Hans Rosling, using Gapminder to ‘wow’ audiences with his ability to cross-reference UN and similar data sources
As a practicioner of ‘due diligence’ I consider one of the key skills to be seeing through numbers/data and finding a way to present, in a very succinct manner, a key trend/issue. There are several key elements to this:
- the ability to hypothesise the outcome, and then consequently choose and develop the appropriate data sets and presentation format (without random trial and error, which is too slow);
- analyzing the data to determine key trends, even when the data may not appear to exhibit trends; and
- the communication of the finding in a compelling, interesting and succinct manner.
Preparing “killer charts”
It is often said that “A picture is worth a thousand words” – in this sense, a chart might be worth a thousand data points, or a thousand words of explanation, etc. Graphical analysis, where the presentation is selected in a way that emphasises (but not distorts) an issue/finding, can be the most effective from of communication of due diligence findings.
Key steps that the preparer of the graphical analysis needs to consider are:
- what are the axes that I want to present? how many axes to present? both/all variable? include time on an axis?
- should I ‘zoom’ in on a part of the presentation (by selecting the scale of the axes)?
- should I color code elements of the analysis for greater effect/interpretation?
- which graph/chart type(s) most effectively highlights my presentation? (bar, pie, bubble, line, etc.)
- how many data sets do I want to overlay in the chart (and whether multiple axes help with comparison of different data sets)?
- which data sets should I cross-reference? (cross-referencing different data sets, that you might not have previously put together, can sometimes result in surprising trends and insights)
- do I need to ‘create’ a new data set/metric which most effectively highlights a trend? (see below)
The variety of options means that it is very easy to get lost – many people frequently use the same charts for all analyses or don’t change the scale of the charts as suggested to them by the software they use. On the other hand, using two many different chart types, or the wrong type of chart can also reduce the quality of the presentation.
The skill in preparing analyses, which I have termed “killer charts“, which achieve the goals above, is therefore the ability to see ahead what it is that you are trying to prove/understand – I believe that this comes from experience, and having the right mindset (a mix of inquisitiveness, imagination/creativity). An approach of randomly trying different data analyses, without hypothesising, is also not effective.
Mastering the “killer chart” allows one to effectively communicate trends that others might not immediately see – doing so can often result in a ‘wow’ factor, or people claiming to have ‘epiphanies’.
Creating new data from existing data
Sometimes it is not sufficient to plot different data sets against each other, but to create new data and segment data:
- creating new data can often mean ‘dividing’ different data sets by each other – for example, taking corporate expense items and deriving the cost per employee, cost per hour worked, cost as a % of sales, annual % increase in cost, etc.
- segmenting data to summarize large volumes of data which is distributed over a wide scale – eg, segmenting project performance into quartiles, or ranges of data (eg, 0-$50m; $50-100m, and >$100m) – trends which might not appear obvious for the total data set might then jump out at you. This can even be done by qualitatively banding different products, for example, where quantified data does not exist or is inaccurate (for example, considering market share, but without market research, one might band products into ‘growth sectors’, ‘flat market share’, and ‘declining market share’.
Sometimes the key trends are not always the most obvious from the original data set – data can appear to be without any underlying trends; if this is the case I often wonder whether the data has in fact been appropriately analyzed or segmented.
When management (the owners of the data) tell you that there is no clear trend, and that this is simply how it is, then you should consider whether they really know and understand their data, and therefore also question their management, if it is based on that data.
Remember, it is not necessarily about buying expensive market data, or investing in collecting research data, but often a question of how the data that exists is analyzed.
Expensive analytical tools can help in data manipulation, but I consider that spreadsheet analysis, based on the principles set out above, is sufficient, and easier enough to use/learn, to allow deep and insightful data analysis. While Microsoft Excel is one of the leading commercial spreadsheet tools, Google Spreadsheet (part of ‘Google docs‘) is freely available online (once in Google docs, click ‘Create new’ to reach spreadsheet option). All you need to access Google docs (a ‘cloud computing‘ version of word processor, spreadsheet, etc.) is a free Google-account.
The dangers of statistical analysis
One does of course have to be careful with statistical analysis:
- forming hyptheses, and identifying trends in data can blind one from other issues or trends that have not yet been seen (be cogniscent that the way in which data is presented can distort a reader’s perception of priority/importance of issues/trends); and
- some analyses can be used to purposely mislead (highlighting certain positive/negative trends to drive a certain opinion/outcome) – as they say, there are “lies, damn lies, and statistics“.
Nevertheless, I do not see these as reasons not to do statistical analysis, rather that the preparer/reader needs to be cautious and mindful of these risks.
Hans Rosling – a master of “killer charts” (wikipedia entry for Hans Rosling)
Hans Rosling, a professor of global health at Karolinska Institute in Sweden, is a master at using “killer charts” to communicate a message. His video (at the start of my blog post) on ‘What Stops Population Growth‘ is a good example of this, and one that Bill Gates says that the Bill & Melinda Gates Foundation use in their communications (eg, in answering the common question they face of “If we solve health issues in the developing countries won’t we see a surge in population growth?”).
The web provides incredible opportunities to learn from people who are both experts in their field and great communicators. Hans Rosling, a doctor and researcher who is a professor of global health at Karolinska Institute in Sweden, is a great example.
He was very good at explaining things even though they gave him only 10 minutes this year. I was hoping I’d have a chance to talk to Hans at TED but I did not.
This is impressive stuff, from someone who is no stranger to success or fame.
Hans Rosling uses ‘Gapminder‘ software, which he developed with his son, to show animated’ motion charts (showing two axes of data which are then run to show a development over a third dimension, time). The video at the outset of this blog post shows him demonstrating one of the motion charts. While I think this is indeed powerful software, it is important to note that this is not the only way of presenting data, and in many cases alternative approaches may be more powerful.
The Gapminder software (the charting software) containing also the various sets of data on social and economic trends (most of the data is publicly available, sourced from the UN, IMF, World Bank, etc.) is available to ‘play’ with on the gapminder.com site, under ‘Gapminder World‘ (click on the axes to choose different data sets, or select/deselect different countries, then ‘play’ time development). The data sets are huge and numerous, so there is a good chance that you might find some interesting trends, but I suspect that Hans and his students have already examined the data at length, and already identified some of the most interesting developments (covered by Gapminder’s videos).
Gapminder sold its ‘motion charts’ tool to Google in March 2007, and so the tool can now be accessed as a ‘gadget’ in Google Spreadsheet (which, as I stated above, is free to use). In ‘motion chart’ you can convert your own data-series into a Gapminder-like graph and put it on your web-page or blog.
Alternative presentation formats to graphical analysis
Interestingly, the underlying skills of spotting issues and trends (perhaps using data analysis) can still be very effective even when not communciated graphically, but in words. Sure, it might take a thousand words, but the reader might then visualise the issue better than if the analysis were shown in a chart (along the lines that a novel can often build a greater image and form a deeper impression than a movie of the same story might achieve).
Malcolm Gladwell’s brilliant book ‘Outliers’ achieves this over and over again – he describes, with case studies, various developments in a highly effective and enjoyable manner, mostly using words (and the occasional table/chart).
Physical presentations allow the presenter to convey charm and enthusiasm for the topic with body language and tone and volume of voice, while authors and report writers need to be highly skilled in ‘story telling‘ to make up for the inability to use body language or tone and volume of voice (but this might also highlight why it can be so effective to ‘present’ a report to client, not just deliver it by post/email).
The combination of verbally presenting a chart (as Hans Rosling does so effectively in his videos, or in a report, or book) can be very powerful, so long as the analysis has managed to separate out the trend/issue that you want to highlight.