Disruptive Proactivity

The SOCITM studies usually cited are a cherry picking of those who care about the answers because they look good, based on the selective users who do reply.

But we know something else about Local Government websites: they’re supposed to do things, not talk about things, and so we can look at service not fluff. Local DirectGov is a database of services, and is supposed to be right…

Over 5% of Local DirectGov links are clearly broken.
Over 6% are links to the same content as something in a different category. In total, ~10% of links are identifiably broken.

So here’s a breakdown of which Local Authorities might care least:

 top London Borough of Brent (some false reports, but lots of broken stuff)
     Borough of Poole
     St Helens Metropolitan Borough Council
     Wiltshire Council (Unitary)
     Solihull Metropolitan Borough Council
     Torbay Council

(But see below for why this is a undercount).

Methodology

When we turned off DirectionlessGov.com, while we knew that @GovUK would be better for national services; there was the big question about Local. DirectGov was of variable quality, there was much less variation in Local DirectGov.

Directionless knew what a postcode was, and if you included one in your search, it did the right thing. Directgov, did something else.

Various incidents and processes kept this issue from being addressed; and like many of the DirectGov problems, it festered on multiple levels which made solving any a pain.

And while this data has been released before (I first FoIed it about 5 years ago, and got a copy last year), what has changed?

When I first requested the data, I was told that what I was asking for didn’t exist. It did, but style of engagement continued which didn’t aid progress.

Then GovUK showed up.

There are lots of criticisms that could be levelled at the data posted, but fundamentally, a substantial chunk of it is wrong. The file originally published by Local DirectGov , the file that comes out of their systems for others to use, included characters that were not only not UTF-8, they weren’t UTF-anything. Odd characters (Windows Pound signs) are understandable in titles, random junk, not so much.

That junk was in links.

Fundamentally, those links could not ever have worked. And at no point did anyone care enough to point this out. I’m sure the people submitting that data didn’t intentionally think “I’m going to put crap in this box” (simply because those characters required yogic-guru-class key combinations), it’s just no one told them it had happened.

The Local DirectGov project has had issues, some of which aren’t at all easy to fix, but the fact that their infrastructure doesn’t check “does this link work” has directly led to a large number of those problems being visible to users. I appreciate that they’re trying to get a bureaucracy to do a thing that’s really hard for a bureaucracy, but that they were using bad tools isn’t an excuse — but it is fixable.

And so we’ve now got a 100,000 line spreadsheet of links, and they don’t know where to start. So I wrote some code (it’s on github).

The output of data is here. It currently only does the advanced checks against a random subset of files for duplicate content (since 404 pages may lie about everything, but 404 pages within the same authority are generally pretty similar). The full check will take about 4 days to run (it could be parallelised). The script as is currently takes about 2 hours.

@GovUK: If you get council to fix 5% of these 10,000 broken links, you’ll start on the way to having a better service for users, by not dropping them onto a page which certainly doesn’t work. How many of the remaining 90,000 pages are also wrong, is a somewhat different problem…

I have a few ideas.

But I also know, that the next time SOCITM publish a report on council websites, that there’s code available to look at how irrelevant their figures are in practice, and it can be done in 2 hours…

The above was done on or around the 14th Feb 2012 – the link state now will be different (hopefully better, probably worse).