Can we (yet) rely on computer algorithms and machine-learning to make sense of that most unpredictable substance in the universe – human behaviour?
There’s a great article in the Financial Times describing the use of data in football. The focus is on the Everton manager David Moyes who is the likely candidate to replace Sir Alex Ferguson at Manchester United. But for this post, the interest is in how Moyes used data to improve performance.
The background into the club highlighted how it is one of the most under-funded football clubs in the Premier league yet consistently out-performs wealthier rivals. Data is used in a range of different ways, but the value has come from when it is personalised to the individual situation.
Rather than have a signature style of game plan like most football managers, Moyes analyses data to determine the specific tactics for each game. If the data isn’t available, the in-house team gathers it from video coverage of football matches. Capturing statistics about the performance of each player and the movement of the ball. Identifying patterns but also putting the statistics within context to help predict behaviours for the next game.
The same approach is applied to talent recruitment. The largest signing at Everton to date has been for Belgian player Marouane Fellaini. Everton needed a specific style of player to replace one who had been sold on. Fellaini was not well known at the time, short of being sent-off early during the 2008 Olympics. From the article:
There were few match stats for Fellaini, because there was then no data available for Belgian league matches. And so Everton watched videos of him to compile their own stats, using key performance indicators that seemed relevant.
The tactics worked and, as a result of his performances since joining Everton, he is expected to be transferred this summer for a far bigger fee than Everton paid. It’s a great article – link at the end of this post to read in full.
Using data to improve performance is not a new science. If you want to read up on the subject, there are plenty of books to choose from. One of my preferred is ‘Competing on Analytics’. (Yep, the title of this post is not the most original.) But I am surprised how little analytics are used to innovate within business, compared to sport. The difference is in the level of personalisation. Within sport, the use of data is highly contextual to the individual scenario to maximise potential within that moment. In business, it tends to be more generalised to look at the ‘big picture’ but often loses value in the process. Much like attempts to capture and re-use knowledge.
General predictions tend to lead to general results (read ‘average’). To compete requires raising performance above average and that means doing something differently to everyone else. Personalisation enables that difference.
- Everton: how the blues made good – by Simon Kuper, Financial Times, 3 May 2013 (may require subscription to access)
- Competing on Analytics – by Thomas H. Davenport and Jeanne G. Harris, published 2007 by Harvard Business School Press
Side note: the image is me 🙂 a sports-related post is too good an excuse for the occasional horsey pic
Big data is a key digital trend right now. Along with social media, mobile devices and cloud computing, it is seen as a potential disruptor of businesses, if not entire industries.
But data, big or small, is useless if it isn’t applied within a process. And the process is more about people than either information or technology.
Last year, the New York Times covered a story – Armed with Data, Fighting More Than Crime – about how a US city improved performance through applying data analytics. The mayor of the city – Baltimore – applied a method first adopted by the New York City Police Department.
The key principles were:
- Accurate, timely intelligence
- Rapid deployment
- Effective tactics
- Relentless follow-up and assessment
Baltimore did not put in a big complicated technology system to introduce performance management. They used Microsoft Excel and PowerPoint, and the dashboard cost approximately $20,000 to set-up. Instead, the investment was in people – four analysts and an investigator – whose salaries cost about $350,000 per year.
The original article covers the story in far more detail. But amongst the titbits are the details about how the process was implemented:
Every two weeks, each agency head would come to a CitiStat meeting, facing Enright (the investigator), sometimes the Mayor, and the heads of the departments of finance, labor, legal and information technology. Before the meeting, a CitiStat analyst went over the latest data from the agency, pulled out the important stuff and put it in graphic form. The agency director stood at a podium, his senior staff seated behind him, and the information was projected onto the wall… When solutions were discussed, there was no need to schedule a meeting with the city’s solicitors or budget director, because they were in the room. By 5 p.m. that day, the CitiStat analyst had written and circulated a short list of commitments the agency had made or information it needed to provide. “We expected that when you come back in two weeks you will have answers to these questions,” said Enright.
The meeting applies all four principles. They have accurate and timely intelligence. Everyone who could potentially stop change from happening is in the room, enabling both rapid and effective decisions. And there is relentless follow-up to make sure those decisions lead to action, that in turn becomes data to feed into the next decision.
Within 3 years, overtime was reduced by 30% and absenteeism reduced by half within targeted agencies. Within 6 years, it was estimated that the the program had saved the city nearly half a billion dollars and city services had been significantly improved.
It’s an excellent case study in the effective use of analytics and how the process matters more than the tools or the data. And also highlights how important senior leadership is when difficult decisions need to be made. Two non-performing heads of departments were fired early on in the process, signalling just how seriously the initiative was being taken.
Read the original NY Times article for full details – Armed with data, fighting more than crime
- Compressing processes – how productivity is going social and mobile (for A|B analytics example)
67% chose the blue pill*
We so often see summarised data encouraging us to buy or do something, that it is easy to forget to be critical. Critical thinking is essential to help avoid making poor decisions due to misleading ‘facts’.
The image above is an advert that appeared on LinkedIn this week encouraging people to vote in a poll. But the question is clearly biased. What if your favourite feature isn’t listed? An accurate response would be ‘Other’ or ‘None of the above’. But those options aren’t available which means you either a) make up an answer or b) don’t respond. This is a common fault with most surveys. Made up answers are more likely if an incentive is given such as free gift for responding to the survey. Otherwise most people would just go ‘Meh’ and not bother voting. But surveys that cover all options can be just as bad. Because people who are interested in the question are still more likely to respond (and respond accurately) than those who are not. Self-selection to participate in surveys will always lead to biased data and unreliable results. It’s the easiest way to make bad predictions.
So next time you see claims such as “84% of CEOs said that <Insert product name> helped increase sales”, check the underlying data for bias (and size! How does the sample size compare to the total population?)
It’s one reason why A|B testing is moving beyond web site design and being used in other industries where the practice can be applied – improving systems by leveraging feedback captured from actual uses rather than what people say they would do. An example is included in the recent presentation: How digital trends are compressing processes.
* The number was made up 🙂 Don’t let it influence a future choice
Featured image is from Flickr: Red pill or Blue pill by Paul L. Dineen
Social Media Analytics are a big area of growth… but beware the wrong focus. A lot of analytics can be like looking in the rear-view mirror whilst driving. Tells you where you’ve just been, allows you to make some course adjustments safely but you really want to keep most of your focus on the road ahead.
Instead, think more in terms of awareness. Real-time analytics can be hugely beneficial if it makes you aware of a conversation taking place that you should be involved in. That means monitoring the various channels, currently including Facebook, Twitter, LinkedIn, Pinterest, Google+… and having a social media strategy that enables you to respond effectively to whatever the monitoring brings up. By all means have your regular reviews, rank people’s contributions if you must. But don’t let the bells and whistles distract you from what matters.
The Economist has an article ‘Don’t lie to me Argentina‘ explaining why they are removing a figure from their indicators page:
Since 2007 Argentina’s government has published inflation figures that almost nobody believes. These show prices as having risen by between 5% and 11% a year. Independent economists, provincial statistical offices and surveys of inflation expectations have all put the rate at more than double the official number…
What seems to have started as a desire to avoid bad headlines in a country with a history of hyperinflation has led to the debasement of INDEC, once one of Latin America’s best statistical offices…
We see no prospect of a speedy return to credible numbers. From this week, we have decided to drop INDEC’s figures entirely.
Whilst we often talk about how statistics can always provide the answers people are looking for (hence the popular quote used as the title for this post), there is another angle to consider – are the underlying numbers telling the truth? It is a critical question when decisions are based on increasingly complex calculations that are then converted into summary data visualisations to assist decision-making. Was the original source data automatically scraped from systems or keyed in by people? Just how balanced is that scorecard..?
One of the current hot trends on the Internet is the emergence of ‘Big Data’ – being able to scrape massive quantities of information automatically generated, such as the search and surfing habits of everyone who ever logged into Facebook… and then analysing for patterns. One of the potential attractions is eliminating human error – or influence (in terms of truthfulness) – over the underlying data sources. Doesn’t solve the challenge of influence through gaming of the system but that’s perhaps a post for another day.
If you are interested in the use and abuse of statistics, there’s an excellent short book that walks through historical examples of where statistics simply don’t work – The Tyranny of Numbers: Why counting won’t make us happy, by David Boyle – Click Here for an old book review I wrote. And naturally, the book is listed on Amazon.