This is the tidied-up transcript of a talk I gave a month ago at the Content Strategy Forum 2013 in Helsinki, Finland. It was the third consecutive year I’d spoken at the forum and it was quite possibly the most enjoyable talk I’ve given to date.

Note: Though I’ve included screenshots of the most important slides, you might prefer to read along with the full deck.

The importance of visualisation – mapping the way forward

As our species’ love affair with maps is about as old as civilisation itself, I’d like to begin today’s talk by briefly going back all the way to the beginning of human existence. Long before humans could write we have been making and using maps to make sense of the world around us (Figure 1). Each example offers a snapshot into a different time and culture, as well as a unique insight into the political, cultural, and spiritual forces that drive society.

Figure 1 – Babylonian map of the world. Clay tablet, probably from Sippar, southern Iraq (c.700-500 BC) © Trustees of the British Museum

Every blank surface you can possibly think of – be it of rock, clay, wood, parchment, paper, or tapestry – has been used to plot places and objects according to their relative spacial positions. And it’s this synthesis of science, art, and history that combines to create these beautiful objects. Every map has its own story to tell and harbours secrets it may never fully reveal. A map can delight, surprise, and unsettle in equal measure. And maps can reveal deep truths – not just about where we’ve come from, but about who we are.

Throughout history the map has demonstrated its versatility. They’ve had an administrative use in marking out national boundaries or individual plots of land, a social use in showing who lives where, a military use in depicting the layout of enemy positions, and a political or propaganda use in showing one country or faction over others. But beyond their two-dimensional depictions of a physical world, maps also afford us the freedom to express the cosmos; to make all kinds of ideas about the spatial relationships of multiple components unexpectedly clear. To draw one is an effective way to establish order on an otherwise chaotic environment. To make it navigable. To make it rational.

City maps are a fine example of creating this impression of order. The whole objective behind creating such a map would be to somehow capture, contextualise, and impose order on an environment which is always moving, growing, and changing. An environment which is falling apart and burgeoning at the same time.

As content professionals, we’re often tasked with making sense of complex, unpredictable, and mostly disorderly environments – home to micro communities with their own agendas, rules, and systems of government. Managing content, processes, and people can be a complex endeavour, and managing them in large systems only increases the complexity.

In his famous map of London published in 1747, John Rocque shows off the perfect enlightenment city (Figure 2). It’s clinical, controlled, and beautiful. This is a map of London at its most picturesque. It’s a map that imposes order on a city of unrelenting change, and it gives all the appearance of objective truth. Rocque has deliberately left out the ugly, unsavoury elements of the city.

Figure 2 – John Rocque’s 1746 map of London, which deliberately ignored the city’s unsavoury elements.

To Rocque, it was more important – and lucrative – to present and promote London as the greatest city of its time than it was to present some of its grimier realities. Indeed, mapmakers through the ages have always responded to the mentalities, and met the requirements, of the societies in which they have been created. They quickly learnt that the quality and effectiveness of a map couldn’t simply be judged by its scientific precision, but by its ability to serve its purpose. Without question one of the huge appeals of Rocque’s map was that it successfully imposed order on chaos (Figure 3). This was a visual interpretation of London which offered its inhabitants a sense of promise and a sense of pride as well. To Rocque this map was about imposing clinical precision onto a city which he knew couldn’t literally be accomplished. Why? Because of the human element. And this is the main issue: The thing about cities, like large multi-departmental organisations, is people, and people just make it into a bit of a mad-house.

Figure 3 – One of the huge appeals of Rocque’s map was that it successfully imposed order on chaos.

If we took it upon ourselves to map a complete and literal visual representation of the way content flows within a mid-to-large organisation, we would, in a sense, be mapping a sprawling city. Such a multiplicity of information would need serious reining in if it were not to cancel itself out in a sorry mess. In the end all drawn city maps, however much they distort the truth, are trying to take on the impossible: They are trying to impose two-dimensional order on the chaos that is urban life.

So we’ve learnt that the quality and effectiveness of a map cannot simply be judged by its scientific precision but by its ability to serve its purpose. Aesthetic and design considerations are every bit as important as the mathematical, and often more so. Indeed, maps of this nature have long served as vital political and negotiating tools. In these cases the intended information inherent to the map has to be conveyed in a way that attracts the eye to certain features. Believe it or not but all this can be achieved with even a simple sketch map. Everything that maps are used to represent: placement, proximity, overlap, distance, and direction can consist of only of a few lines and a few letters, with nothing but the essential features remaining (Figure 4). In short: The simpler the better. Maps like this can be used to quickly and clearly communicate our ideas to others. And if sketched in their presence, you can even use them to tell stories.

Figure 4 – Everything that maps are used to represent: placement, proximity, overlap, distance, and direction need only consist of a few lines and a few letters.

Every day we come into contact with various kinds of maps which, on the surface, may look wildly different, but they all share the same basic building blocks (Figure 5). It’s all about picking out the people, places, and things – indeed anything whose relative positions we want to compare – and picturing them in a geographical landscape.

Figure 5 – Different maps, same basic building blocks: Exploded view, Concept map, Star map, and Topological map.

The only real challenge is coming up with a meaningful coordinate system (Figure 6). We’re all well versed in the north-south versus east-west coordinate system, but we can make a map of anything using other pairs of opposites like expensive-cheap versus high-low and fast-slow versus small-large.

Figure 6 – Finding your coordinate system. Anything can be mapped using pairs of opposites.

Once we’ve defined our coordinates we can start to plot our landmarks (Figure 7), beginning with the most prominent feature of our landscape. This could be an object, person, or an idea.

Figure 7 – Plot your landmarks, starting with the most prominent feature.

And then we can move outward (Figure 8), adding more and features and details, illustrating everything from borders and distances to pathways and sets of shared traits. Let’s make a map!

Figure 8 – Move outward by adding more landmarks to you map.

The unfamiliar environment – finding our way

When it comes to getting a grip of that complex trinity of content, processes and people, governance is an important tool (Figure 9). Put simply, governance determines how key decisions are made and who has the authority to make them. Governance defines who is allowed to create, approve, and publish content, and how these decisions are initiated and communicated.

Figure 9 – When tackling that complex trinity of content, processes, and people, content governance is an important tool.

One big content governance challenge is maintaining consistency with messaging, communication, editorial and content standards throughout an organisation. This is made particularly difficult when many contributors have other responsibilities and priorities besides content. Governance is primarily a human management issue, but one useful way to help maintain these standards is through content management system (CMS) workflow. This functionality helps manage the sequence of steps from adding or editing to publishing.

Let’s imagine we’ve been hired by a regional English Council to design a new content workflow for their CMS. This council is one of 27 large two-tier non-metropolitan counties used for the purposes of local government in England. Approximately 15,000 staff across seven departments are responsible for providing services to 700,000 local people including education, social care, transport, and culture and leisure. Unsurprisingly, it’s a varied, siloed environment. Some departments take responsibility of hundreds pages on the public website, others only a few. And in the middle of it all is the relatively small Digital Services department trying to keep everyone’s content balanced and consistent.

If you were about to be airdropped into unfamiliar, and possibly hostile, territory, you’d want to be carrying either a map or pairing up with a native guide. So as an outside consultant arriving fresh into this large, diverse, politically-charged environment, your first instinct in making sense of it should be – besides asking for directions to the nearest coffee machine – to draw a map. In order to work out how decisions about content are made around here we need to figure who best to talk to besides Mandy, our main point of contact from the Digital Services department (Figure 10) – particularly those who might be influential to the way things work without even realising it. And to start this process we first need to map out our landscape.

Figure 10 – To discover how content decisions are made and how they could be improved in our varied, siloed environment, we need to figure out who best to talk to.

Since we’ve just arrived in this new environment let’s start by placing ourselves and the rest of the Digital Services department on the map. Next we want to add in the six other council departments (Figure 11): Adults & Communities, Chief Executive’s, Children and Young People, Community Planning, Corporate Resources, and Environment and Transport.

Figure 11 – To begin mapping our environment, we’ll mark out each of the seven departments.

Now we need to introduce a coordinate system (Figure 12). So let’s plot our departments by their relative sizes from small—large, and we’ll also ask Mandy how quickly each department takes, on average, to move through the gears of creating, approving, and publishing content.

Figure 12 – To devise our coordinate system, we’ll rank each department by their relative sizes and how long their content takes, on average, to go from creation to publication.

Our department’s are plotted, so what can we see? If we’re looking for best performing department we need look no further than Corporate Resources, who, despite their large size, are one of the most efficient at delivering content. On the opposite end of the scale are the Community Planning department, who are small and yet one of the least efficient.

Next we want to understand a little more about the politics of the place. So after surveying various members of the Digital Services department, we can map the pathways between departments that they believe share closer communicative ties (Figure 13).

Figure 13 – Mapping those pathways helps us get a clearer picture of the politics of the place.

So what jumps out now? Our marked pathways tell us that there are two distinct groups, with one markedly quicker at delivering content than the other (Figure 14).

Figure 14 – What can our marked pathways tell us?

Perhaps if we wanted to bridge the gap between these two groups the Digital Services team might consider trying to form closer ties with the similarly-sized Community Planning department (Figure 15). If we could improve the efficiency of this relatively small department it might provide a positive knock-on effect. To me, the Community Planning department looks like the ideal place for a pilot project.

Figure 15 – Where are the opportunities for improvement?

Now we have our map. It’s by no means scientific or pretty but as I mentioned earlier the quality and effectiveness of a map should primarily be judged by its ability to serve its purpose. Our map may only consist of a few lines and letters but we now have a quick and useful overview of the council’s departmental structure. Not forgetting, of course, that we’ve only just walked through the door.

The unfamiliar environment – mapping out a process

Designing an effective CMS workflow requires meeting the needs of the organisation and each of your content players. In the case of our English regional Council scenario, this is our team of authors spread far and wide across the seven departments. To understand how we can improve the author experience we need to ask each of our content stakeholders to describe their current process for planning, creating, and publishing content.

In our conversation with Mandy, Content Officer in the Digital Services department, she kindly talks us through the typical process for publishing a press release on the council website. I’m not an especially quick note-taker so I always ask permission to record the conversation.

I was always taught never to read your slides, but hey-ho.

“On Monday she received a request from John, Highways Manager at the Environment and Transport department, to publish a press release to inform the community of the start of major road improvement work. She first called Helen, Technical Services Manager at the Environment and Transport department, who emailed through the source material and directed her on the technical details that needed to be included.

She continues…

“After completing the draft in the CMS environment, she marked the press release as ‘ready for review’ and emailed Rachael, a Content Manager and colleague in the Digital Services team, and Helen again for their feedback. After one or two line edits, she was ready to mark the press release as ‘ready for publishing’, so she emailed David, a Digital Media Officer to prepare and publish the press release to the test server.

She continues…

“John, Helen and Rachael were all emailed to help her perform a second review of the content, where they duly approved its publishing to the public website. David then got all the glory as usual by hitting the button marked ‘publish’.”

Now, I don’t know about you but I’m lost. The hardest part of these conversations is keeping track of all the different people, roles, and departments as they’re reeled off. When we hear four, five, six names or more it can be difficult to process them; to find form of the relationships between them. To make sense of this complex drama we can make a map that shows who each character is, how many there are, and how directly and indirectly involved they are in the story.

Let’s begin by listing all the characters we heard just then and their roles before sketching their likeness (Figure 16). Don’t worry too much about accurately portraying their looks. You may not have even met them before, but it’s always a good idea to put a face to a name.

Figure 16 – Who are the players in our story?

Next we need to separate them by department and plot each interaction and task in chronological order from the initial request to the press release’s publication (Figure 17).

Figure 17 – Plotting each interaction and task belonging to each character in chronological order (click image to zoom).

This is good. We can clearly see the different roles each character plays and the interactions between them. We can follow this process from the initial request, the research and gathering stage, the creation and subsequent review of the first draft, publishing the draft to the test server, the second round of reviews and approvals, and finally publishing the press release to the web. But we’re still missing features of this story which would further enhance our understanding of this process and how it could be improved. Namely, which interactions and tasks took place in the CMS environment and how long each took to complete.

As Mandy almost practically saw the process from start to finish, she’ll be able to help us fill in these details. Let’s first stretch out the timeline to accurately reflect the varying speeds of each interaction (Figure 18).

Figure 18 – Stretching out the timeline gives us a more accurate measurement of how long each task took to complete (click image to zoom).

What jumps out now? We can see that John and Helen from the Environment & Transport department took longer than Rachael to review the first draft and to review and approve the test server version. Might that just be because they have other departmental responsibilities besides producing, checking, and updating website content? Possibly. Let’s see what happens when we mark out which interactions and tasks took place within the CMS environment (Figure 19).

Figure 19 – Highlighting which interactions and tasks took place within the CMS environment might help us reveal where the bottlenecks are in the process (click image to zoom).

Aha! Interesting! John appears to be carrying out his reviews outside of the CMS. How come? To answer this we need to return to our first map.

We can see the Transport and Environment department are on the slower end of the scale in terms of moving through the gears of creating, approving, and publishing content (Figure 20). After interviewing Helen we learn that the department is comprised of three sub-departments: Transportation, Highways, and Environment.

Figure 20 – Breaking down the Transport and Environment department.

John resides in the Highways sub-department, who we discover has no direct access to the central CMS because of their geographic location elsewhere in the county, and thus is effectively an external contractor (Figure 21). Helen tells us that John has to use email to communicate his requests, edits, and approvals.

Figure 21 – John’s failing sub-department might be down to CMS workarounds adding unnecessary time and cost.

It all makes sense now. The reaction times of the Highways sub-department is adding unnecessary time and cost to this and many other content production processes. It’s compromising the efficiency of the Transport and Environment department and the entire organisation. The intrepid explorer would be wise to investigate whether other departments on the slower end of the scale may be affected when broken down like this. There could be incredible cost-saving potential across the board and opportunities to form closer collaborative ties between departments.

And all this from a couple of fairly simple hand-sketched maps.

The power of drawing – mapping the way forward

Anything to do with aligning content, process, and people is bound to be tough, messy, and complex. But remember that we and others who inhabit these environments are not completely constrained by what has happened before. It is a natural reaction of ours to try to order an environment by fitting it into the categories of our expectations. So when we cannot find a way to fit any new ideas and concepts into these simplified slots, it’s really no wonder we experience those sharp pangs of panic. But we and everyone else around us don’t have to rely on the old established categories, we can always create new ones. Our reserves of intellectual capacity are vast. Perhaps if we’re continually looking upon new ideas and concepts as chaotic and threatening this can only be because we’ve never seriously tried to make use of our potential ability to cope with the unexpected. Believe it or not, we are all amazingly inventive and resourceful, and one of the ways to bring out our innate inventive qualities is to step away from our screens and just draw.

Maps like this can be used to quickly and clearly communicate our findings and ideas to others. And if sketched in their presence, we can even use them to tell stories. It can be an eye-opening experience for everyone involved. We have to fuel our innate curiosity. We have to be willing to look stupid if it eventually leads an insight. Mistakes are just part of the process.

Just remember to keep it simple. You’re not doing it to show off. Often it’s the overly-elaborate pictures that draw too much attention to the art, rather than the idea – and it’s the idea that we want them to remember and to develop.

There really isn’t anything about this level of mapmaking that is beyond us. Make use of shapes, arrows, faces. Find your axis. Make mistakes. Start over and over again until you get it right. There is nothing to be apprehensive about. Just keep it in mind that we’ve communicated using sketched maps from the very beginning. Mapmaking fulfils one of our most ancient and deep-seated desires, which is to understand the world around us and our place in it. It’s very much a basic human instinct.

Recommended reading

• Communicating Design: Developing Web Site Documentation for Design and Planning by Dan M. Brown
• You Are Here: Personal Geographies and Other Maps of the Imagination by Katharine Harmon
• Blah Blah Blah: What To Do When Words Don’t Work by Dan Roam
• The Back of the Napkin: Solving Problems and Selling Ideas with Pictures by Dan Roam
• Maps of the Imagination: The Writer as Cartographer by Peter Turchi

Visualising Data: Seeing is Believing

As humans, our ability to observe and analyse the contents of the world around us is both unique and astonishing, but so too is our capacity to form verbal and visual concepts. These seem to be the principal factors which have worked to our adaptive advantage in competition with other animal species. We are, in one respect at least, superior to other animals because we have developed a greater variety of systems of communication and expression, and one of these is art.

Figure 1 – Paleolithic Era cave paintings of giant elk and bison. Discovered in France and Spain respectively.

Indeed, some of the earliest known preserved examples of human expression (Figure 1) demonstrate our incredible ability to bring chaotic and complex environments under control through the magic of art, because to illustrate something is to transform it into whatever form or shape we want. And though we’ll never know for certain what our prehistoric ancestors were thinking when they painted pictures of cows, horses, bison and deer on the walls of caves, it is thought that because their paintings showed large and dangerous wild animals rather than humans, that this was their attempt to bring them under control — to tame them. It’s an interesting theory, particularly when we consider how many of these animals would come to be domesticated by humans thousands of years in the future.

So if art and other forms of creative expression are the power to transform and interpret, then science is the great identifier and unifier, and there a few better collisions of these two cultures than a diagram.

Figure 2 – Examples of Famous diagrams (clockwise from left): Da Vinci’s Vitruvian Man, the Pioneer Plaques, Florence Nightingale’s coxcomb diagram, Copernicus’s heliocentric universe.

Our history is littered with instantly recognisable diagrams (Figure 2). At their most potent they have the ability to express complex ideas simply, and an intellectual and artistic beauty that has the power to shift our perspectives or change our mind about things. Often it’s that desire for simplicity and beauty that leads to the truth.

Figure 3 – Copernicus’s heliocentric universe diagram. He had to move heaven and earth to draw it. (c. 1543).

Take, for example, Nicolaus Copernicus’s heliocentric model of the solar system (Figure 3), which would come to revolutionise the way we look at our place in the universe. For over two thousand years scholars and religious scriptures were steadfast in their belief that our planet was the static centrepiece of the universe, but Copernicus — then a little-known Polish cleric — dared to think along different lines. His concept revealed the solar system as we know and understand it: not with Earth at its centre, but the Sun. This radical new arrangement of the universe just seemed unreasonable and ridiculous at the time, with Earth millions of miles away from where it was supposed to be positioned. So how did Copernicus support his theory?

Figure 4 – Copernicus’s ‘On the Revolutions of the Celestial Spheres’ contains thousands of years worth of astrological data (c. 1543).

One thing you immediately notice as you scan through his life’s work “On the revolutions of the Celestial Spheres” are the amount of pages filled with numerical data (Figure 4), but they weren’t all his own. Copernicus not only relied on his own astrological data to recalculate the planetary positions, but thousands of years worth of previous observations by others. All this data underpins the completed diagram.

But despite the weight of its foundations, the beauty of his diagram lies in its simplicity. If you or I wanted to quickly explain the arrangement of our solar system to someone, you’d probably sketch something a lot like it. That it still hasn’t been bettered is testament to his achievement.

The data revolution

For someone whose investigations involved using data from several external sources, you wonder what Copernicus would have made of today’s huge democratisation of data. Right now we’re increasingly seeing barriers lowered between ourselves and rich data sets containing information about our communities, our politics and our governments. It’s not some fanciful idea to suggest that wider public access to numbers and statistics offers us a clearer picture of what’s really going on in the world, and that with this knowledge we can begin to make our lives better at a local, national and international level. As citizens, we should all be very excited indeed about the pace of this digital data revolution.

Tools

One of the exciting offshoots of this has been the emergence of powerful new tools for interrogating and presenting data. Tools which can help us all make better sense of our environments, help us find out if the things we think and believe are actually true or not, and communicate our findings in a way that our audiences understand and can act upon.

What’s great about this sort of work is that a lot of the stuff used to do it is freely available to everyone. None of these tools used for data extraction, exploration and visualisation will cost you money at their basic level to access and use at their basic level.

• Gephi
• Google Chart Tools
• Google Fusion Tables
• Google Refine
• Lucid Chart
• ManyEyes
• OpenStreetMap
• ScraperWiki
• Tableau Public

Where we fit in

I’m at pains to point out that I’m not a data scientist, nor am I a statistician, or mathematician; I’m just someone who has found visualising data to be both an effective way to gain a deeper understanding of the hidden processes that exist within organisations, and to amplify and simplify the communication of my content strategy recommendations and arguments to other audiences — namely decision makers. Essentially, it’s about trying to have a two-way conversation with the people who can effect change within an organisation using data-driven visual communication. What I’m not advocating for needn’t be a great deal of added work on your plate. The idea is to make effective use of some of the existing tools and deliverables you would typically call upon during a project’s discovery phase. Let’s take a look at a few ways we can do this.

The hidden people networks

One of the many effective methods of analysing the lifecycle of an organisation’s content is to conduct one-on-one interviews with key members of the authoring team, but finding the ideal people to speak to in a multi-departmental organisation isn’t always that easy. Consulting a hierarchical organisation chart can be helpful, but what they don’t reveal are those hidden relationships that are forged by everyday collaborative content work.

Figure 5 – A sociogram visualises the interpersonal relationships within a group. It is composed of nodes (individuals) connected by edges (relations).

A sociogram (Figure 5) is a visualised representation of the structure and patterns within a social network. They can be a powerful tool for discovering deeper meanings behind the relationships and communities within a network of people, and can be used to quickly reveal community clusters and calculate network science parameters such as degree and betweenness centrality.

Degree

The degree of a node calculated by the number of edges that are adjacent to it. So by ranking each node within a social network by degree, we can distinguish which individuals have the most connections (Figure 6).

Figure 6 – A sociogram ranked by degree. The more connections an individual has the higher their degree.

Betweenness centrality

Betweenness Centrality measures how often a node appears on the shortest paths between nodes in a network. So by ranking each node within a social network by betweenness centrality, we can distinguish which influential individuals have the most connections across distinct community clusters (Figure 7).

Figure 7 – A sociogram ranked by betweenness centrality. The more connections an individual has to different community clusters the higher their betweenness.

Let’s build a sociogram

All content management systems (CMSs) and intranets worth their salt automatically generate and store logs which maintain a history of activity performed on them. With information on each page requests, including when they were created, edited and who was responsible, to name but a few, it is possible to chart the lifecycle of content pieces or whole pages. If you, your database or CMS manager, or someone equally clever, extracts a set of site-wide logs for a specified period, then you can put together your very own sociogram using the data.

Once opened in a rich text text editor, your log file might look a bit like this:

#Fields: date time c-ip cs-username s-ip s-port cs-method cs-uri-stem cs-uri-query sc-status cs(User-Agent)
2012-09-03 00:10:19  XXX.XXX.X.211 clarke_n  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=84 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:10:39  XXX.XXX.X.17 olson_b  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=37 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:11:12  XXX.XXX.X.40 zajac_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=37 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:20  XXX.XXX.X.29 arecchi_f  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=168 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:50  XXX.XXX.X.107 chalmers_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=174 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:52  XXX.XXX.X.178 harding_a  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=174 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:14:38  XXX.XXX.X.107 chalmers_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=73 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)

That might have been too much, too soon. Let’s take a closer look at a single entry.

2012-09-03 00:09:53 XXX.XXX.X.104 russell_g XXX.XXX.X.103 80 GET /admin/pages/content.php?id=12 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)

The highlighted fields are the ones we’re particularly interested in. They are: date, time, client ip address (c-ip), client username (cs-username) and the content accessed (cs-uri-stem).

With this mind we can return to our scary log file. What we’re looking for is the same content or page accessed by two or more different authors within a certain timeframe. I’ve highlighted two such examples below:

#Fields: date time c-ip cs-username s-ip s-port cs-method cs-uri-stem cs-uri-query sc-status cs(User-Agent)
2012-09-03 00:10:19  XXX.XXX.X.211 clarke_n  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=84 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:10:39  XXX.XXX.X.17 olson_b  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=37 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:11:12  XXX.XXX.X.40 zajac_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=37 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:20  XXX.XXX.X.29 arecchi_f  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=168 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:50  XXX.XXX.X.107 chalmers_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=174 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:13:52  XXX.XXX.X.178 harding_a  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=174 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)
2012-09-03 00:14:38  XXX.XXX.X.107 chalmers_s  XXX.XXX.X.103 80 GET /admin/pages/content.php?id=73 Cmd=contents 200 Mozilla/4.76+[en]+(X11;+U;+Linux+2.4.9-ac7+i686;+Nav)

Next we want to extract these usernames (cs-username), including others that fitted our criteria, and add them to a two-columned spreadsheet (Figure 8) with the first set of usernames under a column heading of Source and the second under Target. Our examples had olson_b and chalmers_s acting as the sources and zajac_s and harding_a as the targets.

Figure 8 – A two-columned spreadsheet with extracted usernames.

Once all the usernames have been added, we can save the sheet as a comma-separated values (CSV) file and import them into some graph visualisation software, such as Gephi. We might initially see something like this huge mess of nodes and edges (Figure 9).

Figure 9 – Gephi rendering of usernames extracted from log file (click image to zoom).

Running a layout algorithm

Though it’s not making a very clear job if it right now, what this rendering is showing us are the links that exist between all the people who have edited content pieces consecutively. If we were to run a good quality layout algorithm, we would begin to see the clusters of connected nodes forming (Figure 10).

Figure 10 – ‘Force atlas 2′ layout algorithm applied to Gephi rendering (click image to zoom).

I’ve used Gephi’s own ‘Force Atlas 2’ to position these nodes in an aesthetically pleasing way. This is now looking much more like a sociogram. We can easily see the different communities to which these individuals are connected identified in the graph.

Ranking nodes by degree

To help us see the most important individuals in the network, we need to rank these nodes by degree. We’ll illustrate this by changing the colours of each node depending on the number of connections they have (Figure 11). In our sociogram, the ‘greenest’ nodes have the highest degree.

Figure 11 – Nodes ranked by degree (click image to zoom).

Ranking nodes by betweenness centrality

Next we’ll find which influential individuals have established the most connections across distinct community clusters. To do this we’ll rank the nodes by betweenness centrality (Figure 12). In our sociogram, the largest nodes have the highest ‘betweenness’.

Figure 12 – Nodes ranked by betweenness centrality (click image to zoom).

Filtering nodes to reveal strongest group

Despite using both colour and size to help us see who are the most influential nodes in this network, it is still a little too crowded. So we’ll use a filter to remove the weaker nodes in our sociogram (Figure 13).

Figure 13 – Nodes filtered to reveal strongest group.

Now we are left with the three individuals who carry the most influence. One or all of durning_j, scott_f, and pearson_r might well occupy lower positions on an organisational chart than others, but their importance to the way content flows around this organisation means they are clearly worthy of our attention.

Lessons

By extracting and visualising the data contained in a log file we were able to:

• See the different community clusters to which people were connected
• See which individuals held the most influence over multiple groups
• Create an alternative organisational chart useful to any content strategist who’s been air-dropped into a project

Relationships between content

NOTE: I am indebted to Dorian Taylor for this idea. Do yourself a favour and read everything he’s ever written, particularly his article “No Longer No Sense of an Ending” which featured in issue 3 of Contents Magazine.

In a piece on his website titled “visualizing paths through the web“, Dorian introduces us to a rendering of the most frequently-trodden paths through his website:

When auditing content for the Web, it’s important to remember that although many of us still write Web content as isolated documents, they are very rarely read that way. It’s entirely feasible for a reader to encounter inconsistent or confusing writing between one page and the next. In order to fully appreciate the story we’re telling our audience, we should look at it in context.

Using the same techniques as Dorian I was able to create my own rendering of the most frequently-trodden paths through the website of a fictitious UK bus and coach operator (Figure 14). The potential is there to learn to learn a great deal about the way the content relates to one another.

Figure 14 – Mapping the movement of users between the web pages of a fictional UK bus and coach operator.

But before we begin the task of rendering one ourselves, let me use a couple of simplified examples to explain what we’re looking at (Figure 15).

Figure 15 – Simplified network of the flow of users between web pages.

Each node represents an individual page on the website, while rach directed edge represents the flow of users, or traffic, between two pages.

Figure 16 – Weightier nodes and edges indicate key paths and stops.

In this example (Figure 16), the larger nodes in the graph have a higher PageRank, which is essentially a higher importance in relation to the other pages in the network, while the thicker directed edges indicate a higher frequency of flow between two pages.

In his piece, Dorian goes on to point out that web servers log information on every available referring resource (the previous page) and each new request (the next page) we make.

When opened in a rich text editor, your web server log file might well look a bit like this:

XX.XXX.XXX.86 [30/Aug/2012:11:09:27 +0100] GET /contact-us/ HTTP/1.1 200 14728 - "http://www.crosscountrycoaches.com/destinations/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"
XX.XXX.XXX.86 [30/Aug/2012:11:09:29 +0100] GET / HTTP/1.1 200 12007 - "http://www.crosscountrycoaches.com/destinations/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"
XX.XXX.XXX.86 [30/Aug/2012:11:09:29 +0100] GET /contact-us/view-your-ticket/ HTTP/1.1 200 14084 - "http://www.crosscountrycoaches.com/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"
XX.XXX.XXX.86 [30/Aug/2012:11:09:37 +0100] GET /services/ HTTP/1.1 200 13428 - "http://www.crosscountrycoaches.com/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1" XX.XXX.XXX.86 [30/Aug/2012:11:09:38 +0100] GET /login/ HTTP/1.1 200 17284 - "http://www.crosscountrycoaches.com/services/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"
XX.XXX.XXX.86 [30/Aug/2012:11:09:42 +0100] GET /reprint-your-ticket/ HTTP/1.1 200 27788 - "http://www.crosscountrycoaches.com/services/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"
XX.XXX.XXX.86 [30/Aug/2012:11:09:42 +0100] GET /services/terms-and-conditions/ HTTP/1.1 200 11638 - "http://www.crosscountrycoaches.com/reprint-your-ticket/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"

You could be forgiven for thinking it exactly like the log file sample we saw earlier, but there are some subtle differences. From this log file sample we want to locate each referrer-referent connection (previous and next page) that come under matching client ip addresses. To simplify things let’s focus on two particular entries. The fields we’re interested in are (in highlighted order) client ip, next page URL and previous page URL.

XX.XXX.XXX.86 [30/Aug/2012:11:09:29 +0100] GET /contact-us/view-your-ticket/ HTTP/1.1 200 14084 - "http://www.crosscountrycoaches.com/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1" XX.XXX.XXX.86 [30/Aug/2012:11:09:42 +0100] GET /reprint-your-ticket/ HTTP/1.1 200 27788 - "http://www.crosscountrycoaches.com/services/" "Mozilla/5.0 (Windows NT 6.0; rv:14.0) Gecko/20100101 Firefox/14.0.1"

With the help of our friendly web server manager it is possible to turn this log into a list of page-by-page connections, each weighted by the intensity of the traffic flowing between them. Cleaning the data is simply case of stripping it of all non-human visits by web crawlers like Google Bot, Bing Bot and others. Unless your site is about robots or creepy-crawlies you can safely remove any mentions of ‘bots’, ‘spiders’ and ‘crawlers’.

As before, we want to extract these URL’s and others that fitted our criteria and add them to a two-columned spreadsheet (Figure 17). The referring URLs should be placed under a column heading of Source and the referent URLs under Target. Finally, we need to save the sheet as a comma-separated values (CSV) file.

Figure 17 – A two-columned spreadsheet with extracted page URLs.

Introducing Cross Country Coaches

Before I continue I wish to formally introduce you to our fictional UK bus and coach operator, who will be acting as a working example for the remainder of this talk. Cross Country Coaches’ main services include airport runs, day trips to major UK towns and cities, holiday camps and amusement parks, as well as sporting and music events. Everything on their website should be geared towards the journey planner; the starting point from where e-tickets can be purchased.

If we were to import the extracted URL data into Gephi we might see something like this even larger mess of randomly distributed nodes and edges (Figure 18).

Figure 18 – Gephi rendering of referrer/referent URLs extracted from web server logs (click image to zoom).

Running a layout algorithm

We’ll run the same layout algorithm as before (Gephi’s Force Atlas 2) to position our nodes in an aesthetically pleasing way (Figure 19).

Figure 19 – ‘Force atlas 2′ layout algorithm applied to Gephi rendering (click image to zoom).

Ranking nodes by PageRank

To help pick out the most important pages in the network, we’ll use colour and size to rank our nodes by ‘PageRank’ (Figure 20).

Figure 20 – Nodes ranked by PageRank (click image to zoom).

Filtering nodes to reveal strongest group

The colour and size of nodes are helping us to see what is the most influential content in this network, but we could benefit from clearing away the clutter. This level of complexity may suit someone who is close to the data but for those that are not we should consider filtering out the network’s weaker nodes (Figure 21).

Figure 21 – Nodes filtered to reveal strongest group.

Probably the most interesting part of this simplified network to note is the amount of arrows which point towards /contact-us/, particularly from the /journey-finder/. That many are choosing to leave this process of buying an e-ticket to contact Cross Country Coaches is interesting and would be well worthy of further investigation.

Nevertheless, what we have ourselves is a sample of the most frequently-trodden paths through this website — a very useful starting point for any investigation into a website’s content. So we’ll export the node data as a .csv as this will come in handy for the final part of this talk.

Lessons

By extracting and visualising the URL data contained in a web server log file, we were able to:

• Provide a different perspective on the stories we’re telling our audiences
• Filter the data to reveal the key paths and stops users are making
• Generate a sample of the the most frequently accessed content for this website

Playing with numbers

Using internal data

We can add further value and depth to our visualisations by informing them with data from our own investigations. We’ll take our exported node table and manually add a set of new columns full of data extracted from a recent audit of Cross Country Coaches’ website (Figure 22).

Figure 22 – Set of new columns added to exported node table (click image to zoom).

For example, we could add data on which department (Responsibility (Dept)) and individual (Responsibility (Individual)) is responsible for maintaining each page, whether the content is maintained in-house or by external parters (Source), when the content was last updated (Last update) and criteria for measuring the quality of the content (Actionability, Accuracy and Usability).

To speed this process up we could enlist the help of Google Fusion Tables to merge data across two spreadsheets by pairing up one (or more) columns with matching values (Figure 23).

Figure 23 – Merging two tables together by linking one or two sets of matching column data (click image to zoom).

To help us take a deeper dive into the additional data we’ve added to our exported node table, we could use a rapid chart creator like Tableau Public. Designed for PCs (although a Mac version is in the works), Tableau Public makes it simple and easy to make pretty complex visualisations with up to 100,000 rows.

Measuring content quality

Let’s pose a question for the data to answer:

Q. Is there a perceived difference in the quality of content maintained in-house by Cross Country Coaches and through their external partners?

By importing the data I’ve been able to create a simple set of bar charts which tell us what each content source scored out of five for Actionability, Accuracy and Usability for the content they are responsible for (Figure 24).

Figure 24 – Bar chart created in Tableau Public showing the average scores for each content source (click image to zoom).

From the chart it is clear that Tourism Partners scored below average for each criteria. As they are primarily responsible for the content for each town and city destination, let’s filter the data further to generate a breakdown of the scores for each (Figures 25, 26, 27).

Measuring the actionability of the content for each town and city destination

Figure 25 – Bar chart created in Tableau Public showing the Actionability scores for each town/city destination (click image to zoom).

Measuring the accuracy of the content for each town and city destination

Figure 26 – Bar chart created in Tableau Public showing the Accuracy scores for each town/city destination (click image to zoom).

Measuring the usability of the content for each town and city destination

Figure 27 – Bar chart created in Tableau Public showing the Usability scores for each town/city destination (click image to zoom).

Looking at each of these scatter plots there a few recurring destinations which scored below average, particularly Brighton and Hove and Edinburgh who scored two out of five for both content actionability and accuracy. While it would be fair to say that these destinations should be first in line for a review, could data from external sources help with prioritising our efforts? Let’s bring some into play.

Using external data

To investigate whether external data could help us, I downloaded data on domestic tourism statistics between 2009-11 from VisitBritain.org. The only problem was that the data wasn’t in a raw state, but rather housed inside a PDF file. Thankfully, you can save hours of re-keying and checking by using some of the free PDF-to-Excel conversions available. I’ve had success with both PDF to Excel Online and Zamzar.

Once more we can call upon the services of Google Fusion tables to merge our data for each town and city destination with the downloaded VisitBritain.org data (Figure 28).

Figure 28 – Using Google Fusion Tables to merge our data for each town and city destination with the VisitBritain.org data (click image to zoom).

Now let’s map the data using Tableau Public (Figure 29). Included are bubbles marking the location each town and city from the Cross Country coaches data. The bubble sizes represent the total domestic tourist trips made between 2009-11 (in millions), and their colour intensity represents the total tourism spend between 2009-11 (in millions).

Figure 29 – Map of British isles showing tourism data for each town and city destination (click image to zoom).

Unsurprisingly, London take the prize for the largest and darkest bubble. As the capital of England will always remain a priority, we’ll remove it from the map and, in the process, observe the rest of the British Isles breathing a huge sigh of relief (Figure 30).

Figure 30 – Map of British isles showing tourism data for each town and city destination excluding London (click image to zoom) (click image to zoom).

Let’s now use the sliders to filter out the destinations which scored the highest for Actionability, Accuracy and Usability (Figures 31, 32).

Figure 31 – Map of British Isles showing tourism data for each town and city destination that scored between 2 and 4 for each content quality criteria (click image to zoom).

Figure 32 – Map of British Isles showing tourism data for each town and city destination that scored between 2 and 3 for each content quality criteria (click image to zoom).

We’re now left with Birmingham, Edinburgh, and Brighton as the destinations which had the highest amount of trips and spend combined with the lowest scores for Actionability, Accuracy and Usability. Might they be our priorities?

Lessons

By extending our exported node table, we were able to:

• Dive deeper by filtering and partitioning data from our own investigations
• Explore possible possible new angles and ideas by importing and presenting external data
• Develop a basic narrative around our data by adding interactive elements to our visualisations

You can do it!

Don’t be afraid to play and experiment with your data. Have fun asking questions of it. I have found that approaching it with such a mentality means it often yield its secrets and stories with surprising ease. And though we often associate numbers with authority and certainty, uncertainly can be a great way of raising new questions and sharing them with others. Getting your work in front of people might mean you get help and co-operation back.

Here’s a video of my talk from last month’s CS Forum 2011 in London. I had the pleasure of sharing the stage with some incredibly smart folk, so do make yourself comfortable and watch them all, particularly the excellent lightning talks by Shelly, Matthew, Nicole, and Sara.

Presentation links

• See the finished diagram on visualizing.org
• View the survey results on Flickr
• Study the survey data spreadsheet

Let me begin by extending a huge and sincere thanks to everyone who played their part in this, particularly those who responded to the survey, encouraged others to do so, and remained patient as I worked out what to do with the results, and to the attendees of CS Forum 2011 who offered such kind words following my sole destroying (you had to be there) attempt to squeeze the last six months into 20 minutes. I can’t deny it’s been fun.

Though the finished diagram is by no means perfect, I can say with a measure of confidence that not only are these the six commonest paths today’s practising content strategists have taken to reach the discipline, but that they demonstrate the extent of our varied skills and approaches. It only serves to emphasise how much we need to continue sharing a little of what we’ve picked up along the way.

As you’ve probably gathered, that lad was me and until earlier this month that occasion was the last time I had done any speaking in public. You see, I’d long considered public speaking to be a domain other people inhabited; those who could tell captivating stories laced with wit, suspense and meaning; those with something interesting and intelligent to impart; those who looked as though they were *enjoying* themselves up there in front of an audience. Not for me, thank you very much. I was more than happy having the option of sinking into my seat on those odd occasion when audience members were invited to take to a stage.

That was, however, until the beginning of this year when I drew up a longlist of things that would take me out of my comfort zone, only to find that speaking in public occupied no less than three items. I waited until the very last hour before submitting a talk for CS Forum 2011 before gasping in an equal measure of delight and horror weeks later upon seeing my name and mugshot amongst some of our industry’s most informed and celebrated pracitioners. My mind rapidly turned to the question of how I was going to close that 18-year speaking gap between a school production and an international conference. No amount of quirky mock Victorian attire would be able to mask my monotonous delivery this time.

So imagine my relief when the chance arose to deliver a 5-minute lightning talk at the inaugural London Content Strategy Meetup event “Wrapping up Confab, unwrapping CS Forum”. I now had an excellent opportunity to put myself through the ordeal of speaking and see how I react under the pressure. Would every other word be an expletive? Would I start referring to myself in the third person? Would I somehow lose my trousers? Pleasingly, whilst I’ll happily admit to being a little overawed at the prospect of sharing the bill with seven excellent speakers, I felt I didn’t do too bad a job of blowing the cobwebs away.

The cake was no lie

The subject of my talk was Confab 2011 and in particular a session led by Erika Hall, co-founder of Mule Design Studio. At the time, I was in the middle of conducting my survey of today’s content strategists. So when Erika briefly took us back to the life of a content strategist during those halcyon days of the mid-to-late 1990’s my ears immediately pricked up. Such was her experience, it was was also interesting to hear Erika’s take on the recent rise in interest in content strategy and how she believes it sits alongside the other design sub-disciplines. I’d heartily recommend flicking through her slide deck.

With a few seconds left to spare I ended my talk by (energetically) paying tribute to Confab’s dedicated army of live tweeters. With the brain-to-finger speed of a court reporter and the anticipatory skills of an air traffic controller, they helped keep the #Confab hashtag competitive at a local to national level throughout the duration of the conference.

To finish, I must thank everyone who made it past the heavy police cordon to be with us that night. Thanks also to Jonathan and Elizabeth for organising, all the brave souls who spoke as well as the event sponsors for making it happen. Let’s do it again soon.