What if you swiped your phone, brought up Google Voice Search and before you could say anything it started to preload pizza restaurants in your area guessing that you were kind hungry. If you had a close friend who worked with you he’d know that on, let’s say Thursdays if you were in a particular area, on foot, around lunchtime you’d treat yourself to a pizza.
These are just the kind of human relationship-based value judgements we make all the time about people who are in our close circle (which I shall now call a little more formally, our close relationship graph) based upon our picking up of all sorts of small cues that accumulate, over time. We then extrapolate those cues to those we know nothing about, guessing that a total stranger, in a very similar situation, when he is consulting his watch and looking up maps on his phone may also be looking for somewhere to spend lunchtime at.
What’s easy for humans is hard for machines. This has always been the case. But the milestones in the cognitive landscape keep on moving and the point at which machines can learn and act like humans is getting broader with every passing year.
The news that Google Maps has added a predictive function to its toolbox that uses all sorts of accumulated clues to guess your intention each time you bring up Google Maps and preload choices for you to pick from is only an expansion of what Google did with the Autocomplete dropdown list you see as you type in Google’s search field. It is also in keeping with the direction the company has employed in YouTube Suggested Videos which preload on your YouTube account before you even start viewing.
All of the above are examples of Predictive Search (Google was awarded a patent in 2014 which comes under the domain of predictive analytics and Big Data which is to say that, really, it’s all about semantic search, again.
All Roads Lead to Search
Pretty much everything we do or hope to do on the web and, increasingly, offline involves some form of information retrieval. This is not an overstatement. Semantic search creates structure upon the chaos of information that surrounds us and enables us to make sense of the world, through it. Semantic search principles are at work in almost all forms of digital interaction that involve some kind of relational exchange.
Our social media activities, posts, reshares, comments, content creation activities, search and web surfing habits, emails and video calls using Google technology (Hangouts on Air http://www.amazon.com/Google-Hangouts-Business-Communicate-Real-Time-ebook/dp/B00JJPCMLM) form a complex digital footprint that’s defined by our strong graph of close ties and the weaker graph of extended connections surrounding it. Every point (node) in that weave has a value and so does every connection (edge), and these values keep on changing based not only upon our activity but that of others.
It takes a machine intelligence to make sense of all that complexity in its grandest form and this is exactly where predictive search comes in.
Why Do We Need Predictive Search?
Predictive search is a refinement that provides convenience and efficiency. As we use small, mobile devices to do things on the go the combination of small screens, small buttons and the need for haste create an environment that we are not designed to operate in, comfortably.
If, however, the devices we use became smart enough to save us the trouble of having to tell them what we want half the time, that would create a much better UX which would increase the number of times we use them and the tasks we use them, which in turn would result in them getting better and smarter as the data about us that they mine would accumulate.
Predictive search then is the next step in removing perceptual barriers between us and the things we use to navigate the digital and physical realms. And predictive search means machine learning. If you’re new to machine learning my piece on it at Plus Your Business will demystify the concept.
You don’t have to exhaustively understand how semantic search works in order to generate accurate predictive models, however we do live in the 21st century and knowledge is key to almost any kind of meaningful activity we undertake, so I will provide a crash course in the next couple of paragraphs.
The moment search tries to accurately predict intent it runs into the issue of rare events that need to be predicted from a sequence of events with categorical features. Search cannot hope to earn our trust if its reliability is poor or if it can only predict 50% of what we want to do. If anything, uncertainties introduced in this halfway house model would undermine both brand value and our trust in Google search and draw our attention to the search “fails” rather than its successes. In order for it to work then it has to work sufficiently well all the time.
Predictive search gets around the problem of the rare event by using genetic algorithms. A genetic algorithm is a search heuristic that mimics the process of natural selection, and uses methods such as mutation and crossover to generate new constructs in the hope of finding good solutions to a given problem. The algorithms employed in this scenario are known as evolutionary algorithms, as they use a rules-based approach to classification tasks which then become the leaping point for independent learning.
Genetic algorithms are exceedingly good at mining predictive patterns in a large series of events (called multi-dimensional Time Sequences) and generating patterns that accurately predict the next sequence of events. In plain speak Google search (in all its formats) uses the data crumbs we give it plus the data crumbs of millions of other users to create robust, predictive models which are then used to make the end-user experience of using specific Google services even better.
In the Google universe data is the fuel everything runs smoothly on so everything from Google’s self-driving cars to Google’s Augmented Reality and Virtual Reality projects http://davidamerland.com/seo-tips/1060-making-reality-porous.html would benefit from having sufficient data available to accurately divine our intent.
The “So What?” Question
Unless you’re enamored by search and semantic technologies what you really want to know is how you will benefit from the increased use of predictive search in Google services.
There are two parts to this answer. One is from the end-user point of view. As consumers of information you and I are only interested in the utility of it. What it allows us to do, how easily and through what devices, in what circumstances. In this case the answers mostly are, “everything”, “instantly”, “always” which means that the easier it becomes to accomplish something through a device the more likely we are to want to use the service.
The greater impact of this approach, however, will be in terms of marketing and business owners. If you have any kind of business you really want it to be found. This ‘finding’ of a business will depend upon it being visible to search, understood in terms of its context and capable of being categorized on the fly in terms of its potential value to those looking for something like it. All of this requires effort.
It is no longer enough to think that “search will find me” and all you have to do is have a website and produce some great content. As a business person you need to be sufficiently active in the online world to have an impact in the minds of those who see what you do and how you do it and relevance in their lives.
Impact’ requires you to have the attention of your audience and relevance means that they care enough about what you do to comment, share, engage with your content and make it part of their lives regardless of whether they are going to make a purchase form you or not.
It is this kind of approach that translates into “top of mind” attention which then leads to the desired outcome of a purchase from you when the time comes for your audience to make that purchase. It also means that their activity helps you surface more in search precisely because your data is mined through the mining of the data crumbs of your audience.
The Magic Marketing Formula Moment
Because this has been a long read and I’ve gone deeper into the predictive search concepts than usual the least I can do now is give you that magic marketing formula you need. For greater visibility in a data-crumb driven, predictive search world you need:
- Content that matters to the lives of others
- Clarity of purpose in the content you create
- Clarity of purpose in what your business does
- Clear communication with your online audience
There you go. Four Cs just like a diamond!
If you want a little help in how to achieve them SEO Help: 20 Semantic Search Steps That Will Help Your Business Grow should be on your reading device or bookshelf. For a deeper dive into the entire semantic search concept complete with actionable points Google Semantic Search is your thing.
Predictive query completion and predictive search results (Google Patent)
Genetics-Based Machine Learning for Rule Induction: State of the Art, Taxonomy, and Comparative Study
Effective search for genetic-based machine learning systems via estimation of distribution algorithms and embedded feature reduction techniques
Genetic Algorithm Search for Predictive Patterns in Multidimensional Time Series
Mining Predictive Patterns in Sequences of Events