Wednesday, June 01, 2016
I love this example from Snickers in Australia, which coincidentally uses two of my trends for this year, sentiment analysis and dynamic pricing.
The Snickers HUNGERithmithm analyses the sentiment of Twitter in Australia, and the more unhappy the mood becomes, the cheaper Snickers bars get at 7/11 stores.
Users just have to click on the button on the Hungerithm page of the Australian website, which then prints a money off coupon for a bar.
Friday, May 27, 2016
Last weekend Candace Payne's 'Chewbacca Mask' video acheived over 100mm views on Facebook.
It's the first Facebook Live video to do this - but very few will have seen it 'live'
'Live' refers to the way it was uploaded - no editing, no enhancement. Live is now the simplest way to upload video to Facebook (& YouTube, Instagram and Snapchat etc) - just open the app and shoot away.
The speed is extraordinary. It his 100m in about 48 hours, I think. By comparison Adele's Hello hit 23m views in 24 hours, and Psy's Gentleman (the follow up to Gangnam Style) took 4 days to reach 100m views in 2013, at that time the fastest ever video to 100m.
I suspect that Facebook put quite a lot of muscle behind this once they realised that it was so viral, and did a lot to help it get to 100m, including PR and more.
I also think it shows that Facebook is now the best place for a certain sort of video, better than YouTube, because it can reach so many people so quickly.
I think it also opens up lots of creative opportunities for using live video that is actually 'live' - think about what sort of times everyone will be looking at Facebook at the same time. How about (in England) right after England's first Euro 2016 game? Lots of people will go to Facebook to discuss the result (or look for distraction) and it would be a great place to respond immediately to what has just happened.
Monday, May 09, 2016
One of the things that really winds me up is the lack of decent original content on YouTube. Yes, there are the vloggers, but as far as I can see they're the new Kids' TV. I don't know anyone who watches them, apart from young kids. While they get lots of views, and as a result lots of press coverage, they're only influential to a small group of people - kids. Some of the people who are currently vloggers will go on to do other things when they're older - kids TV examples include Richard Bacon, Ant and Dec, and even Jeremy Irons - but I suspect many won't.
I quite often hear people at YouTube talking about all the great content, but they can never recommend anything that I'd want to watch. If I watch something longer than 5 minutes on YouTube it's probably an old documentary, for example one of Adam Curtis' early ones that some public spirited soul has uploaded.
By contrast, I'm spoilt for choice with great podcasts - Marc Maron's WTF, The Nerdist, Adam Buxton, ReplyAll, Scroobius Pip's Distraction Pieces, RHLSTP, Tech Tent, Talking TV... There's so much great stuff out there. It's easier to do audio only at a budget. These podcasts are usually enthusiasts talking to each other, which works fine in audio, and don't need much in the way of production. They also don't compete with my TV viewing - Night Manager, Line of Duty etc - because I listen when I'm cooking, when I'm on the train etc.
It's now looking like there might be more independent video aimed at an older age group. Since January, Louis CK has been filming and releasing episodes of his new show Horace & Pete on his site. It's a sitcom set in a bar, but more inspired by Abigail's Party than Cheers, and stars Louis and Steve Buscemi as the eponymous characters.
Apparently it cost $500,000 per episode to make, and it's being sold for $5 an episode on Louis's site, or $31 for all ten episodes. (remember that he pioneered the direct sale of his own stand-up film 5 years ago). It was 'dropped' an episode at a time from the start of the year, via an email to site subscribers. It's also likely to be on Netflix or HBO at some point - Louis has the rights to licence it to other people.
It's not a very replicable example - Louis is a big star - but it's interesting to see him making stuff for online release, like Jerry Seinfeld does - with sponsorship - with Comedians in Cars. Maybe if the audiences and advertising grows, and the costs of production come down, lots of my favourite podcasts could become web series, or develop new formats more suited to being watched rather than being listened to.
In the meantime, please tell me what great original content I should be watching on YouTube!
Friday, April 29, 2016
|Image source: http://demo.vhost.pandorabots.com/pandora/talk?botid=b0dafd24ee35a477|
Messaging apps are becoming more and more popular. WhatsApp has over 1 billion users, Facebook Messenger 900m, and WeChat 700m, and there are also lots of smaller ones like Kik and Line.
I've long felt that companies should offer Messenger platforms for customer services, but the issue is always one of providing support quickly without getting overwhelmed by millions of people wanting to get in touch.
Earlier this month Facebook announced that it was introducing an API for people to develop automated chatbots - effectively a way of answering questions automatically.
This is a good solution in principle - it allows companies to offer another access point, and gives them a platform more flexible (in some ways) than sms or email as it can be quicker, and can include clickable links, 'cards' and links to other apps, like maps. It also means that people wouldn't have to download a new app for each company they want to deal with, but just add the company as a contact on messenger, which is much easier.
I expect that early bots will be essentially a new search interface. Take the example of retailer John Lewis. I'd expect that a lot of the search queries are for closed questions to find the address or opening times. ('When does John Lewis close tonight?', "does John Lewis stock Chanel' etc). A bot could take 1,000 of the most popular search queries and for each use natural language AI to try to work out when this question (or similar) is being asked.
You can also see that, for a lot of closed questions like 'When does John Lewis close tonight?' you only need one search results, not pages and pages of them. Just as 'search' isn't great for vague, open questions like 'What's the best thing to do in London tonight?', it's not that great for questions with one correct answer.
I can also see apps being replaced by bots. Again, imagine a cooking bot, a cocktail bot, or even a city guide bot. Again you would use questions to select what you wanted 'A cocktail using Gin and Orange Juice', 'Things to do tonight' etc - which is essentially what apps like Mixology and YPlan do.
But I also think that chatbots could turn out to be an absolute nightmare. In the past month I've had problems (mostly minor) with DVLC, Ryanair, BT, Lambeth Housing and more, and in each case it's been almost impossible to get hold of an actual person who could answer a question, instead I was directed to forums, sent emails from accounts that you couldn't reply to, asked to press numbers on my phone keyboard and so on.
I fear that bots may make this even worse, and will lead to a whole new level of 'computer says no' pain. It also makes me remember the pain of dealing with (some) grown-ups when I was a child:
- What would you like to drink Daniel?
- I'd like a glass of water please
- Water? I've never heard of a little boy drinking water before. Who ever heard of such a thing! I've got some Coca Cola - would you like that?
But let's be positive. I think that the potential is huge, and it's being done successfully (apparently) in China, where some cities are connected via WeChat - you use messenger to arrange medical appointments and so on.
Friday, April 15, 2016
Monday, April 11, 2016
Thursday, April 07, 2016
Monday, April 04, 2016
Instagram is going to start using an algorithm to try to show users the posts they are most likely to appreciate, rather than just use a chronological timeline as it has until now.
The chronological timeline is simple and easy to understand, but it has drawbacks - if I don't look at Instagram within an hour (or a few minutes) of one of my friends' updates I'm unlikely to see it. An algorithmic timeline could be better - this is what Facebook uses, and it puts important news like engagements, and new babies at the top of my newsfeed.
As services like Instagram, Twitter and Facebook get bigger people follow more users, so there is more to see, but also more clutter.
The drawback of an algorithm is that can make some posts invisible - essentially if Instagram thinks that I'm not likely to like a person's post, it's likely to bury it. It's a black box in that it's not clear how the algorithm is worked out.
(I remember a story of someone getting so fed up that their posts about the unrest in Ferguson weren't getting seen on Facebook that she changed her status to 'married' so that it would appear in friends' feeds)
Some Instagram users are getting panicked that their friends won't see their posts and have been asking friends to turn on notifications so that friends will get an alert every time they post. This clearly isn't workable - imagine if you turned this on for 20 people - but it shows that people will try to game the new algorithm, just as they've been trying to game the current system by using lots of hashtags, tagging lots of friends in their pictures, even when they're not there, and messaging friends to ask them to like pictures. In my experience the most popular accounts, beyond the celebrity ones, are ones that are very specific to a genre - cats, landscapes, desserts - you get the idea.
What I suspect will happen once we have the new algorithm is that 'the rich will get richer' - if you currently have a lot of followers and get a lot of likes, you will start to get even more. Instagram is doing this both to make it a better experience for users, but also to make the platform a more engaging place for casual users, so it will show you more of what you have already said you like. This way you will return more often.
(Instagram already uses algorithms of course, in the 'search' tab, to show you popular pictures like the ones on the sites you already follow, and in recommending new accounts for you to follow, based on who your friends follow and sites that are similar to the ones you follow or have visited).
So what is likely to be included in the Instagram algorithm, to choose what pictures to show you?
The strength of connection to the account, including on Facebook - so you'll get to see more pictures from people you know, and even more from people you know well.
The frequency of linking pictures - if you often like Nigella Lawson's pictures you'll be shown all the news ones
The frequency of liking a type of picture - if you like pictures with the hashtag #dessert you'll be shown more #dessert pictures from accounts that you follow
The location - if you've posted a picture somewhere, you'll probably be shown friends' pictures from the same area.
The frequency of posting - Instagram will want to reward the people who put the work in - and besides their pictures are likely to be better.
At the same time I think it's likely that Instagram will start to penalise the spam tactics that people currently use to game it that I mentioned above - too many hashtags, too many users tagged - in the way that Facebook did in penalising the brand posts asking questions ("It's Friday - what's everyone doing tonight?) and other ways of fishing for interaction.
Instagram has always been gamed to a degree, and now that there's a 'black box' algorithm the gaming will increase - companies will offer to make brands' instagram pictures more popular - but essentially if you post good pictures, often, then you should be fine.