We live in a world full of data. Each of us generates vast amounts of it, and every day all over again. This happens when we browse the Internet, but also when we shop in a supermarket with an electric checkout system or move from A to B with a cell phone in our pocket. Until a few years ago, this incredible mass of information we leave behind was mostly captured unused, but not really utilized due to a lack of computing power and application examples. However, due to technical progress in the processing of data and decision-making procedures based on them, so-called algorithms, it is now possible to calculate and also use behavioral probabilities. This is precisely where the fullistic approach in "pay per click" advertising comes in, and has thus made significant progress in recent years in increasing conversion rates - i.e. the proportion of sales or leads from paid traffic. But where exactly do these automations take effect and how can I use them to my advantage? Machine learning and algorithms are used in various areas of paid online advertising, which will be analyzed and examined in more detail below.
1. Smart Bidding Bid Strategies
Deciding which bid is the right one for which keyword has been an issue in search engine advertising from the very beginning. Where online marketing managers used to set new values granularly at the keyword level on a daily or weekly basis with a lot of experience, intuition and routine, now every single search query comes into focus. Not the keyword alone, but all the information available beyond that, such as the previous user behavior of each individual searcher, is taken into account when using smart bidding to place bids for each individual search in real time. Smart bidding strategies have now also arrived in social networks as an advertising environment and decide which users should really be addressed.
So instead of using a "one-size-fits-all" approach every day to try to get closer to its target group, it is decided individually which bid fits which user.
2. Target Groups
Even before a user even starts a search or scrolls in his newsfeed, he is assigned to different groups in advance, based on his behavior. These groups range from interests to the presumption of a specific purchase intention in a product area. These can be taken into account both in a customized user approach in the advertising materials or ad texts, and in the targeting of campaigns based on individual target group performance.
3. Ad Creation
Automation has not stopped at ad creation either, creating different degrees. From ads whose text is dynamically fed from the web page content, to ads that select from a number of different pre-set ads the one most likely to lead to a desired action, there are no limits.
4. Keywordmatchtypes / Ad Delivery
Changes due to machine learning also exist in keyword match types, where search engine advertising is used to define how generously search queries are matched with the keywords posted. Here, regardless of the respective match type, the ads are played out on more and more variants of the search queries. This goes as far as matching the content of synonyms.
5. Attribution
Finally, the topic of attribution - the logic according to which conversions are assigned to campaigns - also offers the opportunity to benefit from algorithms. The "last click" attribution model automatically used by Google, in which every conversion is always assigned to the campaign with the last click, has the major disadvantage that measures used further forward in the funnel are often not assigned any conversions and are thus incorrectly dismissed as not achieving their goal and paused. The fact that conversions can be prepared by the measures, however, is not taken into account and so one is surprised afterwards about the decline in sales.
With other attribution methods, it is possible to assign the same share to each conversion, but nevertheless it is not taken into account here how strong the influence of the campaign was on the user's decision. The data-driven model trained by machine learning, on the other hand, tries to map exactly this, thus deriving a more informed action and enabling better analytics.
With all these possibilities, however, several broader questions about the fullistic approach should not be ignored:
First, online marketers need to ask themselves whether they have enough data to deploy the measure in question. Only with a sufficient data basis, which differs from case to case, can added value really be generated here. Furthermore, the large advertising networks, such as Google, Facebook and Co. should not be blindly trusted. Each automation must be questioned, understood and checked for the respective application purpose. Often, the use of these processes requires a great deal of trust from the advertiser, which cannot always be met without prior experience across the board.
The nexts steps ...
The central question should always be: How can I invest my time in a way that adds value? We as experts have had astonishingly good experiences with many of our customers using the PPC Fullistic approach. However, we let ourselves be individually involved with each of our customers and check which measures really make sense and are target-oriented for them. Our diva-e specialists around Andreas Kirchermeier look forward to your call or e-mail.