SEO
 

As we saw in the previous post about the role of artificial intelligence in search engine optimization , Google is relying on “AI First”. It is not surprising that this also influences the SEA sector, after all, a significant proportion of the company's revenue comes from Google Ads revenue.

Automation has played an important role for Google for years, the use of machine learning and AI systems is the logical next step in this development. In this article, we will shed light on where this is already being applied and what it means for the future of the SEA channel.

Smart bidding in Google Ads

Machine learning has played a role in the development of Google Ads for many years, for example in the calculation of the quality factor. As early as 2016, Google presented the bid automation based on machine learning with Smart Bidding for Google Ads and Google Search Ads 360 (formerly DoubleClick Search).

Anyone who manages a Google Ads account of a certain size has known for a long time how Google tries to motivate advertisers to automate their accounts. Instead of submitting bids manually, Google would like to take over bid management itself based on certain KPIs.

Google is driving this development forward with Smart Bidding. Bids are automatically adjusted based on conversions. The system analyzes a wealth of different data and recognizes, for example, patterns and similarities between landing pages, ad texts or keywords.

The optimal bid is determined on the basis of a large number of signals, including the device or the location.

Google Ads bid strategies

Campaigns in Google Ads can be automatically directed towards different goals. In addition to existing conversion data, the Google bid algorithm uses machine learning to gradually increase the performance of the campaigns.

Target CPA / Target CPA (cost-per-acquisition):

Google tries to control the campaign in such a way that the targeted CPA is not exceeded. The CPA expresses how expensive a conversion or acquisition can be on average.

Target ROAS / Target ROAS (return on advertising spend)

This bid strategy takes into account the value of the conversions achieved (in an online shop, for example, the shopping cart value of the respective order). The algorithm tries to optimally adjust the bids of the campaign to the target ROAS. The aim is to achieve maximum sales (a maximum conversion value) taking into account the ROAS target.

Maximize Clicks:

With this setting, Google will try to achieve the maximum number of clicks from the specified campaign budget.

Maximize conversions:

Here, Smart Bidding tries to get the maximum number of conversions out of the campaign budget. Similar to target CPA and target ROAS, machine learning and artificial intelligence are used to optimize the conversion rate as high as possible.

Alignment to search page position / Target Search Page Location:

This option is good for branding campaigns as the search page position helps ensure the visibility of the ads. The display is only limited by the campaign budget. Possible target positions are the top positions (1-4) and the first search results page.

Competitive auction position / target outranking share:

In this case, position control is made dependent on the competitor's ad rank. The algorithm will try to place its own ad above the ad of the competing domain.

Desired share of possible impressions / target impression share:

The percentage of possible impressions expresses the percentage of how often your own ad was shown for a keyword and sets this in relation to the absolute number of possible displays. An impression share of 50% means that your own ad was only shown for every second search query. This percentage value can be specified as a target with the bid strategy.

Auto-optimized CPC:

This option gives Google control over an automatic bid adjustment of up to 30% upwards and up to 100% downwards (based on the maximum CPC bid). This means that the bid for clicks with a higher conversion probability can be up to 30% higher than the bid actually entered.

If, on the other hand, the system uses semantic signals to calculate a lower probability that a conversion will be generated for certain clicks, the bid can be reduced by up to 100%.

The ad is therefore shown less often or not at all. This function should increase conversions and use the budget more effectively.

Automation offers many opportunities, but it is crucial to understand the system behind it and to be aware of potential problems. And finally, the question of the conflict of interest arises when Google determines and adjusts bids and expenses itself. Do you really want to completely surrender control over this?

Attribution: data-driven model

Automation and artificial intelligence are not only used to control the bids. Where previously various firmly defined attribution models (first click, last click or the so-called bathtub model) were used, Google enables the data-driven calculation of the conversion value in Google Ads.

Based on performance data, the system learns to assess the value of the various channels and continuously develops the attribution model independently.

For this purpose, conversion paths are compared with one another, patterns are recognized and probabilities for the conversions after individual clicks are calculated. However, a fixed amount of data is required, which is why data-driven attribution is not available to all advertisers.

Since users only see the final result of the calculation, it remains unclear how this comes about. So we have to rely on the system to determine correct probabilities and that their accuracy depends heavily on the amount of data available.

And ultimately, that Google acts in the interests of the advertiser. Again, this is a possible conflict of interest. Large companies that provide sufficient data and resources and are already building AI-supported attribution models themselves have the advantage here. Smaller companies do not have this option and have to rely on the Google system.

Further areas of application for artificial intelligence


Artificial intelligence can also be used for optimizing ads in the display (and affiliate marketing). One example is the company Jivox that introduced a technology called "neuron" in 2016 presented , which is proposed by machine learning ads.

Recommendations are generated on the basis of a collection of data that contain, for example, information about which color a certain user prefers or how he reacts to animated advertisements. Also media agencies such as Havas Media , in order to be able to buy media and produce content more efficiently in perspective.

In the future, everyone could see exactly the banners that they are most likely to click (and then convert). Personalization in perfection and maximum efficiency.

Challenges: The SEA Manager of the Future

Where is the trend towards automation for online marketing agencies and SEA experts? What does the SEA manager's professional field look like in the future? At least Google is pushing for less manual optimization and more automation. Setting options disappear or are restricted, the control of the SEA Manager is increasingly disappearing.

The operational work on Google Ads accounts will be less, maybe it will even disappear completely, so that in the future Google will manage accounts independently and continuously optimize them in a self-learning manner: for the greatest possible (cost) efficiency for advertisers and the highest possible profit for Google itself.

But there is still a long way to go before full automation. Manual optimization is and will remain very important in the short to medium term.

In particular for new accounts for which no data is available, there is no alternative to manual setup and support. This also applies to certain scenarios such as seasonal fluctuations or other predictable peaks. Even an intelligent system will not be able to replace the work experience of an SEA manager in the near future.

But what can already be determined:


As in the SEO channel, the role of the SEA manager lies more and more in strategy development, in understanding the channels and their interaction. Even with AI-based solutions and increasing automation, experts are and will remain necessary who understand the system and can "train" it so that the results are reliable.

Here, too, the fear is unfounded that the intelligent machines could cost us online marketing experts our jobs in the future. More and more operational processes will run automatically. But the people who understand and monitor all of this remain important. What is needed are qualified digital marketers who no longer see themselves only as experts in a particular channel, but who can demonstrate a comprehensive understanding and experience of online marketing.


Post a Comment

Previous Post Next Post