I was delighted to be nominated and then invited to attend last night’s BIMA Hot 100 and Digital Hall of Fame ceremony and party at the Ham Yard Hotel in Soho. The BIMA Hot100 Digital People of 2015 provides insight and recognition for both the industry and leaders within it. I was therefore honoured to ...
I had the pleasure of speaking at one of the SEO-industry top conferences on Friday, the fantastic Brighton SEO. I was asked to accept the challenge of a "pecha kucha" style 20:20 presentation, which means you have twenty slides, which auto-forward every twenty seconds, if you are ready or not! I chose to talk about what I believe to be the future of search on mobile, which Google describe as a "serendipity engine", using a users total context to present search results before the user has even thought to look.
Here's my slides and explanatory notes:
1. Mobile Serendipity: How Google Plan to Send Users Search Results Before We Have Even Thought to Look!
2. Serendipitous search, or the concept of a "serendipity engine" is something that Google spokespeople Marissa Mayer and Eric Schmidt have spoken about frequently in interviews and at conferences in the past few years.
A product in development Serendipitous search is under Marissa Mayer's remit and has been referenced often since her move to VP Geo/Local. Before looking at these references it's interesting to remind ourselves that Mayer's background is AI - that being her major at Stanford.
3. Of course, it's difficult for an outsider to "predict" the future when imagining what Google may have in development and the required product and data points to facilitate this; however, we can combine information freely shared with known developments in existing product, combined with general evolutions in tech to hone our intuition. So what do we know thus far?
4. As far back as 2009 Mayer and Schmidt were speaking about development plans for "serendipity". A mobile "opt-in" service. Combining a users total context to get a picture of what might be useful and relevant to them in the moment. I first became aware of the concept in a Telegraph interview in 2009, which is from where the quote in slide four is taken.
Required data/intelligence points for serendipitous search:
5. History - To model the future we first need to know about the past. I wouldn't want to hazard a guess as to how much history would be required to found behavioural inferences, however, according to Hitwise there were 2.2 BN searches in February 2012, but for this to be behavioural, it must be personal.
7. Proclivity - Proclivity is inclination, likelihood defined as a tendency to choose or do something regularly. It's therefore behavioural, related to preferences that might be shaped by past experience, cultural-social influences. The slide reference here is to caffeine, and the most significant development we've seen publicly for me as it might facilitate serendipity is with Instant.
8. How Instant Facilitates Proclivity Modelling
The Feedback-on-the-fly facilitated by instant allows for far more sophisticated proclivity modelling. Prior to instant feedback is more limited to interaction post query-completion, whereas now when presented with choices and the ability to edit or select on the fly, there's a much more detailed data feedback delivered in almost realtime.
9. Community - As mentioned proclivity is behavioural and our behaviour is shaped by past experiences and influenced by culture, hence community is also important to behavioural modelling. I'm unsure as to the legal ins and outs of data-modelling in social groups, though technically if our identity and that of our social contacts exhibits behavioural tendencies significantly anomalous to other norms wouldn't that be useful data?
10. Even pre-Google+ connectivity data is/was already available to Google for many users thanks to the social graph.
11. Our identities and connections are already well known to Google. Social circle results already favour documents shared by those in our 'circle' why wouldn't a broader algorithm use this data?
12. Users with a detailed Google+ profile are already providing this connectivity data to Google knowingly or otherwise. Xfn rel attributes and other structured data offer interpersonal entity detail e.g. hCard, hReview 'describe' properties. In this slide example I'm telling Google that the entity "Nichola Stott" on Twitter is the same "Nichola Stott" on G+ etc etc.
13. So we have History, Proclivity, Community and of course Location would be required to offer relevant search results. Products like latitude already allow us to make our location known and trackable and track that of our friends... As a mobile opt-in product I'd expect location=on to be a default requirement for Serendipitous search to be possible.
15. I can not predict exactly what Schmidt means by this - though we might interpret this as the goal of providing machine-discovered layers of data to augment our physical behaviour. It was at this event that Schmidt (talking about the serendipity engine) singled out personal context - with permission. However, given the recent Google Glasses product I'm glad I chose to mock this up using a kind of augmented reality search interface.
16. So, how might serendipitous search look? I'm away from home on business. Google know this because I use latitude and I'm opted into serendipitous search push notifications. I love coffee. Google know this because I search for coffee shops often when I'm mobile. Wouldn't it be logical to imagine a serendipitous search prompt to find a coffee shop when I'm out at 8.00 am in a strange place?
17. Wouldn't it also be possible for Google to notify me as to the whereabouts of contacts that I may interact with regularly that might share similar history, proclivity, community and location? And if they opt-into push notifications or 'friend alerts' notify me of serendipitous happenings? But how real is this product?
18. Whilst of course my mock-up was purely for entertainment and illustrative purposes only; at the last time we heard anything publicly about the product, which was in an interview between Jason Kincaid of TechCrunch and Marissa Mayer in May 2011, Mayer revealed she expected the product to be available 'inside of a two year horizon'. But what can we take from this?
19. My star take-aways for mobile even regardless of the serendipitous search product remain the same. Ensure your site is marked up correctly using schema reviews and localbusiness itemtypes and that your social media strategy is integrated and dovetailed technically as well as strategically with your owned and operated properties, so that if this product does become a reality your ahead of the game. Even if not - or even if it's wildly different to my imaginings this is still the immediate-future direction of search in my opinion.
20. I want to give the final word to Eric Schmidt.