Living in a digital world

Just another WordPress.com weblog

Explaining the Persistence Paradox

Kevin Kelly recently drew attention to a paper by Fang Wu and Bernardo Huberman of HP Labs (PDF) analysing the popularity of almost 10 million YouTube videos. The paper takes the somewhat narrow view that all videos submitted to YouTube are attempts to achieve mass attention – to have the YouTube equivalent of a blockbuster. Within this framework, the researchers observed an interesting result which they termed the ‘Persistence Paradox’.

In essence, the Persistence Paradox is the apparent contradiction between two objective results obtained from their analysis:

The more frequently an individual uploads content, the less likely they are to have a ‘hit’ (where a hit sits in the top 1% of videos by viewership).
The more frequently an individual uploads content, the better they get (using video ratings to measure quality).

The only explanation offered is a kind of Diminishing Returns theory: “when a producer submits several videos over time, their novelty and hence their appeal to a wide audience tends to decrease.” One example of this might be Ray-Ban’s Never Hide Films, which has yet to achieve the same amount of attention earned by the original “Guy catches glasses with face” video. But as noted in the paper, this theory alone does not intuitively seem to adequately explain the apparent paradox.

Being fairly familiar with the YouTube ecosystem, several alternative explanations come to mind.

1) That difficult second album
Just like a musician, the first work someone produces in a particular channel is likely to have been gestating for a much longer period than those that follow. This goes hand-in-hand with the paper’s Diminishing Returns idea. An example would be the channel of Those Lil Rabbits, who despite producing many creative and moderately popular videos have yet to match the original towering success of their first David Blaine parody, shown above.

2) The harder the battle, the more glorious the victory

Someone that produces videos relatively infrequently (such as the Simon’s Cat animations) is likely to be putting in more effort and so produce better quality videos than someone that uploads more frequently. (If a smart match can be found this can be an efficient way for a brand to get involved – see the most recent Simon’s Cat video for the RSPCA). There is also a scarcity factor – something produced regularly is likely to have less apeal than something rarer.

3) Speak up if you have something to say

There are many motivations to join YouTube and submit a video, but people are more likely to take that step if they think they happen to have captured something really worth sharing with the world. It could be because they just happened to capture something particularly impressive (such as this ‘Crazy trampoline stunt‘), or they have one particular party piece they want to share (such as this dog trick). Another possibility in this vein is of course that of an expensive piece of branded video being launched on YouTube along with its own channel, such as the recent LED sheep video put together for Samsung.

With these ideas in mind, the Persistence Paradox does not seem so surprising. The next question we have to ask is: so what? What does this tell us if our intent is to produce a hit?

Apart from unpredictable, Black Swan Theory-type successes, the conclusions are pretty clear and reassuringly intuitive: work long and hard on your content, and if that doesn’t work, try doing something completely different. And don’t forget that gaining the loyalty of a smaller group through relevant and regular content can often be more useful than having a single huge one-off hit.

April 8, 2009 Posted by mantim | Uncategorized | | No Comments Yet

Location Based Services part 3 – New ways of choosing

Making choices in the real world

In the final post in this series, having introduced Location Based Services and described how HP’s Mscape platform is blazing a trail in this space, it’s time to consider the implications for digital marketing.

It’s extremely challenging to fully grasp the potential of having access to location data on a mobile device. One way to approach the problem is to consider that the most useful sites online have proved to be those that help us to search for things and/or choose from the things that we have found. I segment search/choice-making strategies (implemented either by a user or by an algorithm they choose to run) in to four main types: Objective, Curated, Similarity-based and Profile-based.

Making objective choices - taken at the London Transport Museum

Objective choice

This is the simplest strategy in which we have just one or two well-defined, objective criteria in mind, and any choice that satisfies these criteria will do. Perhaps we just want the cheapest (or most expensive) wine on the menu; the most popular camcorder on Amazon that costs under £250; the mobile contract that gives the most minutes for £25 a month.

When we consider how location based services (LBS) might help this strategy, the most obvious benefit is being able to search for places by distance – so we can find the nearest free WiFi zone, or cashpoint, or public convenience. When combined with other criteria, this can become even more useful – find the cheapest beer within 10 minutes, or the most popular tourist attraction within half an hour.

The other benefit of mobile services is having access to time-sensitive information. If you wanted to travel to, say, Baker Street, a service could take into account real-time public transport data, and find the best possible way to get there from where you stand right now.

There’s a further benefit -  if usage of these kinds of services reaches critical mass, we will suddenly have much more efficient ‘load balancing’ across all kinds of services, as people can find pubs, buses, facilities etc that are below peak capacity.

Curated choices from Time Out - perfect for an LBS

Curated choice

When we seek something curated, we use a trusted curator – be it an individual or a group – to help us narrow the field of what to choose. We might listen to a particular DJ to find new music, rely on a few critics and a trusted friend to highlight movies we should seek out, or a community like Digg to direct our attention online.

Although we can already choose places based on a trusted curator – reviews of restaurants being a common example – LBS will still massively improve this type of location-finding. The ease of access to this information while on the move makes it much more appealing, because we can discover interesting places that happen to be close to us that we wouldn’t usually make a specific trip to see, and we don’t have to print out a map in advance.

Possible curated services could include places where a particular celebrity has been spotted, shops that your Facebook friends spend a lot of time in, places that have been mentioned on Boing Boing, great examples of grafitti identified by a Flickr group, or instead finding places to avoid by overlaying geographic crime data.

Similarity matching with the X-Rite color IQ test

Similarity-based choice

In this strategy, we narrow the field by looking for something similar to things we have enjoyed before. Musically, this would be when we try music in a genre that we already know we enjoy. More recently, Pandora.com uses a team of musicians to classify thousands of songs, and is thus able to stream music to a listener that is similar to a given song or artist that they choose. Of course, because similarity matching works in the same way for everyone, this is one of the most useful ways shops can lay out their wares.

When it comes to LBS, passively recorded data of places you tend to visit can be used to find similar locations, and alert you when you happen to be close by to somewhere of potential interest. For example, if you visit modern art galleries whenever you are in a new city, a similarity-matching algorithm could notice this, and any time you are 10 minutes away from one you’ve not yet discovered your device could alert you.

In true permission-marketing style, you might opt-in to a service from Wagamama that will alert you if you are within 10 minutes of one of their restaurants and it’s between 6pm and 8pm. Similarity matching could then take this further by identifying nearby places that serve similar food. If you prefer niche rather than chain clothing shops, similarity matching can identify this and point them out to you. (Indeed, a key factor in the success of chains such as Starbucks or HMV is that people can choose to visit a new one knowing roughly what to expect. LBS has the potential to provide much better information about shops you’ve never heard of, at the very instant you are trying to decide whether to go in or not. As LBS takes off, chains may find themselves under threat).

More advanced similarity-matching algorithms might identify (from automatic time/geo-tagging of your Flickr uploads) that you often take photos of sunsets, and could let you know when there’s a beautiful view just around the corner. They could identify pubs/clubs that are frequented by people like you, or find public transport routes that match your preferred balance between scenery and efficiency. Wherever there is tagged location data, similarity-matching will naturally arise.

Profile picture matching - by Jorge Cham, PhD comics

Profile-based choice

This type of choice has only recently become an option at all. Not so long ago, the closest we could get would have been speaking to that one employee at the record shop that who knows what you’ve enjoyed in the past and can identify what new music you would love.

This sounds like a curated choice, but there is an important distinction. When we choose something curated, we leverage the fact that we know and understand someone’s way of thinking; in profile matching, we instead use the fact that someone (or an algorithm) understands our own way of thinking.

The modern version of this service is provided by Last.fm, which processes the listening data of thousands of users in order to predict for any one individual what other music they may enjoy. One of the reasons this works so well is that Last.fm can capture the music you listen to through your PC or laptop. This gives it access to a greater quantity and quality of data than if you had to manually tell it your preferences.

Similarly, LBS could track your everyday movements (both real and virtual) and use these to build an accurate picture of your preferences. If sufficiently well programmed, it can then make intelligent recommendations. You watch Sci-Fi films at the cinema and go to and Forbidden Planet – you might like this obscure retro collectible shop that’s down the back street you are about to walk past. You go to art shops and once went to Amecon – you’d probably like this manga art exhibition. You often move at a speed consistent with skateboarding – there’s a great skate location just around the corner. You favourited a Banksy book on Amazon – there’s a Banksy on the street across from here. You subscribe to techy RSS feeds – Inamo, a restaurant with a digitally projected table interface, has a table free for two right now and is 2 minutes away.

This is the top of the tip of the iceberg

Those were just a few examples of using LBS to help choose something, but the potential generalises so much further. There are many more possibilities for each of the four strategies outlined, and you can always mix and match these strategies. There will of course be applications beyond choosing, and beyond using just the location data of your own mobile device. Then there’s the possibilities offered by platforms like Mscape, enhancing the real world with parallel virtual worlds. New business models and new marketing opportunities will inevitably emerge.
So what can we do right now?

Despite relatively low penetration of suitable mobile devices, the conditions are already in place for an LBS killer app to emerge - and this would then drive further support. Although several mobile operating systems already have the potential to support LBS (including Symbian, the dominant OS among smartphones), the greatest opportunity right now sits with the iPhone and the G1. This is because they also each operate a single hub which makes downloading applications easy - the iPhone App Store, and Google’s Android Market. Each has its limitations – iPhone apps can’t yet run in the background (crucial for passive LBS), and the Android Market is some way behind in terms of maturity. But these hurdles can be overcome very quickly.

Keep an eye on interesting and well-implemented LBS applications that are coming out right now. Or better still, make one.

November 7, 2008 Posted by mantim | Uncategorized | | 1 Comment

Location Based Services part 2 – HP’s Mscape

In my last post I mentioned that I had recently been playing with Mscape, the geolocation platform developed by HP Labs in Bristol.

In the video “Roku’s Reward” (above), we are shown HP’s vision of what Mscape might be able to do in the future. Apart from a crushingly unimaginative take on the target audience (after the Wii has already established that innovation in games can engage more than just teenage boys), it suggests a few interesting ideas.

Unfortunately the most noticeable one – viewing ‘through’ a device to see a virtual world replace the real one in real-time – is the furthest from fruition, and also arguably the least practical. It certainly seems highly unlikely that their prediction in April 2007 - that such technology was just two years away -will come to pass.

Right now, the Mscape platform allows anyone to program ‘mediascapes’, which are flash- or html-based applications that can take advantage of the GPS data provided by a mobile device. Despite being a fairly hefty compromise on the vision set out in Roku’s Reward, this is still enough to create some very interesting new experiences.

User-created mediascapes can be uploaded to the Mscape site for anyone to download, try, or edit for their own purposes. The site has been running since April 2007, but hasn’t reached a huge audience (the most popular mediascape clocking just 672 downloads at the time of writing), partly because of the relatively narrow hardware requirements.

Another factor that constrains growth is the fact that geolocation applications are, by definition, based on a specific real-world location, or ‘anchored’ in Mscape’s terminology. It is still possible to write ‘portable’ mediascapes, which can be played anywhere, but these can never fully exploit the power of the platform – the ability to lay a virtual world over the real one, and leverage the interactions that emerge.

Playing with mediascapes

Consequently, I began my Mscape experience by travelling to Bristol, where there are many anchored mediascapes that have been written for locations around the city. As can be expected with user generated content in a new medium, the quality and usability of the mediascapes varied widely, but I quickly gained a feel for what was possible, and to my mind it seemed mediascapes tended to do one of two things.

Some mediascapes take an existing idea and play it out in a new way. The introductory game you are recommended to try first is just ‘whack a mole’ – you set up three locations in an open area, and then whack moles you are told are coming up in these zones simply by running over to them. Another example is a treasure hunt, in which clues direct you to certain places in the centre of Bristol, your GPS location confirms that you have found the place and solved the puzzle, and this triggers the mediascape to give you the next clue. In this particular case, the treasure hunt ended at a crepe van and even suggested you treat yourself to a crepe as your reward, immediately demonstrating some of the potential this field has for marketing.

Having run a couple of treasure hunts in the past myself, I was familiar with some fundamental gameplay problems that Mscape was able to easily fix. One problem is that if several teams are playing, everyone can follow whichever team solves a clue first. In Mscape, one can simply set up a number of different clues which each team visits in a different order, a final clue only being unlocked once all the others have been completed. A more significant problem is that a treasure hunt is no fun at all if a team gets stuck on one particular clue, as there is absolutely nothing more they can do. With Mscape, you can time how long a team has been stuck on a clue, and have hints automatically revealed at appropriate intervals.

With the help of my friend Richard Loxley’s consummate programming skills, I designed a mediascape treasure hunt set in Regents Park, and ran it for 36 people. It was notably more successful than my previous treasure hunts, particularly in that all the participants had a lot of fun - not just those that found the treasure!

Of course, the other class of things you can do with geolocation technology is things that were never possible before. Among these is my personal favourite of the mediascapes I tested in Bristol, “In 10 seconds“.

“In 10 seconds” is not just a mediascape. It has an introductory short film and a website to be explored first, which is a smart way of building up associations with the location while also establishing back-story. It centres around a ghost story, which requires a particular kind of atmosphere that one does not tend to find in the public park in which the mediascape is located. However, there is another dimension to be considered in the creation of such an experience – time. The mediascape recommends that you play at dusk, and as this is also the time at which the short film was shot this will reinforce the associations. (I myself ended up playing it at midnight, which was also extremely effective!)

One of the things a mediascape creator has a lot of control over is audio, which as Mark Kermode has noted is a key part of creating a chilling experience. “In 10 seconds” uses location-triggered sound to great effect – ghostly voices rush past you, you hear a gate slam shut just after you pass through it, the creak of a playground swing grows louder as you approach even though you can see it is quite still. The sound design of “In 10 seconds” was the finishing touch to a really fascinating experience that could not have been created in any other medium.

Working again with Richard, I produced something similarly new, although not as profound as In 10 Seconds. A full write-up will shortly appear on the Mscape blog, but it essentially involved using the mobile device as a virtual Geiger counter, with the familiar clicking sound of radiation being triggered by location in order to allow teams to track down virtual radiation hotspots – much to the confusion of members of the public! For the finalé, we shot video in the same location as the players found themselves in, which could then be displayed on the devices themselves to give the impression of having a window on a parallel world – a simple way of approaching the kind of experience shown in Roku’s Reward. The illusion was heightened by arranging for teams in different locations to see footage from appropriately different angles.

Given all the above, it should hopefully be clear that the possibilities offered by this technology are huge – too huge, in fact, to easily comprehend. I will go in to the framework I use to try to grasp how this could be used in marketing in the third and final of these posts.

For more information on Mscape, visit the website, Mscapers.com, or watch this 7-minute video from HP.

October 28, 2008 Posted by mantim | Uncategorized | , , | No Comments Yet

Location Based Services part 1 – Gaming as a Technological Bellwether

I’ve spent the last few months getting involved with some location-based gaming, and I’ll be reporting on my findings and thoughts in three parts. In this first part, I (post-)rationalise my personal interest in location-based gaming.



Nokia’s “Welcome to the 4th screen” video (above) takes an interesting angle on the technological developments of the past century and the way they have come to affect our lives. Taking a dramatic stand against fears that technology is destroying the natural order of things, it optimistically describes how the (implicitly Nokia) mobile brings all that is good about the digital world together with all that is good about the real world, and hints at some kind of transcendental state of being that can be attained at that sublime intersection of realities.

Personally, I think the most interesting and important developments in linking those worlds have to do with geolocation data and location based services (LBS), something hinted at in the video with the line ”We discovered new people and places and experiences.” But is the general public really going to embrace this vision of the future?

Jan Chipchase forecasts a difficult transition period into a world of shared location information, and makes an informed prediction that ”we are all only one or two compelling, highly social applications away from shared location information being indispensable.” Anyone skeptical about what promises to be another digital revolution should bear in mind that the technology required to track the location of your mobile phone has been in place for years, so there’s no need to wait for GPS to become standard in handsets.

Such a technology can only really take off if all the handset manufacturers get behind the idea, and recent acquisition activity shows that this is already the case (not least Nokia’s $8.1 billion acquisition of Navteq).

The real sticking point was arguably the lack of well supported geolocation platforms that everyone could write applications for, a problem that has now been addressed by Google, Yahoo, and Firefox in the form of support for the W3C draft specification of a geolocation API.

So what can we do with this technology?

One way to get an idea of what a technology offers is to look at the games that are made for it. From a certain point of view, games brought computers into the home, they brought portable technology to the masses, and MMORPGs anticipated social networks. (One could reason that early adopters are mainly geeks, geeks often find socialisation difficult, and since games offer a conveniently codified form of social interaction, geeks tend to play and make them… but that’s another story).

Geocaching is the geolocation equivalent of Pong, in that it is about the simplest idea for a game that exploits this technology that you could come up with. But it seems clear that the possibilities go far beyond this.

A step up from geocaching can be seen in the form of Mscape, a geolocation platform for certain Windows mobile devices developed by HP Labs in Bristol. This is the platform I’ve been experimenting with in the last few months, and I’ll be writing about my findings in the next post in this series.

October 23, 2008 Posted by mantim | Uncategorized | , , , , , | 2 Comments

Orange, PMOG, and the evolution of tools

This week sees the launch of the “World’s first internet balloon race” by Orange (arguably an evolution from this idea). Although it may seem like a short-lived gimmick, it could perhaps be more accurately described as the latest attempt at creating a kind of metaweb – a layer of content that sits on top of the internet as we know it.

This is all part of a familiar story: the difficult process in which we collectively work out what to do with the technology we have created. One of the most visible examples of this process is the mobile phone – a well-conceived device to begin with (in terms of global market penetration, going from negligible to 50% in 15 years), it turns out that it can also provide banking infrastructure to third-world countries and has of course become the major focus of device convergence.

A more visceral example of this process can be seen in the field of movie special effects, where it plays out spectacularly in the public eye. For example, Steven Spielberg made use of CGI in Jurassic Park, and as a tool it evolved to a point of fluency when George Lucas came to direct the Star Wars prequels (although even now it still struggles to win over much of the movie-going public).

But the Academy Award for visual effects in 1999 went instead to The Matrix, because ‘bullet time’ demonstrated the first step towards a new tool: the virtual camera. As with CGI, there was a long period in which the new tool was used for its novelty value alone, rather than focussing on how it might be developed further. Almost a decade later, the team that first created the tool (John Gaeta and the Wachowski brothers) have demonstrated its greater potential with the release of Speed Racer. One shot in that effects-heavy film is notably weaker than the rest, and that is the direct homage to the original bullet time-style shot from the end of the original anime series’ opening credits – it feels awkward and contrived, and shows how far the tool has come by contrast with the rest of the film’s fluidity.

Of course, it’s not easy to tell which tools will actually prove useful given further developement. For example, 3D cinema has repeatedly failed to take off (although James Cameroon is convinced that this time it’s different). This difficulty was recently highlighted by Kevin Kelly, who recalls with humility his initial scepticism at the potential of Photoshop, Quicken, and eBay.

While it is hard to see quite how Orange’s balloon race could lead to much greater things, the recently hyped ‘Passively Multiplayer Online Game’, PMOG, while essentially a very similar idea, shows greater promise. As it stands it seems something of an unbalanced, gameable, niche product open to abuse by the likes of myminicity, but it does seem like a significant step towards something far grander than we can quite grasp at this point.

It’s a significant advance on the old web annotation concept, in which users can add notes to any web page they visit that are then viewable to subsequent visitors using the same annotation plug-in. PMOG takes the significant steps of explicitly facilitating links between otherwise separate web pages, and of introducing a half-way practical universal web currency.

Perhaps most significantly it suggests another route towards the anticipated metaverse – a single virtual destination through which eventually all our information and interactions will be filtered. Instead of waiting for a platform such as Facebook or Second Life to reach Google-like domination and thereby gain the power to force all content to fit their own structure and social norms, PMOG is instead a layer that rests lightly atop the web, allowing everything to continue as normal, and simply making things better for those that choose to view things through that layer.

As the Man On A Train says in Waking Life: “Things are just starting.”

June 23, 2008 Posted by mantim | interactive | , , , , | No Comments Yet

Deception is not an option

Last week the IPA announced that legal restrictions on deceptive buzz marketing have now come into force, in the form of the new Consumer Protection from Unfair Trading Regulations, which has quite a few things marketers need to pay attention to.

The IPA drew particular attention to item 22 on the list of 31 specific types of marketing activity considered ‘always unfair’, which reads: “Falsely claiming or creating the impression that the trader is not acting for purposes relating to his trade, business, craft or profession, or falsely representing oneself as a consumer.”

Online, it seems this will primarily act as a welcome disincentive to those companies taking advantage of user review sites to give themselves inordinately favourable write-ups under the guise of consumers.

Perhaps as a side effect, this also rules out the deceptive tactic of releasing content intended for viral distribution under the guise of a passionate consumer.

Broadly speaking, the list of unfair practices will accord naturally with practices consumers already perceive to be unfair (or indeed were already covered by the Trade Descriptions Act). But when all is said and done, consumers will never take kindly to deception, whether it falls under the remit of the new regulations or not – and any content that succeeds virally will fall under the scrutiny of an army of geeks.

An interesting example is the recent viral video of a man dressed as Indiana Jones fleeing a giant LEGO boulder.

The fact that it wasn’t a genuine piece of user-generated content proved just as viral as the initial video, although this wasn’t perceived as a problem by all – as one Boing Boing commenter put it: “those internet marketers are TERRIBLE. How DARE they make something that we found briefly enjoyable.” What was more consistently disappointing for viewers was the fact that the boulder was merely made of styrofoam with a LEGO shell and certainly did not contain 5 million pieces as claimed by the video.

So while there is now a hard line defining what counts as unacceptable deception, that doesn’t mean anything on the legal side of that line is still a good idea. And if you’re not careful, Captain Disillusion will take you down.

June 5, 2008 Posted by mantim | digital marketing | , , , , , , , , , , , | No Comments Yet