Author Archives: Andrew

A study in UX atrocity – the Mercedes Benz COMAND SatNav

I drive a 2011 Mercedes. Like many of its brethren, most of the vehicle is a demonstration of engineering excellence: smooth, efficient, safe and smart (in both senses of the word). Its controls are clear, precise and beautifully weighted. The intelligent brake and start/stop systems are almost telepathic once you get used to them.

But there’s a fly in the ointment – the appalling COMAND SatNav system. It does have a comprehensive UK address database, and will get you close to where you want to go, but that’s about the best which can be said for it. The user experience is atrocious, to the point where one wonders if it was considered at all.

The problems start with the routing algorithms. If it can, the car will recommend a route using motorway-class roads to the maximum extent. That’s fine, but it will then pursue that strategy to ridiculous extremes. Get off the motorway a junction early, and the system will determinedly try to re-route you back on, even if there’s a direct main road from your current location to the destination, and rejoining the motorway would add many miles.

However that pales in comparison with what happens if you either select a non-motorway route, or despite its best efforts the car really cannot find one. Plan B is to impose strict optimisation for distance. This ignores quality and size of road, the number of junctions and all similar considerations. There may be a direct route on a fast dual carriageway main road, but if the Domesday book mentions a cart track which saves a few metres the car will try and take you down it instead.

On one occasion this led me off Dartmoor on a dark November evening down a road with a 100m stretch narrower than my garage, with high granite walls on both sides, both front-end proximity sensors sounding continuously. Recently we had to get from the A4 to the A36, both substantial main roads which link together on the edge of Bath. The SatNav found an alternative – using single track roads and an ancient toll bridge for which we had to hand over £1 in cash!

None of this would matter if there was some way to set your optimisation preferences. Well the COMAND system does, sort of. It actually provides four routing options:

• Dynamic route. The default, described above
• Fastest route. This is the motorways-only model on steroids. It will happily take you 20 miles out of your way to use one. Plus it doesn’t take any notice of traffic warnings!
• Shortest route. I shudder to think what this involves, but levitation and off-roading are safe bets.
• Eco route. There’s not much guidance on what this entails, but since the most economical route is either the most direct or the one on which you can maintain a good speed unimpeded, I suspect it’s going to be similar to the default.

There are also a few "Avoid" checkboxes which you can add to the calculation. "Avoid tolls" might be useful (but see below), most of the others don’t apply in the UK or near continent. I’m not even sure what a "vignette obligation" is – do I have to stop to take a photograph with dark corners at regular intervals?

The options really don’t help much. Instead of aggressive optimisation for distance why not road size/quality (which you can easily estimate in the UK from the letter in the road number) or the expected speed based on speed limits?

To add insult to injury COMAND does recognise some restrictions, but at random. When travelling from England to Wales it always asks about trying to avoid the Severn tolls, no longer in operation, which would mean a 50 mile detour off the very direct motorway route, but it didn’t ask in advance about the active toll in Bath. After yet another cart track, when we programmed in our route back to hotel it had the bloody cheek to ask "your destination is in an area with restricted access do you wish to continue?" The only challenge turned out to be the hotel gate which is a good 2.5m wide, a clearance of at least 30cm either side of my Mercedes, neither proximity detector triggered. My wife has started to anthropomorphise the system as actively mischievous.

Other UX issues are equally frustrating. You can only turn off voice guidance by a long press on the mute button while it’s actually speaking. Once off there is no way to turn it back on without cancelling and re-programming the destination. Surely it ought to be possible to provide a menu option to just turn the voice on and off?

At least the navigation function’s voice is a pleasant female with a clear British accent, who can correctly pronounce most English place names and will have a decent stab at Welsh ones. When speaking she gently fades the music and restores it afterwards.

She shares the box with the traffic warning "lady". I assume this second voice is also meant to be female, but it sounds like Stephen Hawking having a fight with a Dalek. She abruptly buts in, just silencing the music, and her attempts to pronounce UK place names are scary. Reigate (pronounced Rye-Gate) comes out as "Ree-a-gaa-ter", anything more complicated is unrecognisable. The best is probably the motorway between London and Southampton, the M3, which is renamed the "cubic meters"!

As a software architect I find it shocking that a closely related pair of systems, which do collaborate to adjust routes around traffic problems, ended up with two separate text to speech systems, a good one and an appalling one. How on earth did this mess get through the first stage of Mercedes QA? As a user all I can do is wince and try not to be distracted from the road trying to understand the encrypted names of traffic locations.

[That said, this is a schizophrenic car in other ways. It can be a saloon or a roadster. It can burble along with the best of luxury limos, but push the Sport button and it releases a snarling monster which has to be actively restrained. Perhaps the dual personalities of Saint Teresa of the Sat Nav and Mad Traffic Tracy are just a further expression. It would have been the ideal car for Dr. Edward Jekyll.]

Two things are apparent. The "designers" of this system (using the politest term possible) either didn’t think about usability, or didn’t care. They then either refused to get independent testing and review, or actively ignored its results. One can only hope the same has not happened with a more critical but less visible system.