Number crunching

I've been contributing in a small way to OSM for a while now, and I am always impressed by the amount of information that keeps being added to the map. Although there is still work to do on the UK transport network, I think most would agree that it is now well above the level of coverage needed to provide a valuable platform for all kinds of different applications.

What I didn't have a good feel for is how well other aspects of the map are covered. I like to measure stuff, so in the down-time between Christmas and New Year, I've been playing around estimating how many examples of different features we should expect to find in the database, and comparing that estimate to how many are actually recorded.

This is inevitably a bit rough and ready. I needed to estimate how many examples of a feature I should expect to find. Then I needed to figure out a simple way of measuring how many are already in the database; with a reasonable level of confidence that both figures are counting the same things. This doesn't always work. For various reasons it turns out that it's not straightforward to measure things like the number of public telephones, airports, windmills, car parks, cemeteries, and sports grounds. I've included some figures on facilities like restaurants, and hotels, but I suspect that the definitions used in different statistics might not be a very good match to the definitions used by contributors. With more care I might be able to improve these in future.

Still, it's a starting point, and with all sorts of caveats, I think I can identify some features that users of the data can generally expect to find already exist in the database:

Feature	Found	Estimate	Completeness
Bus stations	498	427	117%
Theatres	809	742	109%
Marinas	458	436	105%
Petrol stations	6,451	6,301	102%
Police stations	1,985	2,036	97%
Toilets	4,441	4,714	94%
Fire stations	1,500	1,642	91%

I think I can also identify some types of features where coverage is less complete, but where any particular example is still more likely to be found in the database than not. Some of these are already the target of concentrated activity to improve coverage. Perhaps others are the areas where some more careful analysis would be most useful.

Feature	Found	Estimate	Completeness
Museums	1,505	1,766	85%
Supermarkets	6,681	7,970	83%
Libraries	2,462	3,206	77%
Hotels	5,762	7,561	76%
Cinemas	578	763	76%
Schools	22,969	33,121	69%
Pubs	32,472	49,303	66%
Cafes	7,481	12,140	62%
Breweries	110	180	61%
A and E (England)	91	150	61%
Allotments	3,936	7,286	54%
Post Offices	7,372	14,000	53%
Cycle shops	1,253	2,500	50%

The list of features that I explored is deliberately pretty arbitrary. I covered some that I happen to be interested in; some where I had an estimate to hand, or could easily uncover one; some drawn from OSM project of the month / week activity, and others because I thought they might be of interest to contributors, or of value to map users. The following list shows some where it looks as though contributors will find it fairly easy to find additional examples to add to the map.

Feature	Found	Estimate	Completeness
Pret a manger stores	115	235	49%
Starbucks stores	330	717	46%
Restaurants	10,043	25,226	40%
Ice rinks	13	33	39%
Night clubs	548	1,507	36%
Hostels	454	1,260	36%
Golf courses	709	2,002	35%
Letter boxes	39,329	116,092	34%
Swimming pools	217	641	34%
Anglican churches	4,880	15,976	31%
Veterinary clinics	363	1,271	29%
Garden centres	733	2,621	28%
Community pharmacies	3,732	13,425	28%
Lifeboat stations	31	135	23%
Casinos	37	180	21%
GP surgeries	2,036	10,352	20%
Bowling alleys	103	528	20%
Convenience stores	7,972	48,289	17%
Sewage works	808	5,412	15%
Branches of Greggs	193	1,526	13%
Fish and chip shops	1,334	11,000	12%
National Trust properties	278	2,477	11%
English Heritage properties	64	584	11%
Shops of all types	47,797	458,275	10%
Piers	3	33	9%
Hairdressers	2,729	35,704	8%
Bookmakers	402	5,029	8%
Dentists	722	10,927	7%
Charity shops	559	9,000	6%
Newsagents	801	16,500	5%

Unsurprisingly, my impression is that coverage is generally better for the bigger features that are more obvious (e.g. schools), and for those which particularly interest the OSM community (More pubs than post offices. More bicycle shops than golf courses. Surprisingly few fish and chip shops, though there are some tagging variants that would slightly boost coverage of these if I collected them more systematically). Some of the services that potential users might expect to find are not so well covered (e.g. GPs and dentists).

Even if they are rough  and ready, I think it has been useful to put numbers on some of this stuff. Partly because it helps to measure progress, partly to help think laterally about priorities, and partly as a sanity check on the amount of detail that is both practical and important. And not least because it already demonstrates an impressive level of coverage. However, there is no shortage of opportunity to record the existence of more interesting and useful features in 2012. You can bet on seeing more casinos and bookmakers; while other contributors get their teeth into the missing dentists.

I'll probably come back to this again. It's a bit of an iterative process, and if anyone wants to suggest a better estimate, a source of suitable data, or other features that I should try to measure, then please comment below.