Posts Tagged ‘statistics’

Masabists: Illustrating Gartner’s Q3 2010 Global Handset Shipment Report

// November 15th, 2010 // No Comments » // Mobile

This post originally appeared on the Masabists blog.

I created a quick infographic this weekend to illustrate the trends shown in Gartner’s recent Q3 2010 handset shipments report:

Gartner 2010 Q3 global handset shipments

The bluer a company is the more its market share is growing, the redder a company is the more its share is being eroded (even if handset shipments themselves are up) – which illustrates nicely the slow decline of the old guard, as a more diverse mix of companies invade the handset space. Fragmentation is here to stay.

Please comment on the original post

Canon vs. Nikon – Unscientific Stats…

// May 30th, 2010 // No Comments » // Photography

I’ve been noticing two camera related things when I go out recently – firstly there are a huge number of people carrying SLRs around these days, and secondly they appear to be almost exclusively Canon.

Last weekend I decided to quentify the second point with a little test.  I walked down the (crowded) Embankment in London from St Paul’s Cathedral to Westminster Bridge on a sunny afternoon and counted up every brand I saw, 25 SLRs in total over about 20 minutes split like this:

If you were to take these stats at face value, it would look like Canon are outselling everyone else in the market combined by a ratio of two to one – no mean feat!  Three of those Canons were serious / pro-grade (L lenses), compared to one of the Nikons (and possibly some of the Others – I don’t know enough about the other brands to be able to say).

This is a very small sample taken in one place, but that one place was the middle of a very connected cosmopolitan city during tourist season so it probably represents a reasonable slice of the international market, with obvious UK bias.

I suspect the real global sales figures are not so extreme, but I still find this an interesting trend that quantifies something I’ve been thinking for a while.  Whilst Nikon currently has a signficantly higher mindshare among professionals than its overall market share, how long could they keep that up if most of the entry levels being sold into the market were Canons?  This generation’s pros are already set in their ways, but the next generation will be picking up whatever camera they can lay their hands on – either 2nd hand or their parent’s unused SLR bought back when they thought having an SLR would automatically make your pictures better.  Chances are it’ll be a Canon.

Anyone setting out to buy their own SLR will often end up buying whichever brand their mates have, because of advice bias from those friends – they’ll understand their brand better than the other – and also the potential to share kit.  This would again seem to reinforce market dominance for any brand that can gain signficiant market share.

The other effect this might have is to generate a skew in resources available for R&D.  No manufacturer can afford to have many new lenses being developed simultaneously, but if the sales skew generated significantly greater revenues for Canon it would over time start to pull ahead in the range of lenses it can offer, which might presumably have something of a feedback effect on its proportion of sales.  I have no idea how much of an issue this would actually be – both manufacturers have an excellent array of glass covering most needs, and 3rd parties like Sigma plug a lot of the gaps (like long tele-zooms); once you go above the really basic kit lenses though, most are sitll geared towards full frame sensors and a manufacturer who could make a more compelling upgrade path for crop sensor bodies might be able to do better.  I’ve no idea really but you can potentially see how Nikon might get upset.

However, it’s worth remembering that nothing stands still and a disruptive technology can easily humble a market leader within a few years – Apple is busy proving this in the mobile phone market, after all.

The most obvious candidate for that in higher-end photography would seem to be EVIL bodies – basically, everything people think of as “SLR” (interchangable lenses, easy to access manual settings, etc) without all the actual tedious and expensive bulk of  aSingle Lens Reflex moving mirror and prism.  Canon haven’t expressed any interest in that market that I’ve seen, because their SLRs seem to be selling very nicely thankyou and they’d probably rather not cannibalise those revenues before they have to – who knows if that’ll turn out to be a mistake.  Then again, they may have a crop sensor EF/EF-S compatible EVIL body just waiting in the wings for the day that their competitors have made the case for the concept to consumers, which will flood onto the market backed up by Canon’s entire lens catalogue.  I’d buy one as a backup body immediately.

Data Visualisation – Global Gender Balance

// February 18th, 2010 // No Comments » // Creative

I’ve spent a few hours playing with Tableau Public, a free version of the rather expensive Tableau data visualisation app, and it’s pretty good. After a random discussion of how Estonia has, according to the Economist world stats book, 84 guys to every 100 girls, I prepared this map in a couple of hours as an exercise in understanding how Tableau works:

Click to see full size - global male:female ratios, scaled by population size

The map shows the male:female ratio of every country as coloured crosses, from red (more women) to blue (more men); the size of the cross is proportional to the country’s population.

Quick Caveats

Tableau Public is a cut-down version of commercial software, with no ability to save files locally. In theory you can publish your diagrams to the web on their site, but that feature was a bit broken when I tried it.

This means that you have to retrieve your diagrams using screenshots, and I have no idea what the legal implications of doing that are if you intend to use them for anything but personal interest and satisfaction! It does constrain the quality a little, too.

Source Data

Being lazy, I didn’t want to type in all of the raw data from the Economist book, so I pulled it from Wikipedia: population data came from here and sex ratio from here. Copying the tables directly into Excel brought a load of unwanted links and images, so I copied into Notepad++ – where it appears as tab separated values – and reloaded as a TSV file via disk.
The country names had a few extra spaces and other characters in them – I pulled these out with the LEFT, MID, LEN etc functions.

Tableau theoretically understands full country names, but I had mixed success getting this to work, so I imported a lookup table of two letter ISO 3166-2 codes, which are also supported and are unambiguous.

I then used VLOOKUP to pull together all this information into a single Excel sheet (remember to turn range lookup – the fourth, optional argument – off); as the data came from diffferent sources, there was a little messing round standardising country names. The finished spreadsheet is here if you want to play with it.

Visualizing in Tableau

It really helps to look at this tutorial video before starting anything – Tableau starts with quite a blank slate!

Open up the XLS in Tableau, and it’ll make a first stab at identifying what is what from the Excel column formats. You’ll see the fields seperated into text and numeric lists down the left hand side.
If you don’t see a globe next to the Country Code field, right click, go to the Geographic Role submenu, and select Country (ISO 3166-2). Tableau can now map this to a geographical location automatically. Also ensure all numeric columns are recognised as numeric with the Change Data Type submenu.

To recreate the map visualisation,  follow these (approximate!) steps:

  1. Select the Country Code and Total fields, and click on the Show Me! button.
  2. Select the map diagram type (near the bottopm of the list). You should see some dots across a map.
  3. Drag the Population column over to the Size box, right click and select Dimension to scale the dots by population.
  4. Drag the Total column to the Colour box and they should become shades of green; further down is a green graduation which you can click on to change, and adjust to a red-blue graduation. Click on Advanced and set the midpoint to 1.0, to make the middle grey represent a 1:1 ratio.

Your left hand panels should look a little like this now:

This is really just a very high speed starter and I’ve barely dipped below the surface – I’ve got some quite complex business plan data that I’ll be dropping in later to experiment further.

Masabists: Hacking AdMob Stats

// September 3rd, 2009 // No Comments » // Mobile

This post was originally featured on the Masabists blog, and was voted joint top post in Carnival of the Mobilists #190.

Handset statistics are notoriously hard to come by – only the operators know what handsets are actually being used by all customers on their network, and they won’t tell. Every other purveyor of statistics has an inherent bias, which makes them more attractive for some analyses and less for others. For example:

  • GetJar make comprehensive statistics available for who is downloading applications from their site and have huge cross-operator volumes of downloads
    • …but many of these are downloaded via the web and synced through a cable, a highly unusual activity for most mobile users
    • …which puts a huge skew on the data – I’m pretty certain Amoi aren’t actually larger in the UK market than RIM’s Blackberry.
  • Bango publish a Top 20 handsets list every month or so, with a skew for their operator relationships
    • Top 20 is great but there are plenty of handsets below that which are still worth supporting
  • AdMob have some immensely detailed metrics reports, but again the operator partner skew was always very visible
    • I’m just not convinced that anything by ZTE is actually in the Top 10 most popular UK phones!
    • I’m also unconvinced that the iPod Touch is much more popular than the iPhone…

User Profiles

All of these statistics have great value, but none of these companies can actually give an overall picture fo the entire market. Each set will be skewed heavily based on the type of user attracted to the service and the operator relationships that service has. To find out the overall market picture fudge them all together and treat with care – but by profiling each site you can find out more useful information for specific needs.

AdMob, as the main provider of mobile web advertising, offer a very good view of the mobile data user, who would also be an obvious early adopter of any mobile service requiring net access, be it web-based or a networked application.

Traditionally, AdMob have just given a Top N list of device models and some aggregate manufacturer numbers, which were never enough to tell anything useful – their strong relationship with MVNO Three clearly led to some very obvious weird “popular” handsets showing up and called all of the data into question.

I had always dismissed them at this point and moved on. Fortunately, though, when you look inside the reports they publish per-operator statistics for each manufacturer – so we can actually derive some meaningful national conclusions!

Hacking The Stats

So – why hack? Don’t we have everything we need?

Sadly not – the July ’09 report is a very nice professional PDF given graphical summaries of the manufacturer breakdown per-operator, as a series of proportions in stacked columns:

AdMob handset statistics - UK operator split

Whilst it is flattering, as a Jersey boy, to see Jersey Telecoms given equal weight to Vodafone, I have a sneaking suspicion that the entire population of Jersey is equivalent to a rounding error when counting Vodafone’s UK customer base.

What we need is to convert these into national numbers, by weighting all of the proportions by the size of the operator. The only way to do this with the public information is to screenshot the graph from the PDF at the largest size you can get it, and count the pixels for every bar – tedious, but luckily for you I’ve done it for you.

Telecoms Market Research provide some useful figures from Q1 2008 for each operator (not up to date, but close enough for these purposes):

Once we aggregate all of these weighted proportions, we end up with the following rough proportions for manufacturers among early mobile web adopters:

Handset manufacturer proportions for mobile web users in UK June 2009, derived from AdMob statistics

It won’t help everyone, but hopefully it is of some use for anyone struggling for some statistics!

Please comment on the original post.