Maximising search automation in Google Ads

There are many different approaches you can take to account structures, and most of these are then refined to suit your clients’ business and its goals. But one type of account structure has risen above the rest lately, and flies in the face of certain best practices that online advertising professionals have held dearly for a long time.

Thanks to the constant evolution of artificial intelligence, specifically machine learning, the way to approach paid search account management has changed. There will almost certainly be a Luddite movement within the industry that pushes against giving more control to the machines – and there are questions that need to be answered before simply rolling out these new technologies. However, these concerns can often be addressed with a rigorous testing strategy to validate the changes and ensure that everything is done on the basis of improving performance. Best practices of the past can put artificial constraints on your accounts, hindering their potential. 

This new structure seeks to achieve three things, through layering the latest machine learning technologies: 

  • Maximise audience signals
  • Fully automated bidding
  • Serve the most relevant creative

Whilst these three things may sound like second nature and are often done in most well structured accounts, they often do not take full advantage of the recent and significant improvements to the machine learning available through tools like smart bidding. This increases the amount of time you spend managing the account and reduces its performance. Let’s run through the steps needed to achieve these goals. 

Firstly, account structures should be simplified wherever possible. This will involve the removal of artificial constraints like campaign segmentation by match type (more on this later..) and device. Campaigns should be structured on two things: business objectives and themes. It is still very important to group your campaigns by the objective, such as a specific ROAS target, and the theme, such as running trainers. Artificial constraints like match types, devices and audiences should be removed – all that this type of segmentation will do is reduce the amount of data your smart bidding can learn from, and thus hinder performance. 

At the end of the day, £10,000 in revenue at a 500% ROAS is the same whether it comes from a desktop or mobile device, broad match or exact match keyword – rather than splitting your campaigns into two match types, because you think exact match is more efficient than broad, why not let the smart bidding just optimise to that 500% ROAS target over one campaign? 

It’s important at this stage to note that an algorithm is only as good as the data it is given. The role of the account manager is to set the most appropriate tracks for the train to run on.  Combining match types of similar themed keywords into one campaign is appropriate – the audience you are chasing is going to be similar across all match types, just with differing degrees of intent. Combining different categories of products into one campaign is not appropriate – if you are a multi-channel retailer, combining mens jeans with womens dresses into one campaign is not going to work, it will confuse the algorithm which will find it hard to optimise toward the audience. 

Keyword matching and your approach to match types will also need to change. This is by far the one element of the approach that raises the most eyebrows from colleagues and clients alike. Since the dawn of time, accounts have been structured in a way where match types are split at the campaign level. This has been done to control bids and the flow of users, funneling them toward the exact match variant which typically has the highest bid and quality score. New queries picked up by broad and phrase would be added as exact. 

Over time, keyword matching has evolved from the literal sense to semantic matching, whereby the focus is on the meaning and intent behind a users search. This has been happening slowly over time, for example most recently with keyword variants including synonyms, implied words and those with the same meaning. This has reduced the need for segmentation and also reduces the need for multiple variations of the same keyword, for example plurals, misspellings and stemmings. 

It will also allow you to move to larger ad groups – relevance is important here, as you want to serve the most relevant creative and landing page to the audience. So as a rule of thumb, it should be one landing page per ad group. Also think about the theme of your creative and landing page, it should have the same relevancy to each keyword within the ad group – if it doesn’t apply to certain keywords, then further segmentation is needed. An example of how this might work is included below – notice how keywords with the same intent and audience are grouped together. Even though match types are condensed in this manner, you can still carry out the same optimisations such as adding negative keywords to exclude traffic that is not relevant to your business.

Thinking now at the campaign level, as mentioned before, it now makes sense to combine your match types together into one campaign. This will combine two campaigns, which target the same audience, into one, significantly increasing the amount of data sent to the smart bidding algorithm. This is particularly helpful for smaller accounts, which may have struggled in the past to meet the conversion thresholds required for the Target CPA and Target ROAS strategies – even though Google states there is a minimum 15 conversion threshold, these strategies often work much more efficiently when they have far more – see the table below. 

Fully automated bidding is a crucial element to structuring your accounts for success. Campaigns should be structured according to their business objectives and themes, with these campaigns using shared smart bidding strategies and shared budgets. The benefits are twofold: firstly, you save time managing the account by condensing your bidding strategies and budgets in a logical way, allowing you to focus on other tasks. Secondly, you are fueling the smart bidding algorithms with more data, allowing it to make better decisions which usually results in stronger performance. The table below highlights how this might work in an account.

The final element of this strategy is to ensure that you maximise the audience signals being sent to each campaign. To do this you will need to comb through all the In-Market, Affinity, demographic and similar audiences available from Google, choosing the ones which suit your target audience – you may often find that ones you thought irrelevant actually drive stronger performance for your campaigns. Another thing to do is to ensure that all the RLSA audiences in use are set to 365 days – whilst best practice has usually been to segment them by time, so that you can bid higher for a user who has been on your site 7 days ago instead of 60, this is no longer required. The smart bidding will factor in how long a user has been in the audience list and bid accordingly. 

One of the main concerns when trying this approach is transparency. Clients may be worried about the lack of transparency when giving so much control to the algorithm – and rightly so, especially if the current approach has worked well historically. Any changes should follow a rigorous testing strategy, ensuring everything that is rolled out is generating efficiency gains and growth. It is important to focus on meeting growth and performance targets, and making the changes in the account which work towards these – rather than what has traditionally been considered best practices.

Another concern is how online advertising professionals can ad value, when so much of the job is now being automated. This is another valid point, but any account manager will know that if your time is saved in one area of management it simply gets taken up with another. Freeing up your time from the mundane tasks of bid management and audience optimisation will allow you to focus more on the most important aspects of the account, such as improving ad copy and optimising shopping feeds. 

These steps may seem daunting, but they are at least worth testing, and you will likely see the same success as we have.

Listen to our podcast on this topic here.