... in data

Retro anti-communist fear-mongering

16 Jun 2012 | 82 words | advertisement data

Ran into this ad while reading the economy section of yesterdays NRC on the toilet this morning. It is easily the most amazingly stupid copywriting i have come across in a while. i mean seriously who on earth still worries about ‘the soviets’ anymore?

Stupid Deloite ad

‘i was 250.000 miles out in space. imagine what could have happend if the Soviets interfered in our data. How do you protect your data today?’ Eugene Andrew Cernan. Last man on the moon, 1972

What airlines can learn from how casinos are run

18 Nov 2011 | 627 words | consumerism data airtravel

So the latest Planet Money podcast consists of this absolutely fascinating interview with Harvard professor turned casino CEO Gary Loveman. The interview is fascinating on a number of levels, chief among them this detailed look into data-driven behavioral manipulations:

Gary Loveman: Everybody who gambles knows that the house has an advantage. They are not unhappy that they loose any more that one is unhappy that they pay $40 to walk into the magic kingdom at disney. They recognize that there is a fee to provide all this but they get very unhappy if the loss is surprisingly severe – or as a statistician would put it, if they are in the tail of the distribution rather than the mean of the distribution, right?

So we have programs where a customer comes to play, we can observe real-time who in the casino is there on his first visit. It’s all though that [loyalty] card. So if Jude puts that card in and our system recognizes that this is the first time that we met her, a flag goes up to the casino management: New customer, Jude, playing the 25¢ video poker machine number 275 in the front left corner of the casino. And people on our staff will then begin to monitor all such people.

And we focus on those whose gaming results are way out in the negative tail, that is they are playing a slot machine where they should be getting back $94 out of a $100 but they are only getting back 50 out of $100. I know that this is a bad experience for Jude and she is going to feel that this was a very poor first visit. So the question is then: what are you going to do about this?

And you can imagine a couple of answers: one is ‘nothing’ which is the normal answer in consumer life. The other is that someone comes out, introduces themselves to Jude and says: ‘welcome to – lets say – Harrah’s St.Louis‘. Now the first thing that Jude is going to say is: ‘Hi, this place sucks! i am having a lousy day, i can’t win anything’.

Then the question is what do you do then? first of all you apologize, you say ‘Gee i am sorry to hear this, we do not like our customers to have a bad experience, how can i help?’ and you can help by buying Jude dinner, by giving her additional coin to play in the slot machines because at some point the law of large numbers will bring Jude back to the mean. You could offer her a room in the hotel or a ride home in the limo or any number of different things.

We do these kinds of interventions and then we run tests and control against it to see whether of all the people having a bad first experience, those who have a visit from one of our staff are more inclined to come back for the second visit. And not surprisingly they are dramatically more likely to come back.


Gary Loveman: Now when i was an academic i did this with airline data. So imagine through frequent flyer data that we observe that Jude has a visit on an airline, the flight was delayed and her bag was lost. Is it any mystery that she hates the airline? No, she hates the airline, you know that. So if you want her to visit the airline again the next time, how would you treat her? You would want to acknowledge that the last trip was a bad one and you would want to try to do something to make it up. It is exactly the same idea. But the airlines never do that.

Money & Speed - show me the data...

Two weeks ago the 50 minute documentary ‘Money & Speed‘ by Marije Meerman aired in the tegenlicht (backlight) series of the VPRO (a dutch public broadcaster). ‘Money & Speed’ casts a light into the world of high frequency (or rather algorithmic) trading, a subject that i have been fascinated with for a while. The documentary attempts to explain the event’s behind the 6 may 2010 ‘flash crash‘ and does so by interviewing a number of experts that are more or less involved in the (aftermath of) the crash.

money & speed title screen

While ‘Money & Speed’ stops short of pointing out who (or what) was responsible for the events on may 6th 2010 (there are conflicting theories and tegenlicht’s prime source Eric Hundsader of Nanex clearly has been advised not to mention on camera whoever he thinks caused the events) it is quite a remarkable documentary that gives an fascinating insight into a world that is most likely completely unknown to more the 95% of the television audience.

I also really like how Meerman has woven George Dyson and his theories about differences in time perception between humans and computers into the narrative of the documentary. That guy certainly makes more sense than Kevin Kelly.

What makes ‘Money & Speed’ even more interesting is that next to the broadcast version the VPRO (together with catalogtree) has released an iPad version of the documentary in the form of a free iPad app1 that contains the entire film in high quality as well as a number of supporting assets such as infographics, short bio’s and a glossary of terms used in in the documentary.

At the first glance (and certainly for a first attempt) the app version of ‘Money & Speed’ seems to be crafted really well. The app is really responsive which makes pulling up extra information while watching a pleasure. You simply click on one of the thumbnails in the top right corner and the video stops instantly and resumes instantly after you are finished reading the text-box.

money & speed

The absence of noticeable delay makes this process work very well (it does not feel as an interruption of the flow at all even though the video itself stops). In the case of a couple of the bigger infographics there is an annoying delay before the infographics appear, but the transition back to the video is almost instantaneous.

Unfortunately those big info graphics (tick for tick visualizations of three different stocks, a visualization of the fall of the dow jones index, a map showing possible locations for financial data processing centers in the vicinity of New York City, and a delayed real-time stock market data map) are not really connected to the storyline of the documentary.

These are clearly the most labor intensive parts of the ipad app, but they do not add to the understanding of the events described: The delayed real-time stock market data map is a complete waste of resources (although it has a vey nice darkish background noise that i am listening to right now) and the map of the data centers, while visually striking does not even show the locations of the data-centers that are depicted in the documentary. The tick for tick visualizations (visible on the ipad in the screenshot above) while impressive simply fail to convey much information (Alpher suggests that this is because ‘the information density is low’).

Given these shortcomings one might come to the conclusion that adding extra visualizations to a documentary like Money & Speed provides little extra value to the viewer. I am pretty sure that this is the wrong conclusion and that the real problem at hand is that the creators of ‘Money & Speed’ have simply selected the wrong data to visualize2:

money & speed

Towards the end of ‘Money & Speed’ Meerman juxtaposes two different views on what happened on the the 6th of May. That of the official SEC & CFTC report and that of independent data analysis done a by the Chicago based financial data services firm Nanex. Both identify different triggers for the flash crash and their disagreement seems to come from the granularity of time applied to their analysis. While the SEC & CFTC report works with seconds Nanex claims to be able to identify events on the nanosecond level that are averaged out (an thus invisible in the SEC & CFTC report) when looking at the events with less granularity. During this part of the movie Eric Hundsader is shown pointing at this data on one of his computer screens and if you ask me it is this crucial data-set that should have been included in the iPad app instead of the tick for tick visualizations:

money & speed

Still the ‘Money & Speed’ iPad app (the VPRO calls this format a ‘touchdoc’) shows a lot of promise. i can imagine that once the technology for integrating linear video content with additional assets like data visualizations becomes more common, attention will shift towards editorial concepts that better integrates the extra assets with the main narration. In an ideal case scenario that would mean that the underlying research assets of the creators are available to consumers both as extra information but also in order to independently verify claims made by the documentary makers. Given this ‘Money & Speed’ tastes like one possible future of investigative documentary film making.

  1. At the time of writing the app is only available in dutch and only in the dutch app store. the VPRO is working on an english language version that will be available in other app-stores for a fee. It is free in the Netherlands because for regulatory reasons the VPRO cannot charge the public for content it has produced with public broadcasting money. This is not the case outside of the Netherlands and it will be very interesting to see how much income ‘Money & Speed’ will generate outside of the Netherlands. I could very well imagine that a documentary-app (is that an existing category?) of this quality could easily generate income that exceeds the production costs (of the app, not the entire project). ↩︎

  2. Of course this might not have been entirely voluntary. It is possible that this is the only data that they were authorized to include or that they had to make choices which data to visualize really early in the project and as a result could not react anymore once the more interesting datasets appeared. ↩︎

Data retention (a.k.a Northwest thinks that i am a vegetarian)

11 May 2008 | 298 words | airtravel data food

So just before the meal service during yesterday’s flight to Boston i get a meal tray handed over by a flight attendant. Turns out they have me on the list for a special (vegetarian) meal. Now the strange thing is that i had not ordered such a meal (even cheked my reservation confirmation and it does say ‘no special requests‘ so Northwest’s reservation computer must have come to the conclusion that i am (still) a vegetarian all by itself. Probably based on the fact that i was a vegetarian when i flew NW for the last time (5 years or so ago). All of this would not have been all too unpleasant if Northwest would not have fucked up the definition of a strict vegetarian (VGML) meal. While the main dish had a VGML sticker on it it turned out to be pasta in three cheese sauce. Now everybody who knows me a little bit will know that this is about the worst thing i can imagine in terms of food. Just thinking about pasta in three cheese sauce makes me feel miserable.

Of course the flight attendants did not really care about the whole thing because (a) they simply refused to believe that i had not ordered a special meal (‘is that your name and your seat on that sticker, sir? – eh, yes… then it is your meal’) and (b) because they refused to acknowledge that three cheese sauce is not a a VGML dish (‘does it say VGML on that sticker? – eh, yes… – then it is a vegetarian meal’).

In the end my neighbor offered me his chicken dish (thanks!). now all i have to do is to figure out how to tell the reservation system that i am not a vegetarian anymore…

My mobile phone provider thinks i am an tsunami victim...

15 Feb 2005 | 645 words | data surveillance mobile networks india

… and has a pretty poor sense of geography.

Got my monthly bill from my mobile phone provider (orange Netherlands which an almost completely different entity than the orange operating in Bombay. They just use the same brand identity and parts of the companies are held by the same holding: hutchinson whampoa limited).now getting a monthly phone bill is nothing special, but getting one which informs you have been a tsunami victim an that they therefore credited you with €42.05 for ‘possible extra phone expenses related to the tsunami’ is somewhat strange. especially if you have not been affected by the tsunami as you where safe on a jet airways flight from Bombay to Delhi when the whole thing hit the coasts of India and Sri lanka. now it is no secret that mobile phone providers record the location data a mobile phone generates, but at least under dutch law this data cannot be used by them for anything else except invoicing purposes (and they have to retain it for law enforcement purposes for half an eternity or so). and as far as i can tell remember i did not generate any data at all during the time the tsunami hit. i switched of my phone on the 7th of december in delhi, replaced the sim card by and airtel india branded one and switched back to my orange sim on the 11th of january at the airport in delhi to make a couple of calls before flying back to amsterdam. so as far as they can tell i have been delhi all the time which thanks to its inland location and altitude is probably even less tsunami affected than amsterdam. of course they could also have a look at my call record and would discover that i did not make any calls during that time either….

For sure i am not comfortable with my phone company using this (non) data to become some kind benevolent paternalistic entity that does a little monetary intervention here and then when things get a bit more bumpy that the average european can be expected to be able to cope with. and if i should ever want something like this i will buy a travel insurance policy. i can imagine how these public relation geniuses at orange have seen this golden opportunity to build up a personal relationship with their customers. given all the suffering that this catastrophe has caused to all these poor western tourists (which is the real reason why we Europeans have donated so much to the relief funds), they went strait to their data-minig department and told them to get them a list with all their customers that have been in the hit by the tsunami. so the data miners go a and get themselves a CNN info graphic that shows the tsunami affected countries run a couple of queries based on this information and come up with a list of tsunami victims among their customer that goes back to the marketing department and there the amount of money available for this stunt is divided by the amount of victims and the billing department is instructed to credit each victim with the resulting amount. and now they are probably all excited how they come up with a way of effectively allocating ressources where they are most needed… credit where credit is due!

When i called their customer center today to complain about using my location data for something which they are not entitled the call center agent simply failed to understand my problem. he could not see how i could complain as i benefited from this measure of theirs and told me that i was ungrateful. and when asked about why i was getting the credit when i was in delhi the whole time he told me ‘well that is in the region isn’t it?’

meanwhile... is the personal weblog of Paul Keller. I am currently policy director at Open Future and President of the COMMUNIA Association for the Public Domain. This weblog is largely inactive but contains an archive of posts (mixing both work and personal) going back to 2005.

I also maintain a collection of cards from African mediums (which is the reason for the domain name), a collection of photos on flickr and a website collecting my professional writings and appearances.

Other things that i have made online: