RSS

Système D

May 26
Permalink

TomTom: from PND to FUD

Satnav wrong. Use brain.

(TomTom data is such high quality that people have to put up signs like this. Photographed in Slaithwaite, West Yorkshire, by Tim Brown, licensed CC-BY 2.0.)

You can tell you've got your competitors rattled when they start producing knocking copy about you. It's a testament to the success of OpenStreetMap that TomTom, the troubled, declining manufacturer of satnavs (PNDs) and geodata vendor, has published a rather cynical example of FUD about "open-source maps".

PNDs are Personal Navigation Devices. FUD is Fear, Uncertainty, and Doubt.

It's not really for me to say whether TomTom is good at manufacturing PNDs... but happily, it's really not very good at creating FUD.

Vandalism

One of TomTom's allegations is that "open-source maps" are vulnerable to vandalism:

Indeed the major benefit – the community aspect – has itself presented problems, leaving maps wide open to attack. A highly publicised case saw a leading provider suffer over 100,000 individual attacks, including reversals of the recorded directions on one-way streets.

This is more than FUD. This is plain wrong. TomTom is clearly referring to the case where OpenStreetMap caught Google contractors vandalising OSM. But let's look at what Steve Coast and Mikel Maron's original blogpost actually says:

Two OpenStreetMap accounts have been vandalizing OSM in London, New York and elsewhere from Google’s IP address, the same address in India reported by Mocality.

The most obvious vandalism started around last Thursday last week from these particular users however it may take us some time to do a full analysis. In fact over the last year we have had over 102 thousand hits on OSM using at least 17 accounts from this Google IP.

Can you spot the slight flaw in TomTom's argument? Not all of those "102 thousand hits" were "attacks". In fact, no more than a couple of dozen were - and those were swiftly spotted, and fixed, by the OSM community.

Of course, it's flattering that Google is so interested in OpenStreetMap that they consulted it over 100,000 times. Maybe they were dissatisfied with their current data provider?

Looks like they were. In fact, they've dumped their old data provider in both the US and Europe. Who was the data provider they dumped? That's right, TomTom. Turns out that their "expert map-makers" aren't quite so irreplaceable.

As a side-note (and yes, I know, "the plural of anecdote is not data"), it took the satnav/PND manufacturers several years to get the one-way direction right for the street 50 yards from my home. OpenStreetMap has always had the correct direction.

So I'm not sure that OSM should take too many lessons from TomTom about such "extremely dangerous mapping errors". I think it's pretty certain that the users of satnavs in these cases (1, 2, 3) are more likely to have been using TomTom data when they came to grief than OSM. TomTom would be better advised to put its own house in order before criticising others.

Data coverage and quality

The second thrust of TomTom's attack is that OSM data is of lower coverage:

In one particular instance, a leading open source map was compared against a professional TomTom map, and shown to have a third less residential road coverage and 16% less basic map attributes such as street names.

This criticism is fair enough. OpenStreetMap is a growing project. If you carefully choose the place to make your comparison, it's not difficult to find somewhere where OSM has just 84% of the streets of TomTom.

For an objective view, take ITOworld's OSM Analysis (requires login), which compares OSM against Ordnance Survey data in the UK. 283 areas have better than 84% coverage (the majority in the high 90%s); 125 have worse. So the average figure, in the UK at least, is considerably better than that cited by TomTom - but it's not hard to find areas that back up their contention.

This should be comforting to TomTom... were it not for the inexorable climb in OSM's coverage. 16% this month may be 10% later in the year, and so on.

However, TomTom's second criticism is wrong right now.

Worse still, it blended pedestrian and car map geometry, and included ‘a high number of fields and forest trails’ classified as roads.

This betrays a complete lack of understanding of OpenStreetMap data.

Pedestrian and car map geometry is intentionally blended in OpenStreetMap. OSM is one, seamless dataset. Pedestrians can walk along most roads: therefore roads are linked to paths to preserve pedestrian routing.

The attributes ('tags') for each object indicate whether it is accessible to cars, pedestrians, or both. It is absolutely standard practice when working with OSM data to filter on the basis of tags, so that you only get the data you need. So if you don't want to drive down forest trails, you simply filter out 'highway=track'.

This approach enables sites like MapQuest Open to offer planet-wide routing for pedestrians and cyclists, while sites using TomTom data only have pedestrian detail in certain urban areas. According to a recent study, even back in 2009,

within more densely populated areas, OSM data offer more overall data than the commercial provider [TomTom]

concluding:

for all cities analyzed in the US and Germany, OSM data resulted in shorter shortest paths [i.e. more complete and efficient routing] [than] pedestrian commercial data sets.

Compare the 2012 datasets, and OSM is richer still - and not just in densely populated areas.

TomTom doesn't cite the survey to which it refers, but since their article first came to light, Pascal Neis recognised it as his paper. And once again, TomTom has misinterpreted its conclusions - whether deliberately or through ignorance, who knows. The original context is this:

in the middle of 2010 OSM surpassed TomTom in the total number of streets recorded. However, a high number of field and forest trails caused this advantage for OSM.

Again, these are not "classified as roads", as TomTom claims. They are tagged as "highway=track". Pascal's study chose to consider these as streets (and, if you drive a 4x4, they are), but there is no suggestion that the data is misattributed in OSM. If your application wants to only focus on tarmaced roads, you simply filter out "highway=track".

It's a shame that TomTom didn't have the courtesy to attribute the study by Pascal (with Dennis Zielstra and Alexander Zipf) so that readers could make up their own minds, instead choosing to take an out-of-context quote. Perhaps they were a little afraid that people would read on to the conclusion:

According to our projection for the future, this discrepancy should disappear by the middle or end of 2012, and the OSM dataset for Germany should then feature a comparative route network for cars as provided by TomTom.


If you'd like to talk OpenStreetMap with someone who, unlike TomTom, actually understands it, I'm now undertaking consultancy work: find out more.