Monday 24 September 2012

What's wrong with Apple Maps (and when they will fix it)

Anatomy of a problem

Since the release of iOS6 last Wednesday it's been a case of Schadenfreude all round. However I've seen little analysis of what exactly is wrong with Apple's mapping apps. What people are missing is it isn't just a case of "Maps is crap", but rather a number of separate issues, some avoidable but others not.

Having covered TomTom and Tele Atlas an an equity analyst on and off since 2005, I think I can offer an informed view on these issues. So what I've done in this post is to simply lay out the problems, identify what's gone wrong and think about what Apple can do next to fix it.


IssueImpact of issueDifficulty to fixTime to fixHow to fixExample
DATA ISSUES
Poor road detail (TomTom data)xxxxxGet new data from TomTomTickhill
Poor road detail (non-TomTom data)xxxxxxxRevert to local sourcesLack of Tokyo detailGrand Cayman land shapeSanto Domingo road names
PoIs in wrong locationxxxxxxSource improved PoI dataHelsinki parkStadium in wrong placeStarbucks in wrong place
Lack of categoriesxxxxxxxSource improved PoI dataStratford CBD labelled a park
Local businesses not listedxxxxxxxCrowdsource data, more listings dealsBurger King in churchIBM sock picture
Use old namesxxxxxxRevert to more recent dataValley Forge State Park nameHoly Ghost HospitalAldwych tube
TyposxxxxxCrowdsource dataDoncaster named DuncasterFlorence Cathedral typo
TEXTURE / IMAGING ISSUES
Weird 3D artefactsxxxxxxxxHand edit data, get better meshBrooklyn Bridge (OK on Nokia Maps)Severn bridge bad 3D modelEnd of Oxford Street 3D data
Missing texturesxxxxxxxGet more textures85 Walmgate, York, England
Low res texturesxxxxxGet newer texturesYork, England
Non-matching texturesxxxxxxManually grade/match texturesYork, England
Clouded satellite imagesxxxxxxGet newer texturesColchester
SOFTWARE ISSUES
Poor routingxxxxxxxImprove / acquire bette algosDirections to San Jose runways
Poor searchxxxxxxxCrowdsource more searches, improve engineAntwerp searchStreet name gives wrong searchMan Utd FC wrong search

Note: More xxx = Biggest impact / harder to fix / will take longer to fix

What's wrong with Apple's maps? (and where did they go wrong)

As you can see from the table above Apple doesn't just have one issue with maps, it has a number of issues. I've broken these down into three rough categories.

1) Data issues

  • Poor road/land data data: Apple maps features a number of cases where roads are non-existent (e.g. the example of Tickill in the UK, flagged in this BBC News article). This also affects coastlines and geographical features (e.g. the shape of Grand Cayman here).
  • Missing or inaccurate Points of Interest: PoI is map-speak for a specific location, normally denoting a place of business someone might want to visit. There's a bunch of different problems here. Either they are in the wrong location, wrongly categorised or not there are all.
  • Typos in place names: Sometimes Apple seems to fall back on outdated data for place names. Other times there are straight (and embarassing) typos.
I think there are a number of different issues here. For some of the problems with roads Apple seems to simply have poor quality data (for a full list of their sources see this page). This is particularly surprising in developed countries like the UK as they are using TomTom mapping data which should have highly accurate road data. My theory is that they have overlaid data from other local sources (e.g. in the UK they cite the Ordnance Survey as another source) which is less accurate.

In less developed countries the opposite is true. Here they appear to have used TomTom data, not realising that they do not have complete coverage and only include major trunk roads. For example Belgrade in Serbia is missing only has major trunk roads. If you zoom to the city in TomTom's data set you see a similar issue.

Missing or inaccurate Points of Interest comes down to one things - either their business directories (sourced from Acxiom) aren't accurate enough. You can argue about whether this was always going to be the case and they could only crowdsource the data once they were live, or whether they should have gone to more local organisations and get the data. Either way the data either isn't there or hasn't been cleaned up enough.

Typos in names are a subset of this. Like with the first category of road data, I suspect they have been overlaying older less accurate sources onto the TomTom data (which is generally pretty good).

2) Texture/imaging issues

  • 3D artefacts: Apple's 3D view, involves overlaying satellite imagery over a 3D mesh. Unfortunately for some objects, particularly for man-made structures like bridges, which leads to some weird 3G artefacting (e.g. on the Brooklyn Bridge). Also where the highest level of 3D coverage (including buildings) ends there is an obvious disconnect.
  • Missing textures: For the Hybrid and Satellite views Apple relies on data from Digitalglobe. Unfortunately there are a number of issues with this, in particular missing textures at high levels of zoom, lo-res textures, non-matching areas where different photos are stitched together. Also some textures are obscured by clouds entirely.
Apple's highest-res 3D coverage (which leads to the worst artefacting issues) was acquired with C3, a Swedish company which was also behind Nokia's great 3D city maps. The problem is that the mesh they are using doesn't look high enough resolution to capture man-made objects like bridges, or hasn't been tidied up enough. It is interesting to note that the image of the Brooklyn Bridge shown above the Nokia Maps engine (based on an earlier version of the same tech) renders the bridge correctly - I suspect Nokia's model was tidied up by hand whereas Apple didn't apply that final bit of finish.

The issue with 3D coverage ending is, to be honest, something of a non-issue. C3 coverage is very limited (and I think will always remain so given the requirement to fly over with a plane rather than use generic satellite imagery, and the need to tidy up images by hand). This means there will always be a cutoff.

Missing textures, like missing PoIs, looks again like a straight failure to get adequate imagery. This should be a commodity - I remember viewing top-down imagery on Multimap back in in 2000. There is no excuse for low resolution, missing or clouded textures - either the vendor or Apple screwed up here.

Non-matching areas where photos are stitched together is more forgivable. This is an issue which Google used to have on Google Maps and Google Earth all the time (IIRC they only updated color-graded textures to get round this a year or so ago). Largely a non-issue.

3) Software issues

  • Poor routing: Turn by turn driving instructions sometimes take people to the wrong destination (or not quite the right destination - e.g. the runways of San Jose airport).
  • Poor search: Searching for an address can give incorrect results. Searching for a fuzzier PoI even more so.
In some ways poor routing algorithms in a v1 product are forgivable. Google has similar problems when its turn-by-turn was first released in 2009. In other ways though this is again something Apple could have nailed from the outset - a quick glance at the number of different turn-by-turn apps (from TomTom to Navteq to Navigon to Sygic) shows that even relatively small companies can get this right. With Apple's resources (and given they were apparently working on iOS maps for many years) you would have thought they could have got this one right.

Similarly getting search right from the off wasn't a given, and this will obviously improve once Apple starts to get a decent amount of user data to work with. But the current algorithm does look particularly crude (whilst searching for Christchurch, Mayfair (London) whilst standing in the Church building I get the Mayfair park in Christchurch New Zealand - surely a simple bit of GPS would have figured out I was searching in London?).

How does Apple fix this?

Just as the problem is really a bunch of different issues, so are the solutions (which I summarised in the table at the start of this post). The good news is there are some easy wins, particularly being smarter with the map data Apple already had. The problem is there are other steps which are harder or will take time. I think its unlikely that Apple Maps will be fit for purpose until well into the New Year. I think some of the bigger problems may not even be fixed until iOS7 hits next year.

  1. Be smarter with the assets you have: For the basical map data issues a lot can be improved by being smarter with what they have. The most obvious fix will be to flip to the TomTom data in countries where is is good (basically the US and Western Europe) and prioritise local data vendors elsewhere.
  2. Get better assets: As I said stuff like textures and PoIs should have been better on day 1. First thing they need to do is get their chequebook out and secure some higher quality imagery. Similarly with Points of Interest - its clear that Acxiom and Yelp are not enough.
  3. Wait the wisdom of crowds: Another obvious route to improvement is to wait for the crowd-sourced data to come in. This will help both underlying data quality and search quality. Apple should act to accelerate this process, for example by making the report-a-problem button a bit more prominent (maybe rename "suggest an improvement") or give it a cooler rebranding (like TomTom MapShare). The problem is this takes time to kick in (although hopefully small business owners will have the sense to register their corner shops ASAP), and Apple will need at least some manual vetting of the data before it goes live.
  4. Build bridges: What people often don't realise is that the hardest thing about building a digital map isn't driving all the roads. In fact most digital mapping wasn't originally done with cars (far too expensive) - rather it was done by building relationships with thousands, utility companies, transportation companies and government agencies to get their data. Apple has tried to shortcut this process by going to the existing mapping providers but either the data is not up to scratch or they have failed to integrate it properly. Therefore a longer-term solution will be to go straight to the source. This doesn't mean they have to build their own map from scratch - but rather they need to build those crucial partnerships so they can get the data that matters from all the other stakeholders to build on what they have already.
  5. Acquire more mapping assets: Apple need to get better and it needs to get better fast. Now M&A is not (as investment bankers would suggest) the solution to every problem, but it could help Apple here. Jason Perlow at CNet put up a very good post over the weekend outlining some avenues for investment. For my part I would add that TomTom is an obvious fit because it brings Apple the bridges it so desperately needs, and a top-notch turn-by-turn routing engine into the bargain. Yes they currently licence the data but owning it and having deeper access could make a big different to the quality of Apple's offering.


The ball is now in Google's court

After spending a good few weeks getting their head kicked in over the Samsung, trial, Google would have enjoyed the last week.

For them the near-term issue is over what to do with their iOS Google Maps (they obviously have a working app - at worst, the old iOS5 one). To keep it off the iPhone would be fun, but long-term stupid as its clearly in their interests to have the iOS user base going past their web properties and not Apple. My hunch is they let iOS6 users stew for a couple more weeks to appreciate how bad the current version of Maps is, and then release it before Apple has had a chance to make major improvements. That way they get the best of all worlds - they keep the installed base, and leave users with a profoundly negative impression of the Apple's app.

The longer term issue is how to keep iOS users installing Google Apps and using Google Services. I suspect this means they have to start bringing at least some of the features Android users have enjoyed (turn-by-turn, vector-based maps, offline caching) to their iOS app. By finally opening up Google's iOS map to competition, it will force it to improve.

So ironically Apple's failure with Maps may actually help iOS users - by giving them a better Google app.

3 comments:

  1. Fantastically detailed and useful summary of how they messed this up so badly

    ReplyDelete
  2. Why thank you.

    Bottom line - long term this is a good move for Apple and a headache for Google. Short term this is a headache for Apple and a good move for Google.

    Although Apple is trying to fix stuff, I think most comentators underestimate how long it will take to fix some of the deeper engine issues.

    They will get there eventually, but if they want to kick-start the process the obvious thing to do is acquire TomTom. Move fast and they can get it wrapped by Christmas.

    ReplyDelete
  3. Apple has removed the "Google Maps" application and launched their own Apple Maps, which is of no use. My own experience saying we are now getting directions that are more worse. As for the Maps, I am sure a stand alone app by Google will soon come to Apple store.

    ReplyDelete