Moving goalposts is a familiar sporting analogy often applied to the world of SEO. You take aim, prepare to kick, only to find that Google has moved the target. With the continuing development of semantic search Google is not only moving the posts to a completely different pitch, the rules of the game you’re playing could be fundamentally changing too.
In the old world of SEO, keyword data was critical. Trying to identify the exact terms that the people you wanted to be found by were most likely to type, and then getting your site to rank as highly as possible for those terms was the name of the game. And for savvier searchers, trying to pre-empt the keywords that the exact information you wanted might be optimised for became the key to tracking down what you wanted quickly.
The real world is not like this
In the real world we find things out by asking questions. We might phrase the questions differently, use different words from somebody else and we might have different reasons for wanting that information. The development of semantic search is moving us to a world where the exact search terms we use matters less, and the context and intent behind our question matters more.
Semantic search delivers results based on a range of factors including context, location, what it is you’re really trying to find, and why you’re trying to find it. It can take account of synonyms, related concepts and whether the query is specialized or generalized. Above all, it allows users to search for information or content using natural language.
With the implementation of Hummingbird, predictive search and the Knowledge Graph, semantic search is clearly the direction that Google is headed and Bing too is developing similar capabilities. Both see the ability for users to type a question in their own language, and get back relevant, personalised and good quality answers to that question as significant benefits.
Google has summarised the shift in emphasis as looking at ‘things not strings’. Looking at the essence and intent of the search rather than matching an exact string of words.
The graph approach
The world is full of data and information which, for the most part, is unstructured. If you can understand how all this data is linked at a conceptual level you are in a position to deliver richer and more relevant results for any query. If you can deduce what somebody is looking to achieve when they make a search, and then interpret data sources accordingly, you are in a position to deliver answers rather than just links that might lead to the answer.
Graph databases seek to build those relationships between different pieces of data. If you want to read a bit more about graph databases and how they differ from the relational model why not read “An introduction to graph databases”.
Google’s Knowledge Graph, Microsoft’s Satori, Facebook’s Graph Search are all examples of using a graph approach to query data. They understand, to an extent, the links between different bits of information. And it is an understanding of those links that you need to support semantic search.
What is semantic search?
Semantic means ‘related to the meaning or use of words’. Semantic searches are, therefore, more concerned with the meaning and intent of the search in real terms. Implementing semantic search means understanding how different pieces of information or data points are linked and grouped. In this way the search can be matched intelligently to related and relevant information or entities that might not match the exact search terms used.
This is the focus behind the Hummingbird algorithm update of August 20, 2013: less emphasis on exact keyword matches and more on what the user was really searching for.
With a semantic search, instead of typing ‘bee flight’ and sorting through the results I would simply ask: ‘how does a bee fly?’, ‘how fast does a bee fly?’ or ‘what times of the year do bees fly in the UK?’ depending on what I really wanted to know.
I will get different results for each search and they might not be ranked according to how closely I have matched exact words in the search. Additionally, as I type, Google is trying to pre-empt what I’m looking for and offers me helpful auto-complete options for my query.
In many cases, when you are looking for general information related to well-known people, places, institutions or buildings, Google aims to deliver answers derived from the Knowledge Graph in the SERP. It displays a knowledge panel in the right hand margin with facts related to your question and a collection of links to related information. It also understands how information about people, places and events is linked – even though, currently, it is only making those links across a small number of trusted sources called an “Entities Index”.
In the context of semantic search, ‘trusted sources’ are known as entities. An entity is hard to define but illustrates just how differently the semantic web operates compared to what with been used to. Think of it as a map of the relationships between different data points or possibly a star around which those data points orbit. Google is actively building an Entities Index so that those relationships can be used in generating search results.
While this is great for finding things out, semantic search for business websites has different, but equally significant implications. Which brings us to…
The importance of trust and reputation
For ecommerce and business websites, people are not just interested in who can provide what they want; they also want to know which provider can best meet their specific needs. This is where location, context and reputation start to become relevant.
So, having said that search results are influenced less by exact keyword matches, what is actually likely to happen in future when we search for products and services on Google? Think of the process in three stages: in the first Google is trying to interpret and refine the query; in the second it is identifying content that is relevant to that query; and in the third it is selecting the most relevant (to you) and authoritative content to present.
The final stage will be increasingly dependent on trust and reputation. All three stages are likely to have major impacts on how we do SEO.
For one thing the search results I get might be different from the ones you see. My results for a given query might depend on the following:
- My online relationships or interactions
- Recommendations and reviews by people I know
- The online reputation of relevant businesses
- My location
- Google’s view of those businesses based on more established SEO factors such as links, social signals and how people interact with their content.
The future of SEO is about optimising your site, your content and the whole of your online presence for your customers rather than for Google, which is a fundamental shift in thinking.
It will be more important than ever to invest in content that is relevant to your customers and that they value. Increasingly your social media, content and search strategies will have to be aligned and integrated. Connecting with your customers through social media will create the links on the graph that connect your business to your customers and their contacts.
E-commerce businesses would do well to look at what Google is doing with Actions in the Inbox that will allow customers to rate your service on a pop-up card in their Gmail Inbox. The card will appear if you’ve sent a confirmation email with the appropriate mark-up – users won’t even have to click on it. Will Google+ exert more influence on how your business ranks in individual search results?
If results are going to be determined more by the context and intent of the search, and people increasingly ask questions to find what they want, it seems to make sense to build your content around what people want and the problems they are trying to solve rather than what you want to sell.
Semantic search in e-commerce
So far we’ve looked at semantic searches in the context of search engines and finding things on the web. Semantic searches are being used increasingly on-site by large e-commerce businesses. Large on-line shops or corporate websites will have large amounts of information that may not always be organised according to how people want to search for it. A semantic search engine helps visitors get to the product or information they want more easily by displaying results related to the intent of searches rather than just the exact terms that were typed.
Walmart reported increases in conversion rates of 10-15% following the launch in 2012 of a new search facility that attempted to interpret the context of product searches.
Techcruch published a good summary of this soon after the new Walmart search engine launched. The search engine understands links between products and synonyms such as patio furniture and garden furniture. Once the semantic link is established, the relationship doesn’t have to be managed for each individual piece of product data.
Graph database technology also has the capacity to have a richer understanding of links between products, making the business of upselling, cross-selling and bundling offers much more scientific. There are new levels of refinement that will be possible with email marketing and paid search – targeting specific offers at exactly the right people.
The next big question is how businesses need to respond to make the most of the opportunities presented by semantic searches. It seems inevitable that including semantic mark-up in your HTML will become more important, there is even some evidence that pages with Schema mark-up rank higher in SERP’s. Semantic mark-up based on the schema.org standards allows specified entities such as products, places or people to be identified in your HTML, and allows you to assign attributes to those entities in a commonly understood way.
Coming back to context and intent for a moment; if Google is trying to interpret what you were looking for and which results would be most relevant for you, how does semantic mark-up help?
Sematic mark-up allows you to define entities in your content that can then be indexed and ranked in their own right. So even if your content doesn’t use the exact sequence of words used in the search it’s still possible to make the connection between the intent of the search and your content (back to things not strings).
Semantic mark-up allows you to add data about your products. Already this means that your SERP listings can show price, ratings, and availability via ‘rich snippets’ in the search results. The ability to display this richness of information increases CTRs, sometimes dramatically.
Here are just a few examples of entities you can define with semantic mark-up (you can view all Schemas here)
Adding event mark-up to your content will allow specific events to show up asn rich snippets, giving your event(s) prominent publicity in SERPs:
Products and reviews:
The mark-up will make you less reliant on exact keyword matches and, if you use it correctly, will show user ratings as rich snippet like this one:
If you’re using YouTube as your channel the mark-up is added automatically. If you’re putting video on your own site then the mark-up will give you a snippet like the one below:
Which one are you going to click on: text only or the one with the video?
Fundamentally, by giving search engines richer information about your products or services you are increasing the likelihood that they will show up in a related semantic search that doesn’t necessarily use the keywords you’ve used to optimise your site.
If, for example, you run a hotel or tourist attraction close to an airport, including the semantic mark-up identifying that airport on your website would make sense. It would increase your chances of showing up in searches related to the airport if somebody was looking for flights, for example.
While the idea that having semantic mark-up on your website improves your search ranking in itself is up for debate, the consensus among many SEO’s is that anything you do to help search engines understand what your site is about will help you show up in more, relevant searches. The quality of your rich snippets may then have more influence on CTRs than whether you in position 1 or position 5 in the SERP’s..
And as search engines increase their ability to interpret natural language searches in a meaningful way, helping them make the link between the intent behind the search and your content seems to be a profitable way to go.
About Jon Dunn