Keyword marketing has been a mainstay in campaigns since the earliest days of ecommerce. But marketers need to rethink that strategy with the rise of retail media networks (RMNs), which don’t have keyword bidding engines. Fortunately, RMNs have machine learning, which can deliver better ROAS.
The biggest advertisers will grow only incrementally YOY using outdated tactics like keyword conquesting, while self-serve activation with ML enables parabolic growth potential. Therefore, RMNs (and any organization thinking about adding an ad business) should focus on enabling the most sophisticated ML possible to unlock significant advertising revenue.
A Blessing in Disguise?
Retail media is one of the hottest trends in digital advertising, and its proponents insist it will haul in $54 billion in ad spend next yearOpens a new window . This raises the question: Big brands already spend heavily with the biggest retail media players, such as Amazon and Walmart.com, so it seems future gains would be incremental at best. So what is behind the predicted surge in retail media spending?
Industry pundits are spot on. Retail media is incredibly compelling for brands and merchants of all sizes. All marketplaces have plenty of first-party data on their shoppers, and they’re just as motivated as Amazon and Walmart to see their merchants’ campaigns succeed. What’s more, as closed systems, they too help brands calculate ROAS (i.e., say definitively that this consumer saw your ad, clicked and purchased).Â
See More: How to Use SEO and SEM to Boost Customer Experience Outcomes
The growth comes from a broader range of brands and merchants spending in a broader range of marketplaces. Amazon, Walmart, and Instacart are responsible for the lion’s share of online shopping, but they’re far from the only places where commerce occurs. There are hundreds of ecommerce marketplaces across the globe, many of which cater to a specific audience. Merchants are keen to reach and engage shoppers in these marketplaces. What’s holding them back?
Marketers have long relied on keyword marketing (more than 80% of businesses worldwideOpens a new window rely on Google Ads). And yet, most retail media networks don’t offer a keyword-bidding engine, meaning they cannot support a favored tactic of digital marketers. And given this functionality is costly and time-consuming to build, most RMNs remain unlikely to make such a monumental investment.Â
But here’s the upshot: what seems like a shortcoming may actually be a blessing in disguise.Â
The Problem With Keyword Campaigns
For the performance advertiser, the traditional model is to launch a campaign based on a hypothesis of what terms likely buyers may search, measure results, glean insights, update the keywords with modifiers and test again. This iterative approach makes sense for keyword campaigns, as keyword searches don’t necessarily lead to a conversion for that product.
Herein lies the basic challenge with keyword campaigns. Keyword searches can be the start of a purchasing journey, somewhere in the middle or, in cases of highly specific searches, the end. For example, typing “Benadryl†into the Amazon or Walmart mobile app’s search bar may end in purchasing a completely different allergy relief solution, such as an over-the-counter nasal spray. And yet, if Johnson & Johnson had purchased “Benadryl†as a keyword, the brand must still pay for that ad, dragging their media investment return lower.Â
Or, if someone is a regular purchaser of Benadryl, they may search for the term as a matter of convenience. Why should J&J pay to show the consumer an ad when their sole purpose for using the Amazon app is to replenish the stock of their preferred product?
In other cases, a consumer enters a search term, sees an ad and then gets wholly distracted. They may even quit the app without registering Johnson & Johnson’s ad. The result is wasted ad spend on the part of the brand and a confusing data point that a human has to try and make sense of.
Most advertisers and agencies realize purchase journeys take twists and turns. By following the iterative approach, they use a range of negative keywords and keyword modifiers to fine-tune their strategies. But it is a highly manual approach. It takes substantial amounts of both time and data to analyze the journeys that begin with the keyword search, assess where they land, and optimize the keywords strategy based on results.
Machine Learning vs. Keyword Bidding
Machine learning (ML) addresses these challenges because it isn’t keyword-dependent. ML also has a test, measure and update approach; only it is performed in real-time. Plus, it has an added advantage in that it can look at a much bigger range of attributes (“inputs†in ML terms) and predict the likely outcomes (“outputsâ€). The split-second speed of the learning algorithms allows ML to hone in on the true consumer journeys for a particular product much faster than human-developed keywords.
Another advantage of machine learning is that it doesn’t need the parameters traditional campaigns rely on, such as hypotheses regarding who is likely to purchase a product. Building any demographic, psycho-demographic or geographic assumptions into an algorithm is unnecessary. ML looks at the complete universe of available inputs and outputs and uses those insights to predict those likely to purchase if presented with a brand or product ad.
Here’s where things get interesting: combining machine learning with retail media. The platform owners have a trove of first-party data on who their shoppers are and how they behave when making purchases – data machine learning algorithms leverage to decipher the unique purchase paths for products.Â
This means it can identify shoppers likely to purchase a specific allergy remedy even if that shopper never enters “Benadryl†into a search bar. For instance, a shopper can add tissues and Advil to a cart without realizing they may be suffering from allergies. An ad for allergy relief can set off a bell in their heads and prompt purchases.
The inputs machine learning uses to predict outcomes are far wider than the audience and keyword signals. It looks at past purchases of an individual user or user cohorts, past purchase journeys, product mixes in orders and a whole host of signals that may seem irrelevant but play a crucial role. Finding these hidden associations and automatically taking action on them to drive returns is what advertisers want.Â
 By not requiring keywords, machine learning not only automates and simplifies the ability to advertise but also drives performance. Avoid paying for as many ads shown to people who won’t convert. Conversely, show it to shoppers prone to buy even if they didn’t search the exact keyword.Â
See More: How To Minimize the GDPR’s Impact on Your SEO Strategy
 Time To Kill the Keyword?
Advertisers and agency teams often look for a silver bullet for performance, and machine learning is the closest tool available. When it is time to review marketing budgets, dollars are directed to platforms where they best perform.Â
 It’s why Amazon, Walmart, Facebook and other big tech companies invest so heavily in their algorithms. They want their merchants and advertisers to succeed quickly and easily, so they get new and incremental media investments.Â
 To achieve this, Big Tech leverages ML to drive higher performance automatically. This unlocks truly scaled retail media investments, allowing traditional keyword marketing to be laid to rest finally.Â
Do you think keyword marketing in ecommerce will finally die in the coming year? Let us know on FacebookOpens a new window , TwitterOpens a new window , and LinkedInOpens a new window .
Image Source: Shutterstock