2020 has been the year that we could say Smart Campaigns have surpassed manual-powered campaigns. Performance of Smart Campaigns will almost always be better than the performance of manually managed campaigns (given that the algorithms had enough data to learn from). And besides all that, it’s easy! Smart Campaigns drive themselves and will improve their performance as time goes by.
Some fear that this triggers the end of the job of PPC managers. However, there is still a lot to do to improve the campaigns as Smart Campaigns have significant weaknesses. Here are some reasons why smart PPC managers can still overcome the power of machine learning.
Quick Links
Machines are stupid and can’t think
Smart campaigns don’t share information
Machine learning can only work towards one goal
No differentiation between remarketing and new users
Smart campaigns are cheaters
Often people compare Smart Campaigns directly with manually managed campaigns. The campaign that results in the highest ROAS will win. However, this is not fair. Smart Campaigns are using a wide variety of data to adjust the bids in real-time at the moment of the impression. But some of this data is not even available to humans.
Here are examples of variables that AI can adjust the bids for:
- The competitors’ product prices in the auction (in Shopping ads)
- Location intent, not only the location
- The ad characteristics, such as the messaging of the ad
- The Search Partner
- Web placement (can modify bids, not only exclude)
- The recency of remarketing lists (while humans can modify audience lists, the modifiers would count for the full list, non-dependant on the time a user was added to the list)
- The search query (humans can only bid on the keyword, but not the matched query)
One may wonder, would AI still be dominant if it wouldn’t have had access to those additional variables?
Machines are stupid and can’t think
While machines have a good memory, they are not able to “think.” This is very well shown in simulations where deep learning algorithms start to learn how to perform a task. For example, when teaching an AI how to play a video game, the AI will probably just be running into a wall. It often takes a huge amount of “evolutions” in the algorithm before the AI even starts to understand how to move a character. On the other hand, if you give your grandmother a game like Mario, she will probably already in minutes understand how to move the character.
In Google Ads, the ability to think is a huge advantage. It allows PPC managers to make assumptions on what will work and what won’t, based on logical thinking. If I’m selling, for example, women’s clothes, I know that I can exclude men in advance. Without guidance, an AI would have to test this first and only find out after spending a significant budget that the male audiences do not convert.
A widespread example of this is holiday days. Human PPC managers will often make adjustments for such days. However, by the time automated bidding realizes the differences in web traffic, the holiday has often already passed.
Smart campaigns don’t share information
Another way PPC managers can provide an advantage to a client is by learning from the insights and applying those learnings to the Google Ads campaign and the Business marketing strategy as a whole.
Unfortunately, Google chooses to make Smart Campaigns non-transparent. The Search Terms report is one of the first places a PPC manager uses historical data and improves the campaign. However, Smart Campaigns don’t display Search Terms from Search campaigns and placements in Display campaigns. From the Search Terms, PPC managers can evaluate the intent of the user and catch errors. For example:
- Words as “free” in the Search Terms, may indicate the user is not ready to pay for a service.
- Words like “study,” “books,” or “vacancy” may indicate that the user is interested in the field, but not to buy the product. For example, “study for hairdresser” will trigger the keyword “hairdresser” but will not reach an audience looking for a haircut.
- Words like “who,” “what,” “where,” “why,” “when,” and “how” indicate that a user is early in the decision process and is still asking, for example, “How does an ERP system work?”
Such critical sources of information can be used by marketers to make adjustments to the keywords, creatives, or even landing pages of the clients.
Machine learning can only work towards one goal
Probably in the future, this will change, but for the moment, deep learning algorithms need a specific “win condition” to evaluate their success. Machine learning works so that different execution variations are being tried, and the winning combination will be taken into further evolution. It can determine a winning variation based on how well the algorithm manages to achieve its goal.
The problem with this is that a marketer may be interested in a wide variety of objectives:
- The direct profitability of an ad
- The brand lift generated by an ad (increase in name recognition)
- The change in brand perception
- “Brand safety“
By bidding manually, one can take into consideration those several objectives.
PromoNavi’s guide on how to get more traffic on a lower budget
No differentiation between remarketing and new users
This is one thing that Google is looking to change, but for the moment, it’s a huge downside of Smart Campaigns. Remarketing is generally always separated from other campaigns as the objectives vary from “regular” ads. One reason is that remarketing is meant to target the user who has already visited the website. This user has already had some degree of previous awareness of your brand, which may have been generated by other paid touchpoints.
In Smart Campaigns, there is no option to control the ratio between remarketing and “regular” ads, which can be devastating to a campaign’s success. While the remarketing share of the audience may typically convert better, the risk is that the AI will shift impressions unreasonably much towards remarketing. That results in a decrease in reach and overserving (and annoying) the remarketing audiences.
Conclusion
There is a threat in using Smart Campaigns, which is that control is virtually non-existent. As human managers are giving up their control of the campaigns, more control is being switched towards Google’s hands, who has a long-term objective of increasing its revenue, rather than decreasing the ad spend.
While Smart campaigns and smart bidding algorithms are very effective in achieving the goals, there are still many ways human PPC managers have an edge. For this year, for sure, human managers play a crucial role in controlling the algorithms to apply them sensibly to the campaigns’ benefit.
