How Google Ads campaigns can be optimized to bring their performance to their peak - that's SEA manager’s idée fix over the years. Adference and adSoul have seen a new way to tackle the industry's biggest problem - long tail keyword. The solution takes machine learning approach to the next level: Entity Bidding – that makes the boldest SEA Managers' dreams come true!
What is Entity Bidding strategy?
Google Search is becoming more and more based on the interpretation of the "search terms". In order to interpret the type of search query, Google creates the entity graph, which depicts the meaning and relationship between search terms. This means that in addition to the direct entry of the keyword in the Google search field, other implicit characteristics of behavior are considered, e.g. what kind of device does the user use to search at what time, what is the search history?
With the entity based bidding approach developed by Adference and AdSoul, it became possible for the first time to calculate bids for Google Ads based on the performance of single entities. During the test period, Adference achieved more revenue at 70% higher ROAS than Google's Fullistic Approach including Smart Bidding strategies.
The Promises of the Entity Bidding Approach:
- Activation of the long tail keywords
- More traffic
- Overall better performance than with Google's best practice model
- More sales with higher ROAS
Step 1: Fragmentation of the search terms into the entities
AdSoul handles keywords as the basic building blocks in SEA and breaks down search terms into 34 million units of meaning with the help of Natural Language Processing , which creates a noticeably infinite granular decision level for optimization.
The automatically detected entities are all stored in the database - which is not only the basis for the creation of ad texts but also the feed for bidding by Adference.
Step 2: Entity Bidding
From the algorithmic core of Adference a new offspring is born 🌱- our determined R&D team succeeded bringing the optimization from the keyword to the entity level. Important: Besides the meaning of words in the form of entities, many other factors play a role around the search terms: keyword length, number of terms, match type, numbers, and much more. They all influence the expected conversion rate.
How is bidding in line with that?
The assessment of the keyword "buy Nike shoes" is not only based on the actual performance of the keyword but after decomposition of the keyword into its meaningful components:
Entity → Entity type
Nike → Brand
Shoes → Product group
buy → Buy word
The expected keyword performance is the sum of the entity performance.
Previous forecasts with aggregated data
Adference uses information from all keywords, all campaigns and all accounts of the client in which the respective entities are also included. This allows reliable estimations to be made and the bids to be placed successfully, even for long tail keywords or keywords with few or no clicks.
For instance; let's assume there is little or no data for the keyword "Nike shoes mint 38". This is no longer a problem, as there is always keyword data available for entities: "Nike", "shoes", "mint" and "38":
“Nike for Men”
“Nike performance wear”
"Adidas Shoes mint“
"Running pants mint“
"Shoes Winter black“
"Sandals Size 38“
Keyword: Nike Shoes mint 38
Conversion Probability:10% 8% 3% 9%
Each entity brings along its individual conversion probability.
- Regardless of the structure, we can reliably predict the conversion rate for each keyword.
- In the mid and long tail keywords, we determine individual bids even before clicks and costs are generated.
Step 3: Pilot tests with ABOUT YOU
What does that mean in practice? How successful is the entity bidding approach? These questions were answered by ABOUT YOU’s SEA team. They designed a multi-level test series, in order to make the evaluation as comparable as possible. According to the extensive tests in the Dutch SEA accounts, the approach contains the right formula to leverage the long tail keywords in a completely new way!
Test 1: Adference vs. Adference
In the very first phase of test of the entity bidding approach, Adference competed against itself. The keyword-based optimization by the Adference algorithm was compared with the optimization of the entities. The results were stunning - 100% more sales (??). This was the first cause of the joy, but without relation to the outside world, it was not very meaningful.
Test 2: Adference vs. Google on granular campaigns
Google and Adference competed on the granular campaign structures created by adSoul. The results were clearly in favor of the entity bidding approach. The doubt: The granular structure could be disadvantageous for Google's Fullistic approach.
Test 3: Adference Entity Bidding vs. Google Best Practice
Another test followed. In advance, both parties had the opportunity to set up their optimal campaign structure:
The comparison campaign for optimization on entities is focused on 558,113 keywords spread across 703 campaigns instead of 1108 keywords in 27 campaigns of a Google Ads campaign optimized for smart bidding.
The results speak for themselves: With the entity bidding approach, correct predictions about the expected conversion rate were made much earlier, resulting in significantly lower CPCs and a lower cost/sales relation from the beginning with the help of Adference.
We are confident - there is definitely more for your Google Ads accounts. Bring your campaigns to the peak of their performance.
Interested? Write to us.