Google Knowledge Graph Algorithm Update 2019 ( Budapest)
Google’s Knowledge Graph had a significant algorithm update in July and August 2019. The Knowledge Graph algorithm update is called Budapest. The Budapest Update resulted in massive fluctuations in resultScores or salience scores for Entities and an increase in the number of Entities returning information from the Knowledge Graph API.
The Google Knowledge Graph API is a tool that allows users to search for Entities in Google’s Knowledge Graph. Google has since migrated The Knowledge Graph API to the Google Knowledge Graph Search API.
The Knowledge Graph was introduced by Google in 2012 to shift from matching keywords to return search results and start using Entities, their attributes, and their relationships to other Entities to produce search results. Understanding the relationships between Entities allows the algorithms to understand the context and meaning behind a user search.
The relationships allow Google to return more relevant search results for users and create search result features that allow users to access the information they need quickly. Google uses Entities to produce Knowledge Panels, Knowledge Cards, filter pills, and other helpful features for users in the search results. Entities are identifiable and definable things, like people, places, businesses, and concepts.
What is the Knowledge Graph Algorithm Update of 2019 (a.k.a. Budapest)?
The Budapest Knowledge Graph Algorithm Update in July and August 2019 transformed how Google connects search queries with Entities. The update removed the traditional stability of resultScores or “salience scores” and began ranking Entities based on the highest resultScore.
The dataset for this paper is 7,504 brands and 4,069 individuals that Jason Barnard was tracking in the Knowledge Graph API when the algorithm occurred.
What were the two significant changes in Google’s Knowledge Graph Algorithm update in 2019?
Two things changed In July/August 2019.
- Salience scores fluctuated wildly. The average salience score increased significantly. Entities saw massive increases, while others had a drop in the salience score.
- The number of Entities returning for a ping in the Knowledge Graph API increased.
Salience Scores changes are listed below.
- Salience scores for Person Entities increased five-fold.
- Salience scores for Brand Entities increased 14-fold.
- Salience scores for Brand Entities increased almost three times more than Person Entities.
- The average salience score for Brand Entities overtook Person Entities.
The algorithm update significantly increased Brand Entities’ presence in the Knowledge Graph.
Size / Number of Entities Present
- The number of Entities returning information from the Knowledge Graph API increased.
- The number of Brand Entities in the Knowledge Graph increased by 42%.
- Google has more entities in the Knowledge Graph, or at least is significantly more confident in matching its ‘query string to Entities’.
What is the Knowledge Graph API, and How Does it Work?
The Google Knowledge Graph API is a tool built by Google to give users access to the Entities in Google’s Knowledge Graph. Google has since migrated The Knowledge Graph API to the Google Knowledge Graph Search API. The API returns the Entities that Google associates with a query string made by a user and ranks results by a resultScore. Kalicube calls the resultScore the “salience score.” The resultScore or salience score is the confidence Google’s algorithms have about an Entity.
Jason Barnard, CEO and founder of Kalicube, has been tracking Google’s Knowledge Graph API since 2015 and has since built the groundbreaking proprietary software Kalicube Pro. Kalicube Pro allows users to leverage the data captured daily since 2015 (ongoing) to easily educate Google about Entities and entrench businesses, brands, or names in the Knowledge Graph.
Check your Entity (name, brand, or company) in Kalicube’s free Knowledge Graph Explorer.
What are Salience Scores, and How Do They Impact Search Results?
“Salience score” indicates Google’s confidence in matching the query to the correct Entity and, in case of ambiguity, determines the most probable candidate based on intent perception.
The Salience score measures the following things.
- Google’s confidence that the Entity surfaced by Google is the one the searcher seeks in the query.
- The probability that the Entity is the intended Entity where there is more than one Entity with the same name.
- The closeness of the relationship between Entities in case the Entity does not directly correspond to the query. The API handles ambiguity by considering multiple entities and ordering them by their salience scores.
The entity with the highest salience score is considered by Google to be the most relevant and is presented first in the list.
What Factors does the resultScore/Salience Score Measure?
- Google’s confidence that the Entity returned is the Entity the user refers to with the query (i.e., has it matched the string of characters to the Entity).
- The most probable Entity where there is ambiguity, according to Google’s perception of intent.
- The closeness of the relationships between the Entities where no Entity corresponds directly to the query.
The Google Knowledge Graph API result for Homer Simpson is below.
Primary Entity (Most Probable)
Secondary (Less Probable) Entity
The search produces Entities with the name Homer Simpson with much lower Salience Scores or secondary outcomes, or alternative Entities, for the same query.
The Knowledge Graph contains five songs with the name Homer Simpson. The artists that are linked to these songs are DJ Bomberjack, Scott & Todd, BeebleBrox, The Death Killers, and Feva Da General. This is just for fun and to show Google’s difficulties when dealing with ambiguity.
Though the probability of these Entities being referenced is significantly lower than that of the fictional character, Google acknowledges that the string of characters refers to it.
Related Entity
Homer Simpson is related to The Sitcom “The Simpsons”.
The information pertains to the TV series, it is essential to note that the salience score for “The Simpsons” is 16,200 when considered in the context of the query “Homer Simpson.”
How Were “Salience Scores” Influenced Before the Budapest Update?
Salience Scores were relatively stable before the Budapest Knowledge Graph Update in 2019. Smaller Brand Entities typically scored in the 10s, while more recognized Brand Entities scored between 1,000 and 3,000. Person Entities had Salience Scores in the hundreds.
The average Salience Scores changed minimally before the Budapest Update, typically by 1% to 5% each month, with rare instances of up to 10%. The Budapest update made more noticeable changes in Salience Scores for individual brands and people.
Can You Improve Your Entity Salience Score?
Yes, you can improve your Entity Salience Score.
Actively working to improve an Entity’s Salience Score results in incremental increases. Corroborating statements about your business, brand, or name (Entity) with consistent information on third-party sites like Crunchbase, Wikidata, and industry-specific sites improves Salience Scores. Establishing connections and relationships between the brand and events, C-level executives, products, and partners featured in the Knowledge Graph improves Salience Scores.
Consider the progress graph for an anonymous client at Kalicube. The consistent upward trend over two years reflects dedicated efforts to enhance corroboration.
Kalicube’s solid understanding of the volume, placement, and timing of corroboration efforts has been instrumental in achieving these results.
Salience Scores are able to drop if fresh corroboration isn’t maintained. Google requires ongoing validation, especially for non-ambiguous names. Without consistent third-party support, the Salience Score tends to decrease.
How Does Ambiguity Influence the Salience Scores of Individuals Sharing the Same Name?
The relationship between ambiguous names and Salience Scores is a critical aspect often overlooked.
Jason Barnard is an ambiguous name.
“With ambiguous names such as mine or another entity with the same name (in this case, Jason Barnard, the footballer, the podcaster, the dentist, the ice hockey player, the gravity juggler, the baseball coach, the disk golfer… ), gaining or losing in salience/importance/notoriety/mentions would affect the salience scores for all name-doppelgangers.”
Jason Barnard, CEO and founder of Kalicube
Be cautious with ambiguous names, as fluctuations in salience are influenced by other Entities with the same name and any corroboration obtained.
Unambiguous names typically have higher salience scores, so choosing your brand name carefully is WISE.
What Factors Influence Salience Scores?
Factors affecting salience scores include are listed below.
- Popularity.
- Brand awareness.
- Freshness of citations.
Popularity
Salience scores are influenced by the number of references and user interactions with search results. For example, “Butch Cassidy,” the historical figure was prominent before the Budapest update, but after the update, the film took over, likely due to more online mentions and recent interest. The shift highlights the importance of fresh references and user behavior.
It’s an interesting change, but it’s a bit hard to understand. Just to mention, the person’s salience score dropped from 535 to 330.
Brand Awareness (a.k.a. The Homer Effect)
The importance of brand recognition is evident in the “Homer Effect,” where the character Homer Simpson outscored the actor Dan Castellaneta in salience, showcasing the impact of brand awareness.
Homer Simpson is widely recognized and saw a tenfold increase after the algorithm update, whereas Dan Castellaneta, less directly linked to the famous character, experienced a fourfold rise. This emphasized that brand familiarity was crucial to the Knowledge Graph’s Algorithm Update.
Freshness and Citation Recency
The Budapest Update was not helpful to individuals or brands with a low number of recent citations or no citation growth.
Deceased actors, or people considered “less-legendary” or “less-iconic” had lower scores compared to legendary figures.
The averages were significantly outperformed by those deemed legendary. Maybe this is because legends are regularly cited and remain eternally “fresh”—they can still be relevant 60 years after their passing.
Recent citations’ quantity and quality are significant factors in this update.
Tech Giants’ Dominance and Potential Bias
Tech Giants (excluding Amazon) experienced significantly higher-than-average increases in their Salience Scores. The references to these companies are fresh and numerous, coming from sources typically considered trustworthy by technology companies. The scale of the increases is staggering, with Google outperforming the others by a wide margin.
While it might appear that Google favors itself, the figures reveal a 600-fold increase, with Facebook and Apple following closely at 500-fold and 460-fold increases, respectively.
Microsoft is the exception, showing comparatively modest progress (though a 56-fold increase is not insignificant).
Google is at the center of its Knowledge Graph, creating a bias. Navigating its Knowledge Graph becomes easier if your brand is linked to Google.
Looking forward, this bias is able to increase, so tying your brand to Google could be strategic. But be cautious; this bias could also bring unforeseen challenges.
Subsequent Update: The “Paris Update”
The “Paris update” that followed Budapest revealed a decline in the “depth” of the Knowledge Graph, signifying a decrease in Google’s candidate set for a query. The drop coincided with a significant decrease in the presence of Knowledge Panels, observed in both tracked SERPs for brands and individuals and in a broader range of queries in the UK.
The correlation between the Knowledge Graph’s depth reduction and the decline in Knowledge Panels suggests a potential relationship. The connected drop in Knowledge Graph depth and Knowledge Panels calls for a closer examination of their correlation and implications between these phenomena.
P.S. Jason Barnard named the updates after the cities he was visiting when the update occurred.
Are you ready to elevate your brand’s visibility and ensure Google has mapped your digital footprint? Entities listed in Google’s Knowledge Graph are able to future-proof their online presence, ensuring Google knows who they are, is able to create a rich, stable Knowledge Panel, and appears in the generative AI search results.
The Kalicube Process is a three-stage-complete digital marketing strategy that allows brands to optimize their visibility, message, and acquisition funnel across the entire digital ecosystem. The Kalicube Process is refreshingly simple, straightforward, and grounded in brand, marketing, and search engine optimization (SEO). The SEO is future-proof: The Kalicube Process optimizes for Google’s Search Generative Experience and Bing Chat right out of the box.
The Kalicube Process
- Gives you The ideal digital marketing strategy so your business stands where your audience hangs out online
- Educates Google so it knows, likes, and trusts you – and rewards your brand with a Knowledge Panel and a glowing Brand SERP
- Ensures Google recommends your brand as the best solution in traditional search and Generative AI results.
Book a call with Jason Barnard.
