a customer to be in their journey. This could be keyword based, with higher-level keywords, like “best smart TV,” treated differently than “55-inch LG TV.” It could also be audience based, with cart abandoners receiving different value propositions than folks who only visited the home page. Once again, the key to enabling success is setting up a structure that shows relevant copy variations to users and allows the algorithm to learn about them.

Marketers can also leverage their business data to create more customized ad copy for different users or categories. Tools like ad customizers in Google Ads allow for text ads to be tailored based on a user’s search, device, or location. Search Ads 360 users can use its business data tools for additional ad copy customization to plug in different value propositions at the category, product, or brand level. 

Building First-Party Audiences for Bidding and Targeting

Auction-time bidding already takes many data points about a user into account with its algorithms, but marketers can create new business-specific signals for it with first-party data. For example, marking a user as an email subscriber or a frequent customer gives the bidder new information that it would have no other way of knowing. While it may not understand what the difference is between a subscriber and non-subscriber, it does look at all of the audiences a user belongs to and uses that as a signal in bidding. This applies for customer match audiences as well as remarketing lists built using on-site tags.

First-party audiences also remain important for targeting purposes. They can be used for ad copy customization, as mentioned above, to give customers tailored messaging. They can also be leveraged in campaign structure, which we’ve already identified as a key element for setting bidding algorithms up for success. With the third-party cookie crumbling, it’s important to have clear, defined first-party data sets that reflect the unique attributes that are valuable to an individual business.

Conclusion

There is a misconception that automation is a replacement for human activity – we don’t believe it is. We see automation as a powerful supplement to human activity, so humans can focus on the elements only they are capable of understanding that help drive success. 

Success with automation requires active management, and that begins with thoughtful definition and articulation of goals that automation and machine learning can optimize against. Humans working in concert with automation can create true differentiation in performance and customer experience across paid search.

Paul Koch is SEM Senior Director at Merkle.