Why we make flawed decisions

Three great TED talks demonstrating how flawed the human decision-making process can be and why happiness is always relative. Enjoy!

Barry Schwartz: The Paradox of Choice

[ba-youtubeflex videoid=”VO6XEQIsCoM”]

Malcolm Gladwell: Choice, Happiness and Spaghetti Sauce

[ba-youtubeflex videoid=”iIiAAhUeR6Y”]

Dan Gilbert: Stumbling upon Happiness

[ba-youtubeflex videoid=”LTO_dZUvbJA”]

We are now responsible

Others have pondered whether one of the side effects of the blogging phenomena will be a challenge to traditional politics. I hope so. At times, it feels like the physical world is out of sync. Catching up on some reading today, I stumbled across a quote that neatly summed up what we could/should be doing…

“In the Wiki world, everyone has access to the same knowledge. Many people
no longer feel that experts have additional authority. If knowledge is
power, then today everyone has enormous power. What we need to do is get
all of the people to take responsibility for the knowledge that we all possess.”

– Thomas Homer-Dixon, interview in New Scientist (subscription required to access full article)

Dashboard Design

As a follow on to a previous post – Diluted information, this is a quick review of just two ideas within Stephen Few’s excellent book – Information Dashboard Design.

Spark Lines

Spark lines are an idea from Edward Tufte, author of books such as Envisioning Information. The concept of a spark line is simplicity itself – take a line graph and delete everything but the line itself. This enables you to present a big chunk of information in a very small space, perfect for dashboards. Take the following example:

This is fairly typical of a dashboard or scorecard used to present key performance indicators (KPIs). You’ve got the sales figure, a green light and an arrow going in the right direction. Sales must be good! But how much information do you really have here? You know that your sales number is in a good position today, and compared to your last measurement it is going in the right direction. What does that information really tell you?.

Let’s use a sparkline instead:

The spark line enables you to present a far bigger range of figures with minimal extra space. When you use the traffic light and arrow, you can only see the last two points on this line. But now you can potentially see a whole year’s worth of data. And what does this spark line tell you? That sales are currently good and going in the right direction, but the pattern through the year looks very erratic. Who knows what next month’s figures will look like. Why is that? Can anything be done to settle into a growth curve or are the causes all external and out of your control? Are your suppliers letting you down, disrupting your production line? The spark line encourages you to investigate, providing you with an opportunity to be proactive in resolving issues. Traffic lights create a reactive approach – you only start to investigate when they go red or the arrow is already pointing in the wrong direction.

Bullet Graphs

Bullet graphs are an idea from Stephen Few, providing an alternative to the popular dashboard tool – gauges. Another example:

This is a classic gauge, displaying a measure and the range of values that indicate where you would like it to be. In this case, customer satisfaction is healthily in the green at above 75%. But again, how much information have you been given? A measure and where it sits within a range. That gauge takes up an awful lot of screen estate to give you so little. Enter the bullet graph:

A bullet graph is the perfect diet – it slims down a gauge and gives you extra value to boot. Styled like a thermometer, you can still see the range of values, but now a simple line indicates your current measure. But in addition, the horizontal line can provide a clear alternative measure (this is harder to do with the gauge, where it ends up looking like half a clock). That horizontal line might represent last month’s figure, or could represent the target. Either way, you now have a comparative measure.

To put these two ideas together, it’s worth comparing two dashboards to demonstrate how much or how little information is often presented:

Here is a classic dashboard, with lots of pretty graphs and charts. But, before you even figure out what it all means, how much does it really tell you? Now let’s look at an example that uses spark lines and bullet graphs:

Before you even start to interpret what this dashboard contains, you can see a lot more information is available in a format that you can analyse. And now let’s take it a stage further… Let’s take that dashboard out of the browser…

This screenshot is Excel – your good ol’ spreadsheet (and it’s not even the latest version at that). This simple screen of data inside Excel contains more useful information than I’ve seen in far more complex and expensive business intelligence tools.

How do you create a spark line in Excel? Simple – just create a normal line chart and delete out everything other than the line (go into Chart Options, delete the grid lines, the axis, the heading, the background, everything!).

The screenshots above are all available over on The Dashboard Spy’s web site, a blog I recommend you subscribe to if you are interested in this subject. Having read Stephen Few’s book, it is great fun to critique the different dashboards on display there. I was going to create my own examples in Excel, but others beat me to it and I’m all for not reinventing the wheel. I’ll use a later post to explore if/how Excel 2007’s new data visualisations can enhance dashboards… …and since starting this post, The Dashboard Spy has beaten me to it again.

Final tip

If you are tasked with designing a dashboard or scorecard, there is one important question you should always ask the audience who will be using the end result. Is the dashboard being designed to provide quick answers or prompt for more questions? Too often, the answer is the former when it should be the latter but that’s a whole different political question 🙂 Let’s just focus on designing for the desired role…

Going back to the first image – the sale figure with a green traffic light and an arrow going upwards. That is a classic ‘answer’ KPI. Are my sales good? Yup! Are they going in the right direction? Yup! Great! I can move on to the next task. If that really is all the audience wants to know, then stick with the colourful images and minimal content. The irony is, dashboards that provide quick answers are largely redundant any way. Quick answers are answers you already know and just want validated. If you are interested in sales, you are going to know if the company is doing well or not before you even look
at that dashboard. Just walk down the corridor, lurk by the water cooler, linger in the toilet even – you’ll know from the atmosphere within the office if your sales are good or not. It won’t necessarily tell you whether they are going to stay that way…

The spark line is far more suited to the valuable use of a dashboard – to prompt proactive investigations when the numbers may be good but don’t look quite right…

Technorati Tag: Dashboard

Diluted Information

It’s always funny when culture leads to misinterpreting a joke. Whilst chatting with someone during the SharePoint conference earlier this year, they muttered how BSM was an apt acronym for Business Scorecard Manager 2005. I wondered what on earth business scorecards had to do with the British School of Motoring before it dawned on me that they were referring to a completely different use for the letters BS…

…and whilst they may have been poking fun at the product itself, the same cynicism could also be applied to business scorecards in general as they become easier to build and deploy…

Intelligence in its various forms (such as reports, portals, dashboards, and scorecards) has been on IT’s radar for quite some time now, but 2007 is likely to be the year it finally arrives as an everyday tool on the desktop. (I think it will be 2008+ before it is successfully adopted on mobile devices, given how long it is taking companies to provide even basic web sites in a format suitable for small screens and flaky network connections.) Gartner has highlighted collective intelligence within their top three emerging technologies for 2007. Microsoft is building up a complete stack of maturing business intelligence (BI) tools, having started with the launch of SQL Server 2005 back in November 2005 and now culminating in the launch of Office 2007, a revamped SharePoint Server 2007 containing all sorts of information management goodies, and a BI portal to rule them all: PerformancePoint Server 2007. And I suspect good ol’ Excel will trump the lot thanks to its popularity combined with new features that make data visualisations incredibly easy.

The technology to analyse and present information is improving, and there is no shortage of information waiting to be mined for ideas. During a Gartner conference I attended in 2005, I heard an interesting view point that sums up the ever increasing quantity of information available: “…combined with lowering storage costs, will we start to see ‘write once, read never’?” It’s a powerful thought.

But before we get all giddy and excited about analysing anything and everything, we need to consider what outcomes we desire. There is little point in staring at a dashboard or scorecard if the information served up does not help make better decisions and/or influence actions. And one of the easiest traps to fall into (and there are many) is presenting too much information. Presenting too much information creates two issues: the paradox of choice and diluted messages.

The paradox of choice is a relatively simple concept – when presented with too many options, we are more likely to choose nothing at all. Doh! Barry Schwartz has written a whole book about this concept (guess what the title is!). To find out more:

Diluted messages is a different issue with the same end result. When we see/hear the same information over and over again, no matter how strong the message (powerful, wonderful, shocking, horrific, intense desire… whatever), its effects become diluted – they weaken to the point where we just don’t notice or care any more. A simple example to demonstrate this challenge was provided by film director Steven Spielberg in an interview last year. He was discussing the challenges with filming Jaws and how, when the mechanical shark broke, he couldn’t delay filming so had to create clever camera shots and create an impression that the shark was there even though you couldn’t see it:

“If I had made Jaws today, it would have been a digital shark that never breaks. I would have have used it four times as much and the film would have been four times less scary…”

Take note! As business intelligence tools become pervasive throughout organisations, the need for good BI design will become increasingly important to ensure that the right information is served up in the right format when and where it is needed to make decisions and act on them.

I plan to write more on this subject throughout the winter and cover some of the other traps that can make intelligence worthless. But if you are interested the design side, without doubt the best place to start is a book called “Information Dashboard Design” by Stephen Few published earlier this year. And once you have read it, visit The Dashboard Spy blog for samples of good and bad designs.

Related blog posts:

Hmmm, I’m beginning to wonder if there is a connection between the month of October and my brain getting interested in intelligence…

Sticky Information

[update] Actually, it was the weekend before last. I forgot to publish this post last week…

I was reading a couple of unrelated articles in The Sunday Times last weekend that shared a common theme – the stickiness of information. It made me think (yes, my brain is still in recovery). On the one hand, we can be very lazy about acquiring information (Google trumps the local library for starters), spectacularly avoiding it when we don’t like what it tells us. But on the other hand, when we find information that fits our expectations, it becomes very sticky and we are reluctant to let it go. I believe this has important consequences for information systems design.

Article no.1 – Michael Schumacher

In Formula 1 motor racing, Michael Schumacher has been at the top for over a decade and the results speak for themselves. Seven world championships and in the running for an eighth. With his retirement all but announced (and it was on Sunday), Martin Brundle wrote an article covering Schumacher’s career and the reasons why he won’t be remembered just as a great champion.

“He cannot see when he crosses the line between tough but fair and ruthless but foul… exacerbated by his total belief that he cannot be wrong. He has a default mode in the car: if you’re going to pass him, he will drive you off the road.”

Brundle goes on to list some examples, including two title-deciding races that saw the contenders seeing dust, literally. There were other incidents too, such as when his team-mate was required to move over to let him win, and when he won a race by delaying a penalty pit stop until his final lap to minimise the time he would lose. He didn’t break the rules but his behaviour didn’t fit with ‘fair play’, an emotive subject. And that’s why everybody remembers. For all the races he won, the pole positions, the podium finishes that by far outweigh incidents of questionable behaviour and the sheer brilliance of his driving skills, he’ll be remembered for the sticky information.

Article no.2 – Afghanistan

On a far more serious note, the ongoing conflict in Afghanistan was back in the headlines. A comment in one of the many articles highlighted the challenge facing the NATO-led campaign:

“The Taliban have won the information war”, admitted a senior military officer last week. “Everyone is convinced we’re there to destroy the poppy and thus their livelihood. We’ve not provided any information to farmers so why should they think any different?”

Here is sticky information of a very different and far more dangerous nature, putting lives at risk. The story the farmers have been given fits with the evidence in front of their eyes and directly affects them in a bad way. Whose side would you take if you were in their shoes? How do you persuade them to believe you instead?

Fortunately, most information systems do not involve risking peoples’ lives. But information systems design rarely takes into account the sticky factor. Instead, the focus is usually on prioritising information based on its importance, and importance is usually determined by the sender rather than the receiver. Hence the home page of an intranet will be filled with business goals and news (e.g. “we’ve won a big new order for product X!”) that, at best, gets read once whilst the sticky information (e.g. “they’re going to outsource product X and make us all redundant”) travels through the unofficial grapevine. Next time you are designing a web site or thinking about adding another KPI (key performance indicator) to a dashboard, ask whether or not the information is sticky – will it change actions or will it be ignored? Is there existing sticky information that needs to be removed? Can it be removed?

Sharing birthdays

I have always been surprised how often I have been in a small group with at least two people sharing the same birthday (it’s happened at school, work, on holiday, on projects, basically more times than I can remember). With 365 days to pick from, it seems logical to think you’d need a pretty large group before duplications start to occur…

…but it takes just 23 people for there to be a greater than 50% chance that two people in the group will share the same birthday. For a good explanation of the maths behind the fact and why we so easily get it wrong, read The Law of Small Errors.

I find this simple probability exercise serves as a good reminder to test assumptions when analysing statistics. Sometimes the answer we expect to see is not remotely close to reality, but our brains have an annoying habit of preferring to believe expected myths rather than actual truths, no matter what evidence is presented before us. As we start to integrate business intelligence and performance dashboards into everyday activities, we need to beware our tendency to draw the wrong conclusions from the information presented.

I’ve read a few good books that cover this subject recently, including:

Group behaviour

Wandering through other peoples’ blogs again and I fell upon a neat collaboration experiment…

Discovery Channel has published an article ‘Cockroaches make group decisions‘ describing how scientists placed the bugs in a petri dish containing shelters and observed how they divided themselves up. Cockroaches are silent creatures. The results are fascinating:

After much “consultation,” through antenna probing, touching and more, the cockroaches divided themselves up perfectly within the shelters. For example, if 50 insects were placed in a dish with three shelters, each with a capacity for 40 bugs, 25 roaches huddled together in the first shelter, 25 gathered in the second shelter, and the third was left vacant.

When the researchers altered this setup so that it had three shelters with a capacity for more than 50 insects, all of the cockroaches moved into the first “house.”

I’m always fascinated when learning how different species co-ordinate and compete. Ants are the usual topic of conversation because their emergent behaviours have been successfully applied to robots and simulation programs.

But this simple experiment did get me thinking. How fascinating would it be to conduct the same experiment with a large petri dish (room, perhaps) and a group of people, such as a team or department, given one instruction ‘you have 5 minutes to get inside a shelter’. Observing that experiment could be fun – you’d sure learn a lot about the dynamics of the ‘team’…

Thanks to Clive Thomson who’s blog entry caused the stumble 🙂

[Update] …I wonder why the cockroaches decided to do an even split… it would be interesting to continue the experiment with additional shelters matching divisions (e.g. remove 2 cockroachs, reducing the group to 48, and provide 3 shelters capable of taking up to 30. Would they split into 2 groups of 24 again or 3 groups of 16…)

Web Design Flaws – part 2

So, more on my pet frustrations with organisations unable to get Web 1.0 right and why that should be fixed before everyone gets too carried away with the potential of Web 2.0… (You can read part 1 here)

Part 2 – Email

Why do companies do such a bad job of using email to connect with customers? They force you to enter your email address before you can do anything (e.g. download a document or trial software) and then spam you with useless automated messages from that point onwards but don’t provide a way for you to respond.

Sample culprit: Handango (http://www.handango.com)

I have an iPaq 4150 – gorgeous little skinny PDA that, thanks to a spare battery, can keep running all day (8 hours+ including watching movies). It serves 3 primary functions: my organiser (calendar, contacts, notes, tasks, email + syncs the lot with my work and home PCs), research (AvantGo, RSS reader + viewers for most doc formats), and play (games, movies, music, ebooks etc. – thanks to 1.5Gb of SD storage). Actually, it’s now got a 4th use, having acquired a GPS receiver and installed PocketStreets, but I digress. I decided I needed a new game for distraction during travel times, and went for a browse on Handango’s site. I decided to ‘trial’ Riven (‘trial’ = trailer rather than demo, as it turned out), which naturally required me to enter my email address in order to download the required file. Sure enough, a few days later, I receive an automated email thanking me for recently downloading a trial and including the link should I wish to purchase the full program. The email then includes a ‘few suggestions’, should I be looking for something else. The closing blurb goes as follows:

This is a one time email sent to the email address as entered when downloading a trial application from Handango. You are not subscribed to any additional mailing lists.


Why would Handango not be interested in any reply I might want to make? Had the email instead ended along the lines ‘We’re interested in your feedback on this trial and how we can improve our services, please send any comments to [insert email address]’, I would have clicked the link and sent them the following response:

“Thanks for the email. I have actually already purchased Riven but from another web site. Whilst checking reviews to see what others thought of the game running on a PDA, an advert popped up offering the product at a 25% discount over on ClickGamer and I purchased it from there. Had you also been offering the same discount, I would have purchased it from you having previously purchased from your site.”

At least that way, Handango would know a) the trial worked, in that it led to a purchase, and b) they lost the sale because the identical product was available on discount somewhere else. As a result, they could perhaps work with the supplier to negotiate a discount to protect future potential sales and even, shock horror, go one step further and respond with something along the lines: “We’re sorry not to have been able to offer the product at the same price as one of our competitors. Next time, please do contact us first as we will always try to match the lowest available price for software.” And you know what, if they replied like that, I would check back with them first next time. If you want to increase your potential sales, basic analytics from your web site won’t help much, you need feedback. I don’t doubt I could revisit the web site, locate the ‘contact us’ link that is likely to exist and supply some feedback there, but a) it takes a lot more effort on my part, and b) will disappear amongst unrelated emails within the standard website inbox.

The web offers also sorts of ways to increase sales and improve customer service, but it requires different methods and processes to traditional channels. First and foremost, it needs to be as easy as possible for your potential customers to get what they want. If you want rich feedback, make it as easy as possible to do so (you went to the trouble of collecting that email address…) It seems that very few companies really take the effort to use the Web well to improve their business…

(Again, just like the last post, I must stress, the culprit named here just happened to be a site I’ve visited recently. They are not alone and there will be plenty of other examples out in cyberspace.)

Practical planning

Having been buried in a couple of customer projects for the past few weeks, yesterday morning the sun was shining and I finally had time to ride one of the horses… and naturally spotted some similarities between riding and the projects…

Both projects involved reviewing and updating existing information systems – one focused more on content management, one focused on collaboration, both introducing new working methods into the organisation. And introducing new methods means introducing change.

Project number 3 – Harry, the horse I was riding – shares this last point. Harry has been rested for over 6 months and is unfit with a thick winter coat and not much muscle – looks more like a seaside donkey than the show jumper pictured on my Contacts page. This was our first session together, and the rider’s condition is not much better than the horse, with the exception of the hairy coat 🙂 Some serious change is required.

When you bring a competition horse back into work after a period of rest, you go through three initial steps:

    • assess the current condition of the horse: fitness, muscle tone, suppleness


  • estimate time to first competition (how much work is required before we’ll be ready for our first event)



  • identify target competition (the first ‘serious’ event we want to aim for)



1. Assess current condition:

The only way to do this is to get on and ride, discover what works and what doesn’t. Despite the long rest, Harry went better than expected. Completely unfit, (clipping scheduled for the weekend, goodbye hairy coat), muscle tone not great but not bad, and suppleness is still there (thanks to his breeding – part Lusitano, the breed of horse traditionally used in Spanish bull fights and they are naturally very agile)

2. Estimate time to first competition

Thanks to his suppleness, we’ll get to a competition earlier than with other horses as long as we do some basic fitness and muscle work. Provided he’s ridden regularly, we’ll be ready to start competing in about 4 weeks (normally it would be 6 – 8 weeks)

3. Identify target competition

Here in the UK, we have a Winter National Championships held in April, and he would be ready by then… unfortunately, qualifiers run from October through to 2nd week in March. I should have started 6 weeks ago because we aren’t going to be ready in time for the last round of qualifiers. In a word – tough. The Winter champs are out, so we’ll focus on preparing for the Summer season. The first major outdoor show we usually go to is Leicester County show, held over May Bank Holiday, so that’s our target. Our stretch target will be the Royal Windsor Horse Show, held 2 weeks later with an annual championship class that we will be eligible for this year… and it’s held in Queenie’s back garden, – it’s a cool show to be at. Our ‘quick win’ target will be to compete in the warm-up classes at the Winter champs, that don’t require pre-qualification.

And there’s a fourth step that spans all three – me. As the rider, I also have to commit to change – I’m not much use to Harry if I’m unfit and slopping about the saddle like a sack of potatoes, and I’ve got to be prepared to put in the time and effort riding him – he won’t put his own saddle on and exercise himself in the arena.

So what does all that have to do with I.T. projects? … 🙂

1. Assess current system

Too often, organisations fail to assess what they’ve already got before implementing new systems. I highlighted this in ‘why is KM so difficult?‘. You need to know what you are working with – existing solutions should be examined and a decision made regarding their role in the new system. Understand what those existing solutions actually do – what works and what doesn’t. And don’t guess, don’t make assumptions – find out for real (example: I.T. dept “our people just use Office for basic stuff, word processing and the like”, Accountant: “yup, we pretty much run the business off these 14 spreadsheets…”, Me (on investigation) “you mean, those 14 spreadsheets linked to 100+ other spreadsheets and a handful of databases?”)

Just like with horses, you can’t build an accurate plan if you don’t truly know your starting point.

2. Estimate time to first milestone

How long will it take to provide something (anything) for the business to try out? I often get strange looks from I.T folk when I ask this one. The usual plan is to spend an inordinate amount of time designing the taxonomy, process and other gumpf, and then launch the perfect, tightly controlled, system upon the end-users, perhaps with a small managed pilot along the way. And wonder why it gets ignored as people carry on doing their work as normal…

If there’s one thing I’ve learned over the years with knowledge-based systems, they rarely get used the way you anticipated. You need to test them live, in their real working environment, and then tweak based on the results. With horses, you don’t really know how well the training plan is going until you’ve completed your first competition. Sure, you can hire out an arena and school round a set of show jumps, but that’s just a pilot test. The first competition is the real milestone – seeing how we perform under show conditions. The same is true for many I.T. projects – the sooner you get a live test, the better. A pilot is just a practice run, it’s not real.

3. Identify the target achievement

Most I.T. projects have an end goal, a vision to be achieved (and hopefully a benefits case to prove some return on investment). That’s all fine and dandy, but knowledge-based projects are not 5-minute affairs. They take time to bed-in and become a seamless part of the organisation. A target set 2 years in the distance will seem a long way away at the start of the project. I’ve got an idea where I want Harry to be competing in 2 years time, but to get there I need to focus on the here and now. The same can apply to I.T. projects – identify a strong short-term target to focus your efforts and demonstrate significant value to the business at the earliest possible opportunity. But don’t get all giddy and excited about ‘ideal’ targets. Be tough, focus on what can be done, not what you’d like to do. (Just like I have to accept I’ve left it too late to qualify for this year’s Winter champs.)

Once you’ve got your first big target and your initial milestone, look for the quick-wins – goals that can be achieved along the journey that serve as regular check points that the strategy is on track. Quick wins help maintain momentum and commitment to the changes being introduced – they also provide instant feedback that can help adjust the plan if needed.

And as for that fourth step – I.T. projects cannot be successful on their own, they need the organisation to be fully committed to the change being introduced. Without support across the organisation, any new knowledge-based system is at high risk of failure. Just like with horses – whilst it is the horse that has to do the jumping, the rider has to commit to presenting the horse at the right jump, at the right pace, and at the right time. It’s a partnership.

See, I keep telling people that working with I.T is just like working with horses 🙂

…and then it all goes horribly wrong