GovUK calls off the search

The date for transition from Directgov to GovUK is fast approaching, and we now have a sight of the homepage which will greet visitors on opening day. And as GDS head of design Ben Terrett acknowledges, ‘it’s significantly different from any of the other homepages we’ve released so far’.

Back in April last year, when a small group of people were starting to think about an alpha for GOV.UK, the expression “Google is the homepage” was coined… People often misunderstood this to mean we thought the homepage should look like Google. We compounded this problem by making the homepage look like Google.


A brief review of previous iterations shows just how deeply that thinking went. Note in particular the one with a Google Maps aerial photo being used as a zero-effort equivalent of Google’s ‘doodles’.
This isn’t the only instance of ‘inspiration → implementation’ in the GovUK design, by the way. When Ben was appointed, he wrote on the GDS blog:

In many ways the problem is similar to problem [Jock] Kinnear [sic] and [Margaret] Calvert faced when designing the road signs in the 60′s. Kinnear and Calvert proposed one consistent system. One designed with the clarity of information as it’s [sic] goal. From then on Britain had a solution that became the definitive standard and was copied around the world… Sound familiar? Swap signage systems for websites. Swap vehicle traffic for online traffic. That’s a challenge no designer could resist.

Six months later, which typeface did they choose as the new design’s base? (New) Transport, Margaret Calvert’s digital-friendly update of said 1960s road sign work. Well, I suppose that’s one way to meet said challenge. But I digress.
Instead of trying to emulate Google, they’ve switched to more of a signposting strategy – which looks more like (very) old-school Yahoo. (Or indeed, Directgov.) A bold decision, which almost feels like a backward step… but a decision based on evidence. It all leads to a fascinating conclusion, which Ben describes as follows.

The people who visit the homepage do so because they are lost. They’re not on the right page, and they’re not comfortable using search, so they go to the homepage to try and help them find what they’re looking for.

Or, if I might dare to paraphrase, your own on-site search isn’t worth worrying too much about. If they’re going to be comfortable doing with the concept of searching, they’ll almost certainly have come to you from Google (65% of traffic) anyway. (All the more reason, I’d say, for using Google Custom Search or the paid-for Site Search.)
The move also coincides with the removal of one of my favourite features of previous iterations: search suggestions as you type. When done well, it’s an invaluable navigation tool in itself: and in fact, I’m now finding myself expecting to be offered search suggestions, when I start typing into the Search box of any large-scale site.

But it may not be gone for good:

I look forward to reading that forthcoming blog post. (Update: now published.)

NHS kills Google advertising

I noted back in February that NHS Choices had spent £2.7m in one year on pay-per-click advertising. Well, that’s all changed now: a PQ answer reveals that the Adwords budget has been cut by 100%.

In line with Government policy, NHS Choices no longer has any arrangement, or pays for any search engine activity. No commitments have been made with Google or any other search provider for ‘pay per click’ online marketing since the moratorium on marketing spend was put in place on 24 May 2010. NHS Choices used paid search activity to ensure that it reaches the widest possible audience, and that users can easily find clinically assured health information and access the services they need from Government.

(And sure enough, other top spenders like the Act On CO2 campaign have also scrapped their Adwords activity.)
In my February piece, I looked at two specific search terms – ‘stop smoking’ and ‘chlamydia’. The NHS site is still the top natural result for ‘stop smoking‘… although it comes beneath sponsored links to specific pharma products. The picture for chlamydia isn’t so great: the NHS site comes well down the first page of Google results – beneath the American CDC, interestingly. Time to ramp up the SEO activity.

Constituency maps in under a minute

Opening up geographic data is beginning to bear fruit. MySociety’s Matthew Somerville has just unveiled MaPit, ‘our database and web service that maps postcodes and points to current or past administrative area information and polygons for all the United Kingdom.’ What that means in practice is, postcode lookups and boundary data are now just a URL away.
(Quick update: actually, not for all the United Kingdom as it turns out – the following method doesn’t work for N Ireland. See Matthew’s comment below.)
Here’s a quick example, as much for my own future reference as anyone else’s. Let’s say you wanted to generate a map of a given MP’s constituency – say Lynne Featherstone in Hornsey & Wood Green:

  • You need to find the appropriate reference number for the constituency: either by browsing the list of all constituencies, or searching for places whose names begin with Hornsey. Note – these will produce nasty-looking data files, rather than pretty HTML lists. Hunt through the code, and you’ll find:
    “65883”: {“codes”: {“unit_id”: “25044”}, “name”: “Hornsey and Wood Green”, “country”: “E”, “type_name”: “UK Parliament constituency”, “parent_area”: null, “generation_high”: 13, “generation_low”: 13, “country_name”: “England”, “type”: “WMC”, “id”: 65883}
  • The ‘id’ is the number you need – in this case, 65883. The MySociety API now lets you call the geometry of that area, in – among others – Google Earth’s KML format, using the following URL. (Don’t worry about the ‘4326’ here: it’s a reference to the coordinate system being used, and won’t change in this context.)
    http://mapit.mysociety.org/area/4326/65883.kml
  • Conveniently, Google Maps lets you enter a KML file’s URL as a search query, and it will draw it on a map. Even more conveniently, if you add ‘output=embed’ as a search parameter, it strips away everything but the map itself. So here’s an embedded map of Lynne’s constituency, pulled into an <iframe>. Look at the source code, to see how easy it is.

Boundary data generated by MaPit.mysociety.org which contains Ordnance Survey data © Crown copyright and database right 2010; Royal Mail data © Royal Mail copyright and database right 2010 (Code-Point Open); National Statistics data © Crown copyright and database right 2010 (NSPD Open).
And thankfully, it bears a close resemblance to this map on Lynne’s own website, which took me considerably longer to churn out.

DH reveals £2.7m Adwords spend

A parliamentary question has revealed that, in the year to the end of January 2010, the Department of Health spent £2.72 million on Google Adwords pay-per-click keyword advertising. A big number, but a fair one?
With Google’s Adwords advertising, you only pay on results. An advert is displayed at the top, or down the side of a set of Google search results for a given keyword or phrase; or, optionally, on third-party web pages where Google’s matching technology decides your keyword is relevant. Plus, Google’s technology allows for geo-targeting, so in the case of DH, they can specify ‘UK only’. (Of course, DH is only responsible for England, but that’s for another day.) So in theory at least – we know that people were looking for something health-related; they saw an advert from DH/the NHS; they decided it was of interest; they clicked on it, and were taken to a DH website. Job done.
A couple of factors to bear in mind. There has been a trend towards campaign calls-to-action based on search terms: ‘search online for X’ – and in the free-for-all of Google search ranking, the only way to truly guarantee visibility at the top of the page is to pay Google for the privilege. If it works as a call-to-action, and you have to pay for it, then so be it. And it’s a competitive business – where government finds itself going head-to-head against pharmaceutical companies. Top ranking can cost a lot of money.
DH is rightly cautious about disclosing too much data, citing competitive confidentiality. Similarly, Google doesn’t tell you as much as you might like about ‘what keyword X will cost you’. But their Traffic Estimator provides some clues.
If we look at ‘chlamydia’ for example: Google’s tool suggests that a bid of 30-43p per click will buy you a position in the top 3 adverts for the term, leading to 62-86 clicks per day. That’s something like £27.50 per day, or £10,000 a year. Now of course, Google’s screen layout means there’s a significant premium to being the no1 ‘sponsored link’ – and you might well consider it worth bidding high to guarantee top spot, particularly in the case of chlamydia, where the NHS site is the no2 ‘natural’ result.
For the term ‘stop smoking’, Google’s tool suggests a bid of £1.66 to £2.50 per click, to secure a top 3 slot resulting in 49-68 clicks per day. So for a similar volume of traffic, that’d be well over £100 per day – and an annual cost in the region of £42,500. Why so much more expensive? – because the NHS is in direct competition with bids from anti-smoking drugs, devices and consultants. One wonders what premium they’re paying to guarantee no1 position there – but the Google tool suggests a maximum bid-per-click of over £6, taking us well into six figure annual budgets.
Time and again, when you search for something health-related on Google, there’s an NHS sponsored link at the top of the page. They, or rather we, are paying good money for this. You’ve got to assume someone’s looking at the numbers, and deciding it’s worthwhile. Just because it’s a big number, doesn’t necessarily mean it isn’t cost-effective. It may surprise some people for me to say this – but some things on the web are worth paying for.

Wiped off the map

romseyroad
I’ve come across a rather curious anomaly in Google’s new Street View. Southampton is one of the cities covered in the initial UK rollout… or rather, most of Southampton. You’ll note the bizarrely unavailable stretch of Romsey Road… and the odd interruption of Wimpson Lane. Here’s a link to it in Street View, to see for yourself.
So… who do you think operates from this unmapped building? If you need a clue, read this piece I wrote late last year: ‘Ordnance Survey ban Google Maps‘.
Of course, I’m sure there’s a perfectly good reason.

Guardian Data Store: threat to ONS or its saviour?

When I first saw reports of the Guardian’s new Data Store ‘open platform’, my heart sank. In a former life, I ran the web operation at the Office for National Statistics; I resigned in June 2004, when frustration started to turn to anger. I’ve still got a copy of my resignation letter, in which I wrote:

I have always maintained that the agenda of openness which I espoused is not a choice; it is a reality forced upon us by the modern communication environment. The general public’s expectations have moved on dramatically in the last decade [1995-2004]. Sadly, this [realisation] has not been shared by other parts of the Office on whom my work or resourcing have been dependent.

I warned them that someone would come along, do a better job than they were doing, and supplant them as the ‘primary source’. Once that happened, the statistical sanctity so jealously guarded by the priesthood of statisticians could very easily be compromised. In effect, to preserve the status quo, things had to change. (The message went unheeded, by the way: the six-month ‘stopgap’ site I introduced is soon to celebrate its seventh birthday.)
So today, the Guardian unveiled their Data Store. Editor-in-chief Alan Rusbridger is absolutely clear about the service’s purpose:

Publishing data has got easier [since 1821] but it brings with it confusion and inaccessibility. How do you know where to look, what is credible or up to date? Official documents are often published as uneditable pdf files – useless for analysis except in ways already done by the organisation itself.

Just to be clear, ONS: that’s you he’s talking about. It’s expressed even more starkly in an accompanying blog post by Simon Rogers, subtitled: ‘Looking for stats and facts? This is now the place to come.‘ A quick look down the data on offer reveals a high proportion, a majority perhaps, to be ONS or other HMG data. Their tanks are on your lawn, guys.
Now I’m not for one minute suggesting the Guardian would do anything malicious. I’m simply warning of the uncomfortable position where an outside entity – indeed, in this case, one with an explicit political slant – becomes the gatekeeper to (supposedly) pure statistical data. Can we rely on them to be as comprehensive, as conscientious, as religious in their devotion to updates, corrections and revisions? No, admits Simon Rogers: ‘it is not comprehensive… this is selective’.
So is this the Doomsday Scenario I predicted? Not quite, not yet. How exactly are the Guardian serving up the data?

We’ve chosen Google Spreadsheets to host these data sets as the service offers some nice features for people who want to take the data and use it elsewhere [in] a selection of output formats including Excel, HTML, Acrobat PDF, text and csv. A key reason for choosing Google Spreadsheets to publish our data is not just the user-friendly sharing functionality but also the programmatic access it offers directly into the data. There is an API that will enable developers to build applications using the data, too.

You read that right: the actual mechanics are as basic as: uploading/copying existing Excel spreadsheets, converting/pasting them into Google Docs spreadsheets (price: £0 for 5000 reasonably-sized files), and letting the Google functionality do the rest. By way of example, data on England’s population by sex and race. The Guardian offers this Google Spreadsheet. Now download this Excel file from the ONS website, and look at the sheet labelled ‘Datasheet’. Actually, let me save you the bother: they’re identical.
Cabinet Office minister Tom Watson writes on his blog: ‘Governments should be doing this. Governments will be doing it. The question is how long will it take us to catch up.’ The answer is, the few seconds it takes to sign up for a Google account, and maybe an hour of copy-and-paste. So, tomorrow lunchtime, then.
This afternoon, I thought this was a disaster for ONS’s future. I’ve changed my mind. The Guardian’s move sets a precedent, and lays down a direct, unavoidable challenge. It could actually be ONS’s salvation.

Putting Google geo-location to the Twitter test

Google’s javascript API has an exciting, and somewhat underreported little feature built in: each time a call is initiated, it attempts to establish where the browser is physically located – and reports back a town, ‘region’ (county) and country. I was wondering if it was accurate enough to be used to ‘personalise’ a website automatically: so I ran a quick experiment among my Twitter following.
I set up a quick test page on puffbox.com, which included a call to the Google API, and asked people to leave a comment as to whether or not the response was accurate. Within an hour I’d had 30 responses, from all around the UK.
The results revealed that the function is sometimes bang-on, sometimes blocked, sometimes curious, and sometimes plain wrong… occasionally by hundreds of miles. I can forgive the occasional placing of towns in the wrong county; but several people in the north of England, using the same ISP also located up north, were getting responses of ‘London’. So my conclusion, disappointingly, is that it’s not really good enough to make meaningful use of.
A wasted effort? Hardly. It actually saves me the effort of building something reliant on the geo function, only to discover it’s useless for large numbers of people. And it’s a nice case study for the value of Twitter: a crowd of good folk and true, located all over the country, from whom I could ask a 5-second favour… with a good expectation of getting responses. Thanks, team.

Two weeks with my T-Mobile G1

The release of the first ‘Googlephone’ running the open source Android couldn’t have come at a better time for me.
My Nokia E65 had served me well, but was getting a bit temperamental. I’d already handed all my email over to Google, and was keen to do likewise with my calendar needs (bye bye Outlook!). Plus, Google tends to just get it right, where Apple’s track record is a bit patchy in my experience. Of course I was going to get a Googlephone; and with my contract’s minimum period having just expired, why wait?
Let me straight away: I like it a lot. A heck of a lot. But it’s not without its faults, some of them significant.
The screen is beautiful, the interface is beautiful, the unit itself is pretty (rather than iconic). As you’d expect with Google, everything integrates neatly. When you power up for the first time, you’re asked for your Google Account details – and yes, thankfully, that includes ‘apps for your domain‘ accounts. Instantly, you’re looking at your Gmail inbox, with all the same contacts, and your Google calendar. Change something on your desktop PC, and it reflects instantly to the phone – and vice versa. That’s enough to make you love the phone itself.
The keyboard is small but usable with practice. The web experience is excellent, with the browser firing up instantly. Google Maps is as good as you’d expect, with or without GPS. Oh yeah, and it’s good at making phonecalls too. Google’s contacts app has come in for criticism, much of it justified; but its integration with Gmail gives you a headstart, importing from Outlook was easy, and courtesy of Gravatar, I’ve managed to add photos for a remarkable proportion of my contacts with zero effort.
But there are faults, some of which will resolve themselves over time. Others won’t.

  • Zoom. For me, the key to the iPhone is the multi-touch operation, which the G1 doesn’t have. So if you want to zoom in on a photo or a web page, you’re pressing zoom in/out buttons. It feels so primitive already.
  • Camera. Three megapixels is respectable, and the quality is fine. But jeez, is it slow to snap… almost unusably so. I’m seriously considering carrying a proper camera again; and that’s really not good enough.
  • Lack of homescreen widgets. If you’ve ever seen Nokia’s Widsets, you’ll recognise the horizontally scrolling desktop on the G1. Some aspects are cool: it scrolls beautifully, you can add shortcuts to individual contacts, and there’s widget functionality. But the selection of widgets is shockingly small: an analogue clock, a picture frame, a Google search box, and that’s it. No calendar view. No to-do list. No weather widget. I’m hoping this is just a temporary weakness, forgiveable since Android’s only been properly public for a matter of days. But still…
  • Lack of ‘send’ functionality. I’m amazed it doesn’t include the ability to send contacts or calendar events via SMS, Bluetooth, email or anything else. Yet.
  • Media playback. There’s a respectable ‘Music’ app, and there’s even some Flight of the Conchords in there to start you off. But if you’re hoping to stream anything other than YouTube, or playback video, you’re going to have to visit the Android Market and hope for the best. For now, you’ll be disappointed.

There’s quite a lot to play with in the Android Market: a selection of weather apps, the barcode-scanning programs are fun, there’s a fantastic Twitter app called Twidroid, and who can resist (proper) PacMan? The big names are starting to appear – among them the Telegraph and MySpace, and others will follow no doubt (including MySociety, I hear?). Again, early days.
Android is unquestionably going places, and there’s an undeniable thrill at being in at the very start, watching it all evolve. But I’d recommend steering clear of the ‘free on 18 month contract’ deals. Although there’s no mention of it in the T-Mobile publicity, my local shop sold it to me for £50 on a 12-month contract; by which time I bet better devices will be available, and no nastiness with unlocking / jailbreaking.

Full launch for Met crime maps

Wednesday saw the formal launch of London’s crime maps, which first appeared in beta only a couple of weeks back. Don’t call it ‘1.0’ though: the source code declares it’s actually ‘beta 1.02’.
As before, it shows areas colour-coded for the rates of ‘burglary, robbery and vehicle crime’, based on comparisons with ‘the average’. Yes, that’s an approach which has its limitations – my favourite being that areas containing police stations tend to rate worst for offences, because of (for example) finding drugs on someone you’ve just taken into the custody suite. But as a first step, it’s surely a good one. There are plenty of legal and logistical issues to overcome before we start putting dots on the map according to offences… and as we all know by now, if you try to sort everything out before going live, you never go live.
No mention of the total cost in the official press release, but I’ve seen a figure of £210,000 reported (eg Daily Mail). Given that it’s a fairly straightforward map mashup, using the standard infoWindows and polygons built into the (free) Google Maps interface, I’d be very interested to see a confirmation and/or breakdown of that. Fair play to Boris and his Conservative administration for getting it out the door early; but the next ‘how much did that government website cost?’ argument could be interesting.
Meanwhile, I see the Foreign Office is doing some map mashing of its own, with a cute (rather than useful) map of travel advice notices for the home nations’ World Cup qualifiers. But crucially, they’ve done it using the free ‘My Maps’ functionality; and it’s offered purely as an external link, not even an embed on the FCO page.

No extensions, no Chrome

It was the usual mix of excitement and fear as I downloaded Google Chrome last night: the former to see what Google would do when it had total control of the browsing experience, the latter in case it rendered any of my designs horribly. To be honest, there wasn’t much to report on either front.
Let’s not get carried away: Chrome is a web browser. I like the minimalist interface (with Vista-esque styling even on XP). The ‘omnibox’ is a great idea, as long as you’re happy with the privacy question. It’s kinda fun – and sometimes a bit frightening – to be able to see how much memory an individual page consumes, particularly if you can recall the days of the ZX81 and its 1K of memory. I’ll take their word on the stability and security features, but (happily) I’ve yet to see any evidence. The automated ‘favourites’ on new tabs is a nice touch, I suppose. But for all this, it’s still just a piece of software which renders HTML, CSS and Javascript into pages.
Chrome’s main selling point seems to be its improved Javascript performance – according to one test, it’s 2.5x faster than Firefox, and a ridiculous 22x faster than IE7. That’s going to matter in the future, allowing Javascript interfaces to become more complex. But I don’t imagine many websites will feel ready to take things to that higher level for quite some time.
Will I be switching? No – for one simple reason, and that’s the surprising lack of plugin capability. The help documentation declares:

Currently, Google Chrome supports the most popular plug-ins necessary to display the Web correctly, including Flash, Acrobat Reader, Java, Windows Media Player, Real Player, QuickTime, and Silverlight.

And that’s it – which is a problem. Firefox is my do-everything window on the world. I get an alert when new email messages are detected by the Gmail plugin I use. My bookmarks are fed into the browser courtesy of the Delicious plugin. Without these and many others, I’m feeling lost in cyberspace.
It’s a curious omission, given that Google’s quite happy for developers to write their own Gadgets for Google Desktop and iGoogle. According to Google blogger Matt Cutts:

I’m sure that extensions/add-ons are something that the Chrome team would like to do down the road, but the Chrome team will be a bit busy for a while, what with the feedback from the launch plus working on Mac and Linux support. I’d suggest that you give Google Chrome a try for a few days to see if enjoy browsing even without extension X. A lot of really cool extension-like behaviors such as resize-able textareas and drag-and-drop file upload are already built into Google Chrome.

So it looks like Chrome won’t be able to give me the online experience I’ve grown to expect, not any time soon. It confirms the theory offered by Shane Richmond yesterday, that Chrome is an attack on Internet Explorer rather than Firefox. It’s for people – the vast majority of people – who don’t know or care about customisation. It will give them a streamlined online experience, with all the plugins they need and no more, within a controlled environment. If you want more, well, you’re probably on Firefox already… and, like me, unlikely to move.
Quick update: it looks like the new Javascript engine planned for Firefox v3.1 (known as TraceMonkey) is even faster than Chrome. But for me, the most interesting comparison in this new data isn’t between Chrome’s engine and Firefox v3.1’s engine… it’s between XP and Vista. Chrome’s V8 is 30% slower on Vista; TraceMonkey is 20% slower on Vista. I’m no expert, but that doesn’t sound good.