The Knowledge Panel Course: Six Knowledge Verticals that Trigger a Knowledge Panel
Script from the lesson The Knowledge Panel Course
Jason Barnard speaking: Hi, and welcome. You might think that all Knowledge Panels are directly linked to Google’s Main Knowledge Graph, which we can also call the Knowledge Vault. This isn’t the case. Knowledge Panels are triggered by multiple vertical Knowledge Graphs in Google’s ecosystem.
Jason Barnard speaking: Outside the Main Knowledge Graph, the most obvious example is Google Business Profiles. Now, they are clearly a slight cheat, since they are essentially business listings. However, they are part of Google’s biggest Knowledge Graph, Google Maps. Google Maps is a huge Knowledge Graph that can provide solutions to geospatial entity-to-entity queries it has never seen before, and it can do that in real time. That’s phenomenally important.
Jason Barnard speaking: Think about how Google Maps works. That is where Google is going with the Knowledge Vault. As a quick aside, think about how you use Google Maps. That will give you a good idea how the Knowledge Graph will be integrated into Search and other Google products in the coming years.
Jason Barnard speaking: Also, Google Maps is absolutely massive. In comparison, the Main Knowledge Graph with just 150 billion facts is tiny. But eventually, the Main Knowledge Graph will become the bigger than Google Maps, and I would guess Google will merge the two. When that happens, entity management will become the single most important aspect of SEO, digital brand management, online reputation management, and digital marketing generally. At Kalicube, we are very much looking forward to that day, of course.
Jason Barnard speaking: Now, back to Google Business Profiles. Although the business provides some of the information, much of the content is added by Google’s algorithms. For example, the social profiles, reviews from around the web, some descriptions, additional facts, the popular times, and much much much more. And it’s getting increasingly algorithm driven.
Jason Barnard speaking: Examples of Google’s multiple other vertical Knowledge Graphs that trigger Knowledge Panels on Google SERPs are Google Podcasts, Google Scholar, Google Books, Google Shopping, YouTube Music, and there is even a Web Index Knowledge Graph. The Knowledge Vault is the ultimate place to be. That is where Google is moving all of the entities, whichever vertical they belong to initially.
Jason Barnard speaking: Hugely important is that the contents of the Knowledge Vault cannot be changed directly by any human. And that is an important point. Google’s Knowledge Vault is 100% controlled by the algorithms. No human inside or outside Google can directly change information in the Knowledge Vault. So, once you are in there, every action you take, every change you want to make will be indirect.
Jason Barnard speaking: In most of the other verticals such as Google Books, Google Business, you have some direct control. In the Web Index Knowledge Graph, you have indirect but relatively short term influence. However, ultimately, you are going to end up in the Knowledge Vault, aka the Main Knowledge Graph.
Jason Barnard speaking: Three points here. Firstly, any vertical outside the Knowledge Vault allows you some control. Secondly, don’t rush to get in there. Prepare the groundwork. Use the control whilst you have it. Create a strategy that ensures that the basic understanding Google has is correct before it adds you to the Knowledge Vault.
Jason Barnard speaking: Thirdly, when Google adds your entity to the Knowledge Vault, which it will eventually do whether you want it to or not, your control will be indirect and difficult, and any changes you try to implement will be slow.
Jason Barnard speaking: Given that it is the basis of everything to come in all Google products, aka the Google-verse, accuracy in the foundational information about your entity in the Knowledge Vault today is absolutely essential. You can easily get a Knowledge Panel without the Knowledge Vault. So, the Knowledge Vault generally isn’t the place to start. It is, of course, where you will end up, but the best short and midterm Knowledge Panel strategy depends very heavily on the entity type and the presence, or not, of the entity in one or more of these verticals.
Jason Barnard speaking: The Knowledge Vault is hugely important and very interesting, but the Knowledge Panel is what your audience sees. So, the Knowledge Vault is a vanity metric at this time. The Knowledge Panel is the pragmatic representation of your entity to your audience on Google, be it a company, a person, a film, a music group, a podcast, or whatever. Wikipedia-worthy entities aside, the Knowledge Vault is not likely to be your best strategy. Google verticals are most likely to be your short and midterm targets.
Jason Barnard speaking: Intriguingly, Google’s advantage of having a lot of different human curated data sets such as Google Books and Google Podcasts is both an advantage and a disadvantage for Google. An advantage because they have human curated data. A disadvantage because they can’t easily connect these vertical Knowledge Graphs horizontally. Because they are all built differently and managed by different teams and there are different levels of maturity, connecting them horizontally in real time in Google Search and other Google products is a huge challenge.
Jason Barnard speaking: Currently, to join these vertical Knowledge Graphs together, the different algorithms and services, Search, NLP, Image, and so on, use a shared lookup table that provides the kgmid and the entity name. All the algorithms refer to this lookup table, but they have only shared the most basic information there, which explains some of the seemingly stupid results of mismatching photos, names, attributes, and so on that we see on SERPs.
Jason Barnard speaking: On the other hand, Microsoft has fewer owned knowledge bases, and they have built their Knowledge Graph and provide Knowledge Panels directly from their Web Index. As you can imagine, this makes everything much simpler for them, and that means that they do it very, very well, arguably better than Google.
Jason Barnard speaking: Getting a Knowledge Panel on Bing is relatively easy from the perspective that the work you will do to get a Knowledge Panel on Google will almost guarantee a very good Knowledge Panel and entity presentation on Bing. I do suggest that you keep an eye on Bing, since it will give you insights into where the data is taken from. It is a fair bet that Google is using much the same sources as Bing, and Bing are much more open about the sources they are citing.
Jason Barnard speaking: Now, Microsoft is building a Knowledge Graph primarily using their Web Index. Although Google has all these vertical Knowledge Graphs, it is, of course, doing the same thing. As we’ll see, for podcasts, books, authors, scholars and musicians, the easy way to get a Knowledge Panel is focusing on the relevant vertical.
Jason Barnard speaking: For people and companies who have no presence in those verticals, the Web Index Knowledge Graph is the best entry point. An increasing number of Knowledge Panels are triggered by this Web Vertical Index. For the most part, they are small sprouts with little information. But if you think about the tree and twigs analogy from the lesson about how Google is building its Knowledge Vault, the Web Index Knowledge Graph is the ultimate sprout of a twig that you can then attach to a branch in the Knowledge Vault.
Jason Barnard speaking: A quick reminder, I currently refer to what Google calls the Knowledge Graph as either the Main Knowledge Graph or the Knowledge Vault. Whatever we call it, the important point is that this is a source of fact and information that no human can directly control.
Jason Barnard speaking: So, back to an explanation of how Google reconciles these multiple verticals. It uses a single identifier, the kgmid, for each entity and uses a lookup table to try to join these different Knowledge Graphs as best it can. So, each vertical attributes a kgmid to entities within its sphere. Within Google Books, Google Podcasts, and other isolated services, this isn’t an issue. As users within a single vertical such as Google Books, Google can show us the who, what, and the where very reliably just using that Knowledge Graph.
Jason Barnard speaking: But these vertical Knowledge Graphs cannot directly compare their named entity to the same named entity in other verticals. And all Google services, especially Google Search and Google Discover, need a unified approach and horizontal communication. Otherwise, as users, we see some strange results such as the wrong photo for a person or the wrong book attributed to an author or even multiple Knowledge Panels for the same person. This is huge. And ultimately, the only solution is including all the entities in the Knowledge Vault, but that is literally a decade or more away.
Jason Barnard speaking: So, short and midterm, what is your best bet for a Knowledge Panel that you can manage efficiently?
Jason Barnard speaking: Always focus on the kgmid. Always be aware that something you think is a single entity might be two or more. For example, your kgmid and Knowledge Panel for your Google Business Profile is your outlet or head office, which is an entity. Your company is another entity that will need its own kgmid. As the founder, you are another entity, and you need your own kgmid. Each product you manufacture or your service that you offer will need its own kgmid too, even if you are a uni-service or a uni-product company and your product or service has the same name as the company.
Jason Barnard speaking: Think about this. Kalicube is a company. It’s an entity with a kgmid. Let’s call that number 1. Kalicube has an office in the South of France on Google Maps, entity and kgmid number 2. We have a software called Kalicube, entity and kgmid number 3. We also have an event series called Kalicube Tuesdays, entity and kgmid number 4.
Jason Barnard speaking: And that goes on and on within my company.
Jason Barnard speaking: Boowa and Kwala is a TV series, entity and kgmid number 1. Boowa & Kwala using the ampersand is potentially another TV series, but we need to ensure that Google doesn’t see it as entity and kgmid number 2 since the word “and” and the ampersand symbol are the same thing. Surprisingly, Google often gets this kind of thing wrong because it is very literal. Boowa and Kwala is also a song. That is entity and kgmid number 2. Lastly, Boowa and Kwala is a fictional universe. And that is entity and kgmid number 3.
Jason Barnard speaking: Be very aware of that. If you aren’t clear in your mind and consistent in your presentation, then Google will never understand. And with those multiple entities with the same name, being clear with Google is the only way to sort it out in Google’s mind, so that Google gets it right and understands, in Boowa and Kwala’s case, three different entities.
Jason Barnard speaking: Back to the verticals. If you have a podcast that is on Google Podcasts, it will automatically generate a kgmid and most likely trigger a Knowledge Panel. If you have published a book and it is in Google Books, you as an author will have a kgmid attributed by Google Books. Each of your books will also have a kgmid. You and your books will therefore have a Knowledge Panel in Google’s brain. Whether that triggers a Knowledge Panel on a search or not depends on many factors that I will cover in another lesson.
Jason Barnard speaking: If you have released a song or a music album and it is on YouTube Music, Spotify, or MusicBrainz, then each of your songs, your music album, and yourself will also have a kgmid. You, each of your songs, and your albums will therefore have a Knowledge Panel in Google’s brain. As with books, whether that triggers on a search or not depends on many factors.
Jason Barnard speaking: Google Merchant for products is still in the early stages, but its structure appears to function much in the same way as songs, music albums, and artists. That is we have a manufacturer, a series of products, a specific iteration within that series of products, multiple sellers of that product with different offers at different times. You get the idea. The Google Merchant feed is vital here. Get that right, and you are halfway there.
Jason Barnard speaking: However, Google is now using the Schema you include in your product pages to crosscheck. It seems Schema will take the upper hand, so focus on that and make sure that Google Merchant feed and Schema are always in sync. Our friends at WordLift have an incredibly powerful solution for this for e-commerce sites, with an added bonus that they can automatically generate great product descriptions at scale.
Jason Barnard speaking: Google Scholar will trigger a Knowledge Panel for academics who publish, and it offers great corroboration for people who are cited by academics in their papers. Now, I haven’t mentioned Google Images. That too has a Knowledge Graph, and it attributes kgmids. Google Images is intriguing, because it triggers kgmid but rarely triggers a Knowledge Panel on its own. It is a great springboard for entities that don’t fit into podcasts, films, music, shopping, scholars, and so on. So, consistency in your images is a huge step forwards.
Jason Barnard speaking: Lastly, and most importantly, is the Web Index. This is huge. It is better than Google Images, because this allows entities that don’t fit into the obvious Google-owned verticals to get a foothold. If you can get your entity into this vertical, then all you need to do is follow the three step process we advocate at Kalicube, and please do watch the previous lessons if you haven’t already done so, and your entity will trigger a Knowledge Panel on a brand search and also gradually move towards the Knowledge Vault and end up in the Knowledge Vault. All you need to do is create consistency and build confidence.
Jason Barnard speaking: Depending on the vertical, you have more or less direct control. Google Podcasts offers a lot of control since the information comes from your own feed. Google Business Profiles offer some control. Let’s say two-thirds. Google Books offers immense control when you initially publish the book, but that is a one-shot. Changing it afterwards is a huge challenge.
Jason Barnard speaking: Outside Google’s human curated verticals, the Google Web Index vertical is where you want to start. I would even argue that this is ideal. Since when you make a real effort, you can control this information simply by being consistent and focusing on the right sources around the web and making sure Google has understood your
Entity Home.
Jason Barnard speaking: And once in the Web Index Knowledge Graph, moving into the Knowledge Vault is simply a question of ensuring that the information about your entity is super consistent over time. And building confidence is purely about that consistency and what I would refer to as normal online exposure by third party sources.
Jason Barnard speaking: Short term, you can leverage the different verticals to get a Knowledge Panel. But longer term, Google will move your entity into the Knowledge Vault where you have no direct control. At that point, if you want any level of control, you need to ensure that Google recognises your chosen Entity Home. That will allow you to leverage, influence, and potentially have control over Google’s understanding of your entity in the Knowledge Vault and how it presents you to your audience in its Knowledge Panels.
Jason Barnard speaking: We do this all day long at Kalicube. It is simple and yet, at the same time, complex. Since every entity type has its own vertical, each entity is unique. And how Google then pulls together the multiple sources of information it uses to build a Knowledge Panel is hugely intricate process, that I will explain in an upcoming lesson.
Jason Barnard speaking: Thank you. I hope this gives you a better appreciation of where Knowledge Panels come from and what your ultimate aim is. And I’ll see you soon.
