The Future of Search Marketing & Machine Learning

The infinite goal of search advertising, and in fact all digital advertising, is to serve the right content to the right people, at the right time. Every development takes us further towards that goal, of which there has been many throughout the last decade. To make sense of this digital and information age, humans have created incredibly intelligent search engines, helping us access the content and services we seek out daily. 

Humans, however, are slow and likely to make mistakes compared to our machine counterparts. In today’s age of connected devices, there is an exponential growth in data. In fact, Cisco forecasts there will be 4.8 zettabytes per year by 2022, which is over three times the volume from 2017. As a result, advertisers increasingly rely on machine learning capabilities to make data-led decisions.

Machine learning is a means by which to achieve artificial intelligence and is essentially algorithms that are based on statistical modelling. In operation, they make decisions (or predictions) determined on probability, based on the strength of patterns found in data. The machine learning algorithm becomes more ‘intelligent’ with the amount and accuracy of data it has access to. When we apply this process to real world problems, such as targeting or reporting on digital marketing, we often talk about automation. 

Automating tasks frees up time and money, so the incentive to do so is of great economic and societal benefit. Increasingly, manual working of data is not only inefficient, but results in a different outcome compared to when you make use of automation. Take bidding on search terms for example, there are now an extensive number of contextual signals, which Google uses, in addition to the search keyword itself. 

Many of these are very dynamic and only useful if analysed ‘on the fly’, which is why advertisers are increasingly using tools like Auction Time Bidding. A recent survey of those trialling it saw an average conversion lift of 15% to 30% at the same or better ROI. The implication for consumers is a step change in how relevant paid search results are. Connected to this, is the implication for advertisers, serving content to more of the right people, at the right time.

Bidding within an auction mechanism is one major area where machine learning is resulting in better outcomes, however it is not the only area. Relevance also relies upon the effectiveness of results or content served. Two of Google’s solutions to solve this challenge, built on principles of machine learning, are Responsive Search Ads and Smart Shopping Campaigns. As of last year, only 24% of search marketers were using Responsive Search Ads, in response to an industry survey.

Compared to bidding, there is a lingering perception that we, as humans, are still more effective at the creative aspect of search marketing. It would be foolish to assume this notion is universally false, however it highlights the next area of opportunity for technology and advertisers to fully embrace machine learning. Dynamic Search Campaigns help to address the challenge of longtail search behaviour, since 16% of all queries each day are completely new and have never been searched for previously.

eCommerce advertisers have been the first to embrace such developments, due to the size and complexity of their search accounts, reflecting the broad product catalogues many have. Almost all eCommerce advertisers should look to take advantage of opportunities like Smart Shopping campaigns. This recent innovation, using Google’s machine learning, shows dynamic product listings to in-market shoppers across Google’s different ad networks including YouTube and Display. 

The implication for advertisers is the opportunity to drive far greater performance from investment, with optimisation no longer relying on multiple single campaigns across different networks working in silo. Consumers, on the other hand, are starting to see more shoppable product ads when browsing sites and watching YouTube, often directly relating to a product review they are watching, for example. 

Many aspects surrounding the future of search will be powered by machine learning. This presents advertisers with a new challenge, which is understanding where and when to take advantage of these tools, which will be different in each case. Take the finance and insurance verticals, where CPCs are high and competitor tactics can dramatically influence cost (even when they affect only a few keywords). Machine learning bid strategies can work, but need to be granular, avoiding situations where the machine suddenly pauses large swathes of activity. 

Other verticals like retail eCommerce, where price and availability are dynamic, can benefit from leveraging ad automation. For example, home retailer Bygghemma used Google’s Ads API to connect their own machine learning tool to ad customisers. This enabled dynamic adaption of both promotion messaging and landing pages, increasing the relevance for each search query, therefore delivering a stronger conversion rate.

Teams exploiting such automation are redefining the role of a search marketer. Focus is increasingly on the implications machine learning has for best practice account structure. Plus, where human optimisation is needed, there is a new focus on improving the quality of data inputs rather than interfering in the optimisation itself