SES London 2010 | Realtime Search Panel

Line-up Change: Aaron Kahlow is moderating. Leaving Fischer, Scott and Walk as panellists.

Aaron kahlow is defining reatime as live results, prescient information. Remember Twitter and FB are not the only, nor often the most important/useful source of realtime.

230MM pieces of UGC published every day. A 40BN dollar market. A lot is at stake in terms of who cracks this first. Thinking of the MSN Twitter deal, quickly countered with Google deal shows us that a lot of this is knee-jerk. It’s a struggle even for the big guys.

Example: Remember Michael Jackson example. G shut down results as the volume search was thought to be a denial of service attack.

Realtime search is cool. Normal search is static. Do not get caught up by the cool factor. Lots of this is not verifiable, nor accurate.

We’re going to be looking at challenges, relevancy, unique sources, credibility of sources and of course the age-old issue of spam.

To the panel…

Rob Walk – Nova Rising

Takes us through a case study for BBC Switch, which was “Sound Index”. A tool to collate RT social mention for music specifically.

The Sound Index was a top 1000 chart taken from YouTube, Bebo, LastFM and later… Twitter an FB; which used algorithms for trawl and crawl and semantic analysis for inetrpretation (which had inherent difficulties due to slang and “teenspeak.”)

Relevancy of the product is underpinned by the need to know/understand “what is a social, music-mention?” To do this you need a taxonomy. This product used the lastFM taxonomy however; recency was a recurring issue. As an example Joe McElderry was added to the taxonomy, however “the climb” was not bringing up mention for Joe, as it had not been added to the taxonomy. Such static algorithms are problmatic, which they are looking to improve by allowing user interaction behaviour to inform the development of the algorithm. Identifying feeds, stating preferences, “liking” etc informs and weights algorithm componants.

Other issues – noise vs freshness e.g. when Britney lost the plot. So much mention was non-music related.

Rob hands over to Bill Scott CEO of easelTV.

easelTV – Using Broadband connection with TV to enable brands to deliver content.

2010 – consumer push to upgrade TV for many reasons e.g. analog switch-off, HDTV coming to freeview, BBC iPlayer pushing to TV and more.

When analog stops so will VCRs.

Project Canvas (lead by Beeb), with 4, itv, BT, five and more to establish standards in BB TV, so that there are tangible technical targets.

easelTV working to these standards, will allow brands to use largely web based technologies to do this. So tools and techniques from mobile, web etc. can be transferrable.

SONY, Samsung, Panasonic, all major manufacturers are working on standardised BB compatible TV.

£200 will get you equivant functionality to Sky box, in an accessible consumer device to access this BB on-demand content.

Consumer drivers will be 2012 Olympics.

TV is of course v.different to web. TV is about shared experience not personal information retrieval. Traditional web products will not translate well. e.g. Google search engine. traditional discovery methods will not possibly be the route forward on TV medium.

Also – Serendipity. TV behaviour – you just watch it because it is there. TV is about delivery with minimal viewer/user effort. Must be an element of push, of relevance and of realtime. TV must intuit from prior interaction, (skipping, most watched, recorded etc.) but needs to be smarter than the Amazon style, recommendation engine. TV must combine serendipity with recommendation adn social trends.

Surfacing information via TV is about “Suggestive Discovery” and not search as we may know it. So how do we relate SEO to TV?

  • It is still about meta-data (describing content is very important)
  • It is more (though not exclusively) about B2C
  • More about video assets than pages
  • An integrated discovery with aggregated content proposition

Relationships with the consumer and consumer trust are key (because of the way behaviour informs suggestive discovery.

Bill wraps up with his vision of the future of BB TV.

  • Connected
  • Still Television
  • Retains serendipity
  • Personal and relevant
  • Dynamic

Finally; we have William Fischer of Watchdigital Ltd.

Watchdigital focus primarily on realtime job search, with products such as TwitterJobSearch.

Primarily they look to create a semantic search experience focused on the former classified ad space, which in terms of opportunity size is around a $48 BN market.


Due to the fragmentation of the space. We no longer advertise jobs in trade press or classified ad sections, but in the 6000 Uk job boards and on Linkedin and Twitter.

Seeking to collate this disparate content and information types has it’s own problems of course…

Interpretation (140 characters). Search alone is not enough. Also a tendancy to speak to our communities in social media. Our social peers may know what we mean if “looking for an associate” but that is meaningless without additional qualifying data.  Again they are using meta data such as profile and geo-location.

Again taxonomic databases, semantics, synonym databases are required to add meaning. A search for “web developer” jobs should bring up “php developer” jobs.


Aaron starts us off with the problem of relevance – how do we combat this?

Rob thinks we need to look deeper than demographic plus match and look more towards behavioural information in RT. William agrees, and this is something they too focus on. Bill – blending different data services together is very important.

Aaron asks the audience for input, but we’re temporarily overcome with shyness, so he saves us with a question about signal vs noise.

William – difficulties tend to occur the broader the user pool which is why they prefer to build specific user base products and tailor algorithms accordingly. I’m guessing such examples could extend to industry lingo (PPC means a whole different thing to paid search vs organic search communities). My example. I’ll ask @williamfischer to clarify…

There were a couple of questions from the audience…

1. Are job searchers active or passive?

William 2/3 of activity is from passive (got a job, casually browsing, vaguely happy, might leave if something better turns up. My words.)

Passive searchers are the Holy Grail as they tend to be more lucrative and less desperate (bit of paraphrasing there re: desperate).

2. Do your products have multi-lingual capabilities?

All panellists have multi-lingual technologies underpinning products, however the position seems to be to learn to walk in the UK, before running off into other markets.

In summary: realtime is much more than Twitter and Facebook. Integrating realtime information with broadband TV will be a huge market focus from now on. Issues exist around relevancy, understanding signal vs noise and developing sophistocated interpretative tools that incorporate behaviour, interaction, semantics and traditional taxonomic databases, to inform product behaviour.

Panellist Twitter accounts.




Leave a Reply

Your email address will not be published. Required fields are marked *