I wrote the other week about ‘the implications of free‘: how the widespread availability of high-quality technology changed the rules when it comes to project management. Another example struck me today, around COI’s ongoing consultation on improving government websites.
There’s a lengthy section on measuring website usage, with detailed proposals around the new requirement for website auditing, kicking in imminently with the aim of ensuring that ‘the rules for measuring the number of Unique User/Browsers, Page Impressions, Visits and Visit Duration have been implemented correctly’. Government websites’ data will be audited twice a year, at a minimum cost of ยฃ1,740 per audit.
So what’s the alternative in the post-free world? How about a centrally managed, mandatory, open-source web analytics package – like Piwik?
- It would place the absolute minimum demand on individual departments: all they’d have to do is include a few lines of javascript at the bottom of their page templates – just like Google Analytics.
- It wouldn’t stop departments running their own analytics packages, if they so desired. Not that many would want or need to.
- Implementation of appropriate standards – statistical, technical, privacy, transparency, etc – could be guaranteed by experts at the centre.
- Lower overall cost: in terms of purchase, ongoing licensing & support, and of course, auditing.
- Freedom to tailor it to particular government requirements, if any.
I must say at this point, I’ve got no direct experience of Piwik myself: but the demo looks great, and it’s used by people I respect – such as Sourceforge and MySociety (eg TheyWorkForYou). Plus, as TWFY demonstrates, you can use Piwik alongside other tracking methods: they seem to have two others on the page too. It’s still at version 0.something, but they’re pledging to hit v1.0 ‘in 2009‘. (Actually, can any of the MySociety gang share their experiences?)
Instead, where will the COI guidance leave us? Website owners will face a financial penalty (admittedly a relatively modest one) if they aren’t using a 2-star rated ABCe Associate Subscriber. And how many of these ‘recommended’ analytics tools are open source, do you think?
Perhaps COI might want to take another look at the Open Source Strategy, (re)published just a month ago: for example, the part where Tom Watson says in his foreword:
We need to increase the pace. We want to ensure that we continue to use the best possible solutions for public services at the best value for money; and that we pay a fair price for what we have to buy. We want to share and re-use what the taxpayer has already purchased across the public sector โ not just to avoid paying twice, but to reduce risks and to drive common, joined up solutions to the common needs of government. We want to encourage innovation and innovators – inside Government by encouraging open source thinking, and outside Government by helping to develop a vibrant market. We want to give leadership to the IT industry and to the wider economy to benefit from the information we generate and the software we develop in Government.
I’d be grateful if COI would consider this as Puffbox Ltd’s contribution to the consultation exercise. Thank you.
Responses
“I’d be grateful if COI would consider this as Puffbox Ltd’s contribution to the consultation exercise. Thank you.”
One of the attractive features of consultation documents that are republished on WriteToReply are the paragraph and section level URIs that allow “remote comments” to associated with a particular part of a document. Using pingbacks/trackbacks, consultation organisers can keep track of who has been commenting elsewhere about a particular section of their document, if those commenters link back to the appropriate section of the report.
Seeing your string “..consider this as Puffbox Ltd’s contribution to the consultation exercise. …” it reminded me of the mechanism used by services such as Technorati and feedburner in which a publisher proves to the service that they own a particular website by inserting a unique identifying string/key provided by the service into a web page. The service detects the key in a page coming from a domain they expect, and so recognise the individual who obtained the key for that domain as the publisher of that site.
So… I wonder if there is an opportunity here for a service in which someone wanting to respond to a consultation could just include a particular string that identifies a web page as being a contribution to a particular consultation exercise. A web search for that string would then pull back all these contributions.
Of course, it might be that people running consultation exercises run web searches on key terms, and so capture informal responses in this way. But by adding a small cost/degree of friction “you must use this exact phrase for your response to count” then it’s possible to start sorting deliberate, as opposed to by chance, public responses to a document.
This approach is still quite weak, though, and doesn’t require much effort on part of the respondent to post a response.
So how about this: if the consultation wanted to run a more formal exercise, it could even offer a ‘click use’ [ http://www.opsi.gov.uk/click-use/ ]/”claim this response” service where a respondent embeds a unique string in the page that can be discovered by a web search engine and used to automatically pull in responses to a consultation by virtue of that string appearing in a page on a domain associated with that string. The pages with the appropriate string can then be treated as formal contributions to the consultation with no further effort required by the respondent?
Hmm, I’m not so sure. There’s quite widespread awareness of Google Analytics (based on Urchin) and plenty of government users of it, me included. I suspect many of those who opt for other packages are often doing so because they need more granular metadata-based reporting or a server-side component. Piwiki – whilst I agree it looks great – wouldn’t address their needs, I suspect.
But the broader point isn’t one about availability or cost of tools, but about using data to evaluate effectiveness. What the guidance needs to do – and does to an extent, I think – is make the case for a data-driven, customer-focussed approach to delivering services and information online. That will vary by service and organisation, so the best approach would seem to be to have a suite of metrics and tools recommended, rather than prescribed.
Absolutely: the whole point should be about evaluating effectiveness, with traffic data being part – but only part – of that.
But I’m not convinced that many departments are consciously choosing other packages. Some have just been using the same tool for years, and haven’t yet been shown a convincing reason to change. Some probably have bigger issues on their to-do lists. And some simply aren’t measuring traffic systematically… or at all. For example: Transport admitted they couldn’t offer a complete picture in a PQ answer last week.
I’m not saying Piwik is the best solution – maybe it is, maybe it isn’t. But I’m asking COI to think creatively around the issue of variable availability of data, and inconsistent measurement methods. A Piwik installation, plus a couple of lines of JavaScript on every HMG website would immediately tackle both those key issues.
Points I made here: http://paulcanning.blogspot.com/2009/03/issues-with-better-connecteds-webstats.html
ยท having more than one package is good practice
ยท get expertise in for specific tasks (as with usability)
ยท but also people need training in basics
ยท sharing apples/apples data with others can be useful