Social media Web metrics

What are your Facebook fans also fans of?

Museums have been pretty good at setting up camp in Facebook. Most have fan pages and groups (and there are many ongoing discussions as to whether groups or pages are ‘better’).

But what matters about all of these activities is whether they are reaching and engaging people who otherwise wouldn’t be. Or are you engaging with the same people in new ways?

Pete Warden’s work has been getting a lot of coverage recently – especially now that he is on the cusp of releasing a huge mountain of Facebook data for academic analysis – and his Fan Page Analytics tool is quite a useful way of quickly seeing how diverse your fans really are. It uses anonymised preference data from 100 million public Facebook profiles – which will skew towards the ‘average Facebook user’ – probably excluding tech-savvy privacy-aware readers of this blog!

Whilst it is early days there’s a lot of promise in tools that look at fan data in these ways.

If you can identify similarities between the fan membership of your own institution and those of others you can start to think of new partnerships and collaborative opportunities.

The Powerhouse Museum has several fan pages so let’s see what the overlap is between the fans of say, the Powerhouse Museum, and one of the festivals that we run – Sydney Design.

Here is the data on the fans of the Powerhouse Museum and the fans of Sydney Design.

We can see that fans of the Powerhouse are more likely to also be fans of other museums but that Sydney Design fans are a more diverse bunch in their fandom – magazines, events, festivals. This reveals opportunities and possibilities.

Obviously fans don’t necessarily reciprocate. Whilst Powerhouse Museum fans are somewhat more likely than most to also be fans of the Australian Museum, fans of the Australian Museum are far more likely to be fans of the Powerhouse!

What can you discover?

User behaviour Web metrics

Segmenting your ‘brand traffic’ in your web metrics

Here’s another web analytics 101 post for you.

One of the important segments for your web analytics are those visitors who come to your website intentionally because it is your website – not just because you have content for Texan monarchists on it.

If you are a museum then this segment is the one that is most influenced by your traditional marketing strategies as well as by what exhibitions and events you have on. In terms of Brownbill and Peacock (2007) these are your ‘visitors’ and ‘transactors’.

So how can you set up a filter to segment this group out in your analytics?

First, in setting up any segment you need to think about how this group might come to your site.

In this case there are five probable ways – directly to your URL, via a bookmark, via an online advertisement, via a third party link, or via a search containing your brand name as a keyword.

Next, you need to look at each of these and figure out how to measure them with the tools you have available.

Direct traffic – easy – they will already be a default segment.

Bookmark traffic – these will be captured as direct traffic and you won’t be able to separate them out as a percentage but that’s probably OK. If you are really hardcore then you might want to segment by landing page here so that those who have bookmarked your funny video of ‘cats in the museum’ get excluded (they’re there for the content, not the brand most probably).

Online advertising – easy. Use your campaign identifiers, track them, and pour in that traffic too.

Third party links – now things are getting harder because in our case the vast majority of third party link traffic (like all our site visitation) is content-related not brand-related. This content-related traffic comes from all sorts of websites, blogs, tweets, Facebook and other places that link directly to a piece of content not because of ‘us’ but because it is ‘good’, ‘funny’, etc. If there are some specific URLs you know have been doing work with then add them in too but I’d recommend not including Twitter, Facebook etc as they are often best addressed as a separate segment altogether.

Search – if you have a relatively unique brand then you can segment out the search traffic containing that brand keyword and add that in. If you are really serious about separating the wheat from the chaff then it might be wise to add another level of filtering which only counts search traffic for the keyword resulting in visits that look at more than one page. (The only problem here will be that you will potentially lose traffic that is coming for simple information like your address – especially if you have tailored a mobile version of your site optimised for mobile traffic.) If you have a generic brand like ‘Australian Museum’ then you are going to need to figure out particular combinations of keyword search as well as landing pages.

Then you want to look at the results further segmented by geography – ideally you’ll find that this segment matches the target geography of your advertising campaigns both online and offline.

If you’ve got a growing brand and you’ve done the segmentation correctly you should see an upward trend as well as significant difference in behaviour of this segment and that of ‘all visitors’.

Ideally this segment should make more conversions (whatever they might be), have a far lower bounce rate and, where relevant spend more time on your site and look at more pages. You also want to be checking that this segment goes to the parts of your site that have been specifically set up for them!

Fingers crossed, eh?

In case of the Powerhouse doing a simple 10 minute analysis on this segment revealed that it made up 15% of our traffic for the past 6 months. We have a site heavy on diverse content so this relatively low percentage is not unexpected – if I was a retailer or I had a museum website with predominantly ‘venue information’ then I’d be worried! Visitors in this segment spent 2x as much time on site, looked at 1.6x as many pages on average, and were 30% less likely to bounce. They were also 3x more likely to begin their visit at the home page and also far more interested in the exhibitions and visit information than the collection than other segments of visitor. They were also 50% more likely to be from Sydney. The time on site and page view ratios for that Sydney segment were even higher!

User behaviour User experience Web metrics Young people & museums

“Let’s make more crowns”, or, the danger of not looking closely at your web metrics

Happy new year everyone.

I’ve got a bit of a backlog of posts but there is an ulterior motive for getting this out the door – and, well, it has been more than 18 months since I should have written about this.

Make-a-king's-crown---Play-at-Powerhouse screenshot

Over on our children’s website – Play at Powerhouse – we have a lot of content for children and parents to do at home either before or after they visit the Museum.

The website was launched in April 2007 as a way of segmenting off the ‘family’ audience from our main website and improving the user experience for that important group. Prior to its establishment, parents who just wanted to know what was on for their kids would have to navigate through exhibitions and events to figure out what was appropriate.

When the site was designed the main navigation was split into two halves – two very simple sections covering the practicalities of a museum visit, and two section for online and at home play.

And so in setting targets for the site we kept in mind that ideally we’d have a pool of casual visitors who we’d best serve by providing quick information that better helped them plan their visit to the museum; and a second group who we’d hope to build as ‘regular’ users of the site for craft activities, and to a far lesser extent, a few online games.

(Digressing briefly, we decided not to focus much on making ‘interactive games’ because there were already many established websites – in Australia the work of the national broadcaster the ABC especially – doing that as their main online focus and, frankly, far better than we could ever expect to do both in terms of design and also promotion).

The ‘craft’ section – Make & Do – was seen as a valuable resource that aligned with the Powerhouse’s reputation as a museum of ‘making things’ in a very crowded children’s web space. Importantly, too, we felt that it was good to support parents in giving them activities from the web that purposely meant doing things with paper and scissors, or out in the garden, anywhere away from a screen.

Over the past nearly three years the site has grown (and is on the schedule for a major UI overhaul!). It attracts a significant amount of traffic – peaking around school holidays as would be expected – and the craft activities, especially, are well linked by sites all over the world.

Internally the site has become integrated with the children’s programming as a whole to such an extent that the site’s Online Producer, Kate Lamerton, is moving over to join the unit responsible for general museum children’s programming. (In many ways this decentralising of content production is a sign of the maturity of the online product).

But that’s not the whole story.

As the site has developed we’ve tried to make user-led choices in the development of new content in the craft area. If the web is good at one thing it is data gathering. Very early on it the thinking of the site we felt that it was important to monitor what was popular and then use that as a means of thinking about what other content should be developed for the site.

Just to give you an idea of the resource burden of content generation – a single craft activity might take two to three full time weeks for Kate to conceptualise, prototype, and then create and instruction set for, photograph and upload. Some take considerably more, others, less.

So obviously we’d want to be choosing those craft activities wisely.

Now not every exhibition at the Powerhouse has obvious choices for craft activities for children, so Kate spends a fair bit of time thinking about ‘events’ to tie activities in with – obvious things like Mother’s Day, Father’s Day [1, 2].

And, because we care about web metrics, we are looking at what is popular and thinking about what I called ‘riffing’ on those – generally making ‘complementary’ thematic content.

But web metrics is a little more complex than that as I’ve said many times here (and in my workshops).

Here’s some data from the last two years.

Let’s first take a look at popular sections by page views for all visitors:

Content type % page views
Make & do (craft) 34.63%
What’s on (events & exhibitions) 24.16%
Play & interact (online games) 15.02%
Visiting (basic visiting information) 6.85%
Home page 15.85%

The first impression here is that the site is doing very much as planned.

The craft activities are generating the bulk of content views whilst the what’s on shows that site visitors are also more likely to be predisposed to visiting the physical museum. As expected with search driving most traffic on the web, the home page is less important as a single entity than each of the larger categories.

Let’s drilldown into Make & Do and see what is popular in there – the top ten by page views for all visitors.

Content type % page views
Craft index 24.63%
King’s crown 15.95%
Outback farm 6.02%
Easter index 3.72%
Knight helmet 3.34%
Princess hat 3.12%
Queen’s crown 3.07%
Masks for the ball 2.53%
Wizard’s hat 2.30%
Witch’s hat 2.14%

Here’s where things get interesting and where the initial thinking became skewed.

The clear leader – by far the most popular bit of craft – is the instructions and templates for making a King’s Crown. And appropriately we went along and made a fair amount of other types of ‘headwear’ – all of which have been popular too.

But are we serving our core audience? Who are these people who are coming to download the instructions for making a King’s Crown?

Let’s re-do those data tables again but this time let’s only look at traffic from Australia and then Sydney.

Content type % page views (all) % page views (Australia) % page views (Sydney)
Make & do (craft) 34.63% 21.06% 15.86%
What’s on (events & exhibitions) 24.16% 30.83% 33.85%
Play & interact (online games) 15.02% 16.12% 16.20%
Visiting (basic visiting information) 6.85% 8.64% 9.38%
Home page 15.85% 19.46% 20.58%

A different story starts to emerge.

Those craft activities are viewed by a far smaller proportion of site visitors the closer we get to our Sydney-based visitors. In fact, for Sydney-based visitors craft activities are even less popular than the online games on a percentage of total page views basis. Not surprisingly, though, by being located in Sydney and thus able to physically visit the Museum, the What’s On section increases in popularity.

Here’s those top ten craft activities again.

Content type % page views (all) % page views (Australia) % page views (Sydney)
Craft index 24.63% 35.29% 41.49%
King’s crown 15.95% 3.31% 1.90%
Outback farm 6.02% 3.97% 4.48%
Easter index 3.72% 7.01% 4.47%
Knight helmet 3.34% n/a n/a
Princess hat 3.12% n/a n/a
Queen’s crown 3.07% n/a n/a
Masks for the ball 2.53% 2.68% 1.96%
Wizard’s hat 2.30% n/a n/a
Witch’s hat 2.14% n/a n/a
Science index n/a 2.83% 3.16%
Easter baskets n/a 2.58% n/a
Speace helmet n/a 2.49% n/a
Mascot colouring in n/a 2.18% 2.68%
Healthy living n/a 2.02% 2.56%
Space index n/a n/a 2.19%

Now this is where it gets really interesting and where the team realised the importance of geographic segmentation. That headwear – the crowns and helmets and hats – wasn’t popular amongst local audiences. In fact, the more local we go the less popular it gets!

So much for putting resources into designing and making instructions for them!

Where was all this traffic for the King’s Crown coming from then?

Here’s the answer.

Country % visits
USA 53.96%
UK 14.27%
Australia 9.14%
Canada 5.42%
Mexico 1.34%

I’m glad our King’s Crown has been popular with Americans – in fact, predominantly Californians and Texans – but without the geographic segmentation being picked up early on in the life of the website we could have continued down that path oblivious to the irrelevance of that content to our local audiences (and the taxpayers who fund the museum).

Have you checked your popular content recently?
Is it really reaching the site visitors you are intending it to?

(Incidentally, if you are intending to attend Museums & the Web 2010 in Denver and wish to do my Web Metrics workshop then book quickly as it is almost full!)

User experience Web metrics

The 2 in 100 who might matter most – your core web audience

As some of you know I’ve been doing a series of deep dive web metrics workshops for various institutions around the world in the last couple of months and one thing I’ve been interested in is estimating the size of a ‘core museum website audience’.

Whilst we all like the big figures of casual visitors we get to our websites many institutions, having flirted with social media, we are beginning to realise that casual visitors, much like casual visitors through the door of a museum, aren’t so useful for building sustained co-creative relationships with.

This ‘core museum website audience’ is the one that is engaged enough with your online activities that they return frequently. The patterns and trends in how they behave in your website is likely to differ significantly from casual visitors, and these trends should be closely analysed for insights into which are your ‘stickiest’ and most ‘interesting’ content areas.

Obviously, in looking at ‘repeat visitation’, though, it is critical to exclude all internal traffic. (I’m always shocked at how many institutions neglect, often through oversight, to stop their web analytics tools from reporting internal traffic!)

If we are serious about ‘engagement’ then our websites need to be actively growing repeat visitation as a proportion of the total.

So, how are we at the Powerhouse doing?

Looking at the Powerhouse Museum traffic for the last 4 quarters (Q4 2008 to Q3 2009) I’ve seen a sizeable number of repeat visitors to our website. Like most websites the vast majority of our traffic is new visitors (80.41%), but I’m pleased to find some interesting figures in our repeat visitors – the other ~20%.

Over the last 8 quarters repeat online visitation noticeably different patterns emerge around our in-gallery exhibitions and around our online-only content.

The ‘2 or more visits in a quarter‘ segment fluctuates most with the blockbuster exhibitions (Diana and Star Wars) showing the impact of return visitors booking online tickets and checking public event information. Here we see a rise from 13.63% in Q4 2007 to a high of 21.45% in Q1 2009 (Star Wars) before dropping again to 18.54%.

The ‘5 or more visits in a quarter‘ segment has grown steadily from 2.11% in Q4 2007 to a high of 5.22% in Q2 2009 and now rests at 4.78% in Q3 2009. This segment contains semi-regular blog visitors and those engaging with our collection online for research and study, as well as some of our high school curriculum focussed content.

The ‘10 or more visits in a quarter‘ segment has grown consistently, unaffected by the seasonal blockbusters, from 0.79% in Q4 2007 to 2.10% of traffic in Q3 2009. This traffic is our most highly engaged – again predominantly around our most consistent blogs (Fresh & New, Photo of the Day, Object of the Week), certain areas of our collection, and very specific curriculum content.

This 2.10% is one that needs a lot more analysis as does the ‘5 or more’ category. How do they arrive at our site? What are they looking for? What do they spend most time looking at?

Just for the record, as I’m using Google Analytics this data excludes RSS subscription-based traffic (critical for blogs), and does contain a low level of error – those who actively clear cookies (who may not be well represented in a core museum audience – but would be on, say, Slashdot). Of course, this data is far more reliable that log-based analytics.

I’m digging much deeper into this for an upcoming paper at Museums and the Web 2010 in Denver and of course my metrics workshop there too.

I’d welcome others’ opinions on this sort of audience segmentation.

Web metrics

How much is your website worth?

I’ve noticed that I’ve been tweeting a lot of links rather than blogging them as I used to. And from time to time there are some links that need to be blogged to get to those who miss the tweets or don’t follow.

Here’s one from the Web & Information Team at Lincolnshire County Council in the UK titled ‘Let’s Turn Off The Web‘.

In order to try to calculate how much the local council website is worth, they turn the question around and ask how much it would cost to provide the same services and level of interaction with citizens if they didn’t have a website.

I like this way of thinking as it provides a way of demonstrating the value of your online services to those who see them only as a ‘cost’. (Your organisation probably already thinks in this way when it is trying to calculate the value of a marketing and PR but web units rarely do.)

So discussions of cost per user, as in a recent Freedom of Information Request to many councils, missed the point. It’s not just about cost per user. It’s about value to the user and savings to the council.

If we turned off our web services:

177,000 visitors per month (May 2009 figures) to our web site would find no web site.

If only 10% of these visitors were to contact us by phone – say 17,000 – then we would incur an extra cost of approx £51,000 per month.* Based on Socitm’s costs of phone contact

Obviously a whole lot of things couldn’t be done at all, but I was particularly drawn to these figures quoted by Lincolnshire Council from work by SOCITM called Channel Value Benchmarking:

*The costs of customer contact are…

Face to face £6.56.
Phone £3.22
Web 27p.
(These figures provided by Socitm 2008.)

Suddenly your web unit is looking pretty good value for money.

Social media Web metrics

Virtuous circle – from visitor to speaker

This short post is for everyone who naively asks about the “ROI of social media” and whether “websites can be proven to result in museum visitation”.

Two years ago Bob Meade wasn’t a regular visitor to the Museum (despite being directly in one of our “target demographics”) let alone a user of our website.

Then we released a bunch of photographs to the Commons on Flickr. These peaked Bob’s interest and reminded him that the Museum existed in his very own home town. (You can read more about that in an interview with Bob from last year – part one, part two.)

Now he’s speaking at one of our weekend talks!

Bob is blogging the prospective content (and museum favourites) of his talk over at his own blog.

It is important to understand that this wasn’t the result of a (social media) “marketing strategy” – it was the result of making valuable museum content broadly available and then engaging our communities in honest, personal conversations.

If you are in Sydney, then come along and hear him speak on September 6.

Tools Web metrics

ROI Revolution’s Google Analytics Report Enhancer

Anyone who attended my double web analytics workshops today at the Transforming Cultural and Scientific Communication conference in Melbourne today saw this lovely little Greasemonkey script in action.

And I thought I better link it for everyone who is not already using this to install.

What GARE does, amongst other things is go some way towards addressing the ‘time on site’ problem that is inherent in most if not all web analytics packages. In short this problem is that single page visits to a website are counted as having zero time spent on them and count this zero figure when creating the ‘average time on site figure. Similarly the time spent on the final page of a visit is left at zero. Blogs are especially susceptible to low time on site figures as most readers visit only one, albeit long, page before leaving.

With GARE installed you are presented with the standard ‘average tiem on site’ as well as a ‘true time on site’ which removes these single page visits from the average calculation. GARE also adds a number of other nifty user interface fixes to make your use of Google Analytics even better.

(My longer paper on web metrics from last year is available at Archimuse and the next web metrics for cultural institutions workshop happens in Indianapolis at MW09 – or on request of course!)

open content Web metrics

Attempts at quantifying social behaviour in the Commons

Over at the fantastic Indicommons blog there has been a flurry of activity around generating data from the various collections in the Commons on Flickr.

Patrick Peccatte initially posted on his blog a set of figures extracted using the Flickr API across the institutions in the Commons. Patrick has reworked these figures a little and they’ve been re-blogged on Indicommons.

The Powerhouse Museum figures work out like this – (as on Feb 7/8)

Powerhouse Museum, Sydney, Australia

Launched on 7 April 2008, currently has 1,101 photos in 27 sets.
1,464 comments,for an average of 1.33 per photo. Max = 97
4,619 tags, for an average of 4.20 per photo. Max = 34
305 notes, for an average of 0.28 per photo. Max = 19
Images with no social behaviour (identified in a separate post) – 336 out of 1,101 (30%)

(source: Patrick Peccatte)

Incidentally these images have been viewed just over 600,000 times at the time Patrick generated the data – which gives some indication of the participation rate (0.2% comment/view rate).

Now as with any quantitative measures these figures have problems. For some institutions the way the Flickr API extracts and reports data has been an issue. But for us these figures are useful given the very Australian and Sydney-centric content of images we’ve been uploading.

Anecdotally we’ve seen a huge increase in viewing as our relatively geographically-specific images have been exposed more widely by Indicommons and others.

Some questions worth exploring further –

– Who is doing the commenting, tagging and note additions?
– Are the repeat viewers?
– How diverse are they? Is it a lot of people doing a little, or, a few people doing a lot?
– Do those who ‘participate’ become ‘contacts’ (do they want to stay notified of future uploads?)

Digitisation open content Web metrics

Library of Congress report on their participation in the Commons on Flickr

Michelle Springer, Beth Dulabahn, Phil Michel, Barbara Natanson, David Reser, David Woodward, and Helena Zinkham over at the Library of Congress have (publicly) released a very in-depth report on their experiences in the Commons on Flickr over a 10 month period.

Titled “For the Common Good: The Library of Congress Flickr Pilot Project” it explores the impacts of the project on access and awareness, traffic back to the LoC’s own website, and, importantly, what they have learned about how collections might operate in the broader social web. Given that their pilot was born of a need to explore the opportunities and challenges of the social web, their findings are important reading for every institution that is dipping their toes in the water.

The Flickr project increases awareness of the Library and its collections; sparks creative interaction with collections; provides LC staff with experience with social tagging and Web 2.0 community input; and provides leadership to cultural heritage and government communities.

I am impressed by the depth of the report and the recommendations. Critically they have identified the resourcing issues around ‘getting the most out of it’ and broken these down as a series of options (see page 34).

Even to maintain their current involvement in the project, they have identified a need to increase resourcing. They also identify that ‘just as is’ is no longer enough.

(2) Continue “as is” – add 50 photos/week and moderate account.

Pro: Modest expense to expand to 1.5 FTE from current 1 FTE (shared by OSI
and LS among 20 staff). Additional .5 FTE needed to keep up with the
amount of user-generated content on a growing account—both in
moderation and in changes to the catalog records (both in Flickr and PPOC).

Con: Loss of opportunity to engage even more people with Library’s visual
collections. Risk of losing attention from a Web 2.0 community that expects new and different content and interaction as often as possible.

Download and read the full report (PDF).

MW2009 Web metrics

Better web metrics for museums – a MW09 workshop, April 2009

The Museums and the Web 2009 programme is now out and registration has started. This year the action takes place in Indianapolis and many of us faraway people are looking forward to checking out the IMA.

If you attended MW last year or the recent National Digital Forum in NZ, or maybe your organisation has had one of my private workshop sessions, you might have heard my rant about the dire problems with how museums ‘measure’ the success or otherwise of their websites and online projects.

My paper on the subject from last year’s MW still stands but now I’ve fleshed the content out to a half day workshop.

This year’s workshop in Indianapolis is now taking bookings and is limited in capacity (unlike last year) and we’re going to be doing a lot more digging into participants’ own sites and I’m hoping everyone who attends will share a month’s worth of data for comparison and analysis purposes.

I’m going to be building this into a solid foundational workshop for basic web analytics as well as a specialised look at the sort of metrics museums, libraries, archives and government web projects need to be engaging with.

If this sounds like it is of interest to you and you happen to be coming to MW09, then register and book a place.