Google Data-Driven Attribution Model
How Does the Google Data-Driven Attribution Model Work?
New models are derived to give advertisers advanced information to work around the technique of attribution.
Google Data-Driven Model explained
Data-driven attribution uses the latest techniques of artificial intelligence to analyze the AdWords account. It helps to make each event a reward for advertisers.
This model is composed of the following:
· How many ad interactions
· Visibility
· Ad content
· Other means to find relevant keywords
Google has tried to put a system in place that can gather data for effective campaigns and promotions.
Data-driven is a technique that reads user actions. It keeps a check on which search ads you click on. It analyses this behavior by keeping a track of the click route consumers take. It compares the click route of users who make the purchases with those who don’t. Google’s data-driven model studies the route that made users convinced in favor of the advertisement. The route and pattern of clicks help the Google algorithm understand user behavior. This pattern is used by Google to recommend to various advertisers the path users take. They can judge campaigns of advertisers by their knowledge about consumer behavior and can tell if a particular campaign would be useful or not.
Google’s model interprets clicks from platforms determined by AdWords and not from channels like Facebook or Twitter.
Working Off the Model
Google has provided an example on its own page to better understand Google’s data-driven model.
Let’s consider the example of a tour operator based in the New York City. You can track when customers buy a ticket online. This way you follow the customer whenever a ticket is bought. This is done to keep a record of consumer interest. A user clicks on some ads to get some information about the specific services.
Google uses this consumer behavioral pattern to identify a link between ads viewed and product purchased. This can be explained by taking an example; say ‘Biking in New York’ as one and ‘Biking near the coast’ as the second ad. It is these ads that are tracked to see which one is viewed first and how does this change the users purchasing interests. There will be a pattern, people who view the ‘Biking in New York’ first and ‘Biking near the coast’ later will be more prone to buying a ticket then if the second ad was viewed alone.
The model could then decide the setup for ads and suggest a design that defines the effective clicking behavior for landing a purchase. This could tell consumers about the value each ad could bring to getting more sales.
Expert’s Verdict
To test the Google Data-Driven Model, reactions of people from different walks of life were collected.
Kate Lynch works at CPC Strategy and is responsible for their advertising. She changed from the ‘last click’ approach to Google’s ‘Data-driven’ approach.
Kate changed her technique in May and it must be borne in mind that this change in sales does not happen overnight. So even when Kate changed her settings, it wasn’t until September that she saw a shift in sales.
Kate initially noticed that her off-brand promotions were doing well which could not see with her ‘Last Click’ model before. The previous model made her question the validity of the non-brand campaigns. She is quoted to have benefitted from the data-driven methodology as it led to getting more customers to shift to the brand. This made her restructure the budget to allow a 171% increase in off-brand campaigns. This resulted in more customers shift to the brand and hence increased the revenues by 180%. Kate stressed on the importance of the data-driven method because it studies the whole process of consumer actions rather than just the last ad analysis.
Data Requisites
For the data-driven method to work, you need to provide a good amount of data. The amount of data needs to be around 15000 clicks and a minimum of 600 shifts in 30 days for the brand.
Time frame
Google can make a model once the minimum data requirement is fulfilled. The data-driven method works for each conversion or shift and hence it could be different for one website then the others even if they come under the same umbrella.
Maintenance of data
The requirement of conversions and availability of data does not stop once you start using Google’s data-driven model. Google requires a specific number of clicks and conversions for the model to study consumer behavior. Hence, the model must continue to get sufficient data from clicks and conversions. If the conversions drop below the required level, then Google will automatically shift you to the normal model. It is, however, advisable to keep checking the settings also to see if a problem regarding the model develops.
Other Models
Other models that don’t require data are:
· Last AdWords Click Attribution Model- It is for when users view one ad only. This one ad that the consumer clicks gets full credit for the sale
· First Click Attribution Model- The first click out of all gets all the credit for sales.
· Linear Attribution Model- Each point of click whether its email, social media, search engine, main website share the sales credit
· Time Decay Attribution Model- The last click before conversion gets credit.
· Position Based Attribution Model- The first and last interaction get a 40% share. The intermediary contact gets 20% credit for the sale.
Criticism
Google’s attribution model faces two criticisms:
1. The data needed as a pre-requisite for this model to work is too large. It is hard for new companies to get a lot of traffic, and hence; this attribution model does not work for them.
2. Google’s model interprets clicks from platforms determined by AdWords and not from channels like Facebook or Twitter. With the huge prevalence of social media, it must be incorporated in some way to reflect its influence on the process.
Useful Links:
- https://www.cpcstrategy.com/blog/2018/02/google-attribution/
- https://www.seerinteractive.com/blog/google-attribution-explained/
