AI and PPC automation are certainly very trendy concepts. Their popularity will only grow – but is there a limit? Will they take over? Their attractiveness will vary depending on your job title and your budget.
From this article you will discover:
- The different types of AI and automation, and where they can be implemented for greatest profit.
- How to setup and structure campaigns for greatest integration of AI and automation.
- Tailoring their use to your individual campaigns, for long-term ROI.
Degrees of “automation” – not to be confused with AI
An important technicality here is the different levels of automation available because some are better at certain tasks than others. This is important because a client might be expecting full-blown AI when in reality, they are getting something a lot more basic.
Google Ads and Bing Ads have a built-in feature called Automated Rules, which allows search marketing executives to control a range of basic tasks as often as once a day. For example, you could set bids to be reduced by a percentage if there is high spend and conversions are low, or vice versa. It’s quick and useful, but care must be taken to ensure effective rules are set up properly. When used properly, this is an efficient way of managing some tasks. However, we’ve seen account performance decline because of automated rules being used ineffectively. Automated rules are basic and cannot ever resemble a fully automated account.
Google & Bing scripts are a lot more powerful than Automated Rules because they allow the user to manage a wider range of tasks more regularly. At Fountain, we have in-house scripting, which means that we can create bespoke and complex rules with very specific outputs, built around our clients’ needs. But this is not AI because we humans are still providing conditions for actions to take place.
APIs (Application Program Interfaces)
API’s are more widely available than scripting interfaces and allow a more direct and frequent communication between a platform (eg Google Ads) and another server. Some example systems that use API integration would be third-party PPC management tools, phone call tracking, or even your CRM. API’s are also often available on platforms where scripts aren’t available, such as Facebook and SEMRush.
Full-blown AI (that is to say, an artificial intelligence managing all aspects of your PPC account) could use an API to communicate effectively with Google Ads. However it is worth noting, that APIs may not support all features of a given platform; that is to say that Google Ads may not allow an API to interact with all the potential optimisations available from Google’s own interface.
Artificial intelligence is defined in the Oxford dictionary as, “The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”
Therefore, neither automated rules, scripting, or an API (by itself) can qualify as AI because they aren’t making a decision that would normally require human intelligence. Instead, what we are doing here is just creating a set of rules and associated actions.
Therefore, automation is only as good as the human who created it. It also doesn’t mean the account management is fully automated. However, we argue that there is a balance to struck between humans and automation.
When should automation be used?
Search Term Reports
A search term report allows an executive to review what users typed into the search engine before clicking on an ad. This is an incredibly important report which can make a huge impact on your account’s future success.
Google and Bing show ads when someone has searched for either your exact keyword, a close variant of your keyword, a misspelling, or even a word related semantically.
Take for example Jeremiah who sells park gates. Jeremiah targets the keyword “park gate”. This means that his ads will show for searches for “park gates” as well as searches that closely resemble “park gate”. However, there is a place called “Parkgate”, so his ads will also show for searches related to the location.
This is less than ideal. It will affect click-through rate (CTR), quality score, cost per sale, as well as Jeremiah’s company image. Any type of automation will [likely] require a certain amount of money to be wasted before noticing this type of thing, whereas a human can utilise search term reports, quickly identify this as an undesirable search term and then add “Parkgate” (without a space) to the negative keyword list.
Depending on the complexity of the automation, we would expect most systems to recognise poor performance and simply reduce the bid for the keyword “park gate”, thereby missing business opportunities and potential revenue. In this example, a human executive taking a more analytical perspective can help generate more revenue for the business in the long run.
On the other hand, automation can be highly effective at recognising and highlighting search term “ngrams“. This is where two or more words can appear in a search term which, when appearing together, produce a pattern of performance. For example, an ecommerce gardening business might be targeting the keyword “indoor plants”. There could be 1000’s of search term variations in the search term report, including things like “uk”, “online”, “large”, “artificial” and “best”. An algorithm could be highly effective at recognising that when both “near me” and “artificial” appear in the search term, the cost per sale increases. The algorithm could then flag this to the executive or automatically add “near me artificial” as a broad-match negative keyword.
In certain situations therefore, automation proves and important tool. But, intervention from an analytical human utilising search term reports and negative keyword lists can save money, protect the company’s image, and avoid missing out on future opportunities.
Creative for Ads & Extensions
It is certainly true that automation can free up the much-needed time to allow executives to focus on other tasks like creating great ad copy.
Ad copy should be engaging, creative, and reflect the target website’s messaging and USPs – all whilst standing out from competitors. It’s a crucial touch-point for users and poor ad copy will hinder an overall performance.
Google has recently launched “responsive search ads”; these allow an executive to load in a large number of different phrases at once, and Google will automatically test different combinations of these phrases in ads. So far, we have seen mixed results and several accounts have received a lower CTR from these types of ads. This is perhaps due to our manually-created ads being finely tuned, with deliberate links between the phrases in the ad titles and ad descriptions that we know make sense. However, we also recognise that responsive ads could be a particularly time-saving feature for new accounts where there is no previous data on ad performance.
We are starting to see Google utilise their AI to come up with what they call “ad suggestions”. According to Google, these are created by their machine learning and reviewed by people before becoming available on a Google ads account.
Sometimes the suggestions take an old ad and swap around sections of text, and we’ve noticed that sometimes they add a little something extra, like “Shop Now!”. Performance is mixed; we have seen ads that improve CTR and ads where the CTR tanks. Sometimes they are not worth using: many are not up to our standards at Fountain, and others are not specific enough to the product or service.
The auto-application of ad suggestions (the default setting) can also cause issues with businesses who are protective over their brand image and sign-off process, but it can be disabled.
We suspect that Google’s ad suggestion capabilities will grow, and their quality will improve.
We are also aware of other companies developing their own automated ad-creating solution; Fountain are no exception and we are currently developing our own solution to assist our account managers. However, we can’t be certain that machine learning will be able to provide fresh, original, innovative ad copy until we see it.
Account Setup & Structure
The structure of a PPC account – that is to say, how the campaigns, ad groups and keywords are structured – can make or break the account’s performance.
You’ll want a nice balance between not-too-segmented and not-too-ambiguous in order to quickly collect the most meaningful data and identify patterns. An account’s structure should vary depending on whether or not automation or AI will be used.
With the present level of technology, we believe that experienced humans need to be involved with an initial PPC account creation to strike this balance with the structure.
Whilst we continually experiment with automation, we have so far found that it either over-segments the structure (which means it can take longer to accrue data and identify patterns) or it clumps together too many keywords that aren’t closely related (which means the ads get fewer clicks and the client gets less revenue).
There has been a growth in the marketplace of organisations boasting a purely automated approach – but if there is minimal human interaction in these organisations then the need for employing highly skilled and experienced executives diminishes. Without this skilled human resource, how can the account be setup in the most effective, balanced way?
Automation should also take into consideration the account and campaign settings. For example, if device bid-adjustments are actively automated on a campaign using an enhanced-CPC strategy, you could actually decrease performance.
The account’s conversion attribution model also needs to be up-to-date and suited to the automation’s goals. If your audience typically has multiple touch-points with your PPC campaigns and you aren’t attributing conversions appropriately, automation could reduce overall account performance. This is because it looks at the qualitative data and, for example, bid down on “discovery” keywords and increase bids on brand terms, because it doesn’t know better. Therefore, it is important to ensure the account is set up in a way that is compatible with automation.
In our experience, automation works best with large accounts with a granular set-up , typically separating out one keyword and match-type per ad group (known as SKAGs – single keyword ad groups). This is because automation can analyse large amounts of data quickly. This works well if you have a lot of traffic going into each ad group, and each ad group has effective negative keywords to help “direct” traffic. Without directing traffic with negative keywords, you might shoot yourself in the foot by reducing the bid so much for a low performing keyword that ads start to show from another ad group – for the same search term!
Comparatively, if you have a selection of ad groups each targeting keywords that don’t generate much traffic, automation could struggle to optimise because of lack of data. However, a human could identify that those ad groups contain similar keywords, and identify patterns at a higher, less granular level, therefore being able to increase performance where automation could not.
Before setting up a campaign, an executive should perform keyword research to get an idea of what they will target, how they will structure their account, as well as forecasting potential costs.
There are a variety of tools available for keyword research. These tools allow an executive to tap in a few key phrases and the tools attempt to spit out a bunch of related phrases with monthly search volumes and costs.
Depending on the input, around 80% of the “keyword suggestions” are unrelated and useless and would require a human to comb through the long list. However, we would argue that this is a really important task because even though the majority are not related, we often stumble upon highly relevant keywords that we would never have come up with ourselves.
We wonder if an automated organisation is going to take the time to comb through keyword research suggestions, to identify those hidden gems?
Additionally, human behaviour changes over time. People start using different language to find the same thing. Without performing frequent keyword research, a business could be left behind and overtaken by their competitors. Unfortunately, this often goes under the radar if the account is performing well – particularly if the client’s industry is in growth. But maybe, it could be performing even better.
Our experience with advanced automation platforms is that they are extremely reactive, making adjustments as often as every 15 minutes.
This has its advantages, for example, if there is increased demand for a product, like a coat in cold-snap or memorabilia when England win the World Cup.
On the flip-side, it’s important to consider how much detail the automation is considering, and does it have all the available information?
• How much time does a user take to convert from their first click?
• Has the website changed within the data’s date range?
• Have new negative keywords been added in this date range?
• Have the device bid adjustments changed in the date range?
• How about audience adjustments?
• Which other changes have taken place in that time?
Recently, we are seeing more advanced automation (like the latest updates from Google’s ROAS bid strategy) that consider things like seasonality – which is great. But even this strategy needs to be put on hold and switched over to manual if there is a change in conversion rate, otherwise, you put your long-term performance at risk.
An account should continue to grow with new campaigns, ad groups and keywords even after its launch.
New opportunities can be identified through competitor analysis, new products or service offerings, search term reports and keyword research. As we have previously mentioned, revenue can increase if the demand for a given industry is in growth – leading us to believe that the current strategy is working. But at Fountain, we believe that account growth that is only in-line with the industry’s growth is not growth at all. Identifying new opportunities is a complex task – and automation as it stands simply cannot consider such complex tasks.
One size does not fit all
There are different types of automation. Google Ads contains its own free automation that comes in the form of bidding strategies. This can take the pressure off an executive who has an obligation to frequently check keyword bids against performance. But they don’t always work, and some perform better than others.
There is no one-size-fits-all solution from Google, and compared to third-parties, Google’s bid strategies have access to more information, such as a user’s profile for each search. So how can we be confident that third-party automation or AI software will get the best performance out of the account?
To find out which Google bid strategy is most effective, it is important to test them all.
Campaign experiments allow us to A/B test one bid strategy against another with a 50/50 split of traffic. When one test ends, we test another bid strategy. This is the only way to know for sure. And those tests need to be created and monitored by a human.
Whether or not we realise it, automation is creeping more and more into the accounts that we manage. Features like ad rotation and auto-ad suggestions often go under the radar. Executives are increasingly encouraged to use more automation and AI, particularly with solutions like bidding strategies and the recommendations report.
Once we understand the types of automation available, we can identify and test that technology on a case-by-case basis.
There is certainly not one type of automation or AI that will get the most out of any account due to factors like account structure, the volume of traffic, industry and PPC settings.
Google is now offering Smart Display and Smart Shopping campaign types. We have seen smart-campaigns increase performance, but executives have very little control over them. It’s not possible to exclude a particular product or exclude remarketing lists which is problematic. These results-driven campaigns don’t allow a marketer to increase the budget to a demographic, and therefore might be incompatible with an advertiser’s goals. Of course, the benefit of using these campaign types (along with any third-party automation tools) might outweigh the negatives.
But we have also highlighted that not everything can be completely automated. A successful account that utilises automation still needs be set-up and monitored by an experienced PPC executive, who will ensure that the account structure is appropriate, monitor the consistency of conversion rates, monitor external factors like weather and sports events, monitor search term reports, and be sure to facilitate the account’s growth with new keywords and ad groups.
Additionally, overall performance must still be monitored, as well as the individual performance of multiple campaigns that a user may interact with as part of their journey to a conversion. We have witnessed first-hand a bug in automation software that led to a drop in revenue and an increase in cost per sale. It was only because of our close attention to detail that we were able to identify and address the issue.
Fountain continually tests and utilises automation in PPC, seeking to minimise human error and maximise efficiency, freeing account managers to focus on areas like strategy and conversion rate optimisation where human creativity excels. Many routine tasks are not yet appropriate for automation because they rely on the intuitive understanding of audience sentiment and wider context.
Automation therefore shouldn’t be chosen just to reduce agency management costs, because then time is being sacrificed – time that should be used to perform analysis to look for new opportunities to grow the account. Having experienced human PPC executives on your team will be a huge benefit to the successful long-term running of a PPC account.