How to predict the future: pitfalls to avoid

If we were to transport ourselves back a century to 1909, would we have been able to predict what would happen over the next century? Highly unlikely! Yet, someone with a good grasp of the world and a strong sense of adventure might have predicted some of the trends: the transportation and communication would get faster and cheaper, that the world would continue to urbanise, globalise and move to service based economies, that wars would continue and get more violent and destructive, and that medicine would help us live longer, healthier lives. All of these should have been fairly obvious.

As I suggested in my previous post on this blog, if we want to predict the future effectively, we need to steer clear of attempting specific predictions, and look instead for trends. In other words, using travel as an analogy, the art of predicting the future is more about a focus on direction and speed, rather than looking for specific milestones along the way.

In the next post, I'll outline some specific techniques that can become your habits as you develop your ability to track trends properly. But now, I need to warn you of some common mistakes people make when looking at trends.

We must be careful to avoid some significant pitfalls. For all those of you who studied statistics or logic at university, this is your moment to have flashbacks. I obviously can't reproduce the full textbook of possible errors you might make in a short blog post, but let me try and point out two or three of most common pitfalls you should avoid when tracking trends.

Correlation vs causation

Probably the biggest mistake made in trend tracking is to confuse correlation with causation. Correlation is when two things - possibly unrelated and independent - happen at the same time. Causation is when one thing happens and then causes another thing to happen. Too many people see two things happening, and assume that one causes the other. Let me illustrate this with some examples.

Research recently showed in the UK that men live longer if they marry women 20 years younger than them. The research suggested that the act of marrying a younger woman causes the man to live longer - and there was all sorts of the typically British "nudge, nudge, wink, wink" sexual innuendo in the report. Other research has shown that people also live longer in certain suburbs and cities than in other areas of the world. It is assumed that certain parts of the world are better for your health than others.

While there may be some truth in these claims, it is much more likely that the cause of the longer life is the wealth of the individuals in question. Wealthy men can "afford" (or are more attractive to) younger women, and wealthy people typically live in proximity to each other in "wealthy suburbs". Because they are rich, they can afford better healthcare, have better access to medicines and healthy diets, and therefore are likely to live longer.

There is a correlation between long life and marrying younger women or living in certain suburbs. There is a causative link, however, between wealth and long life.

Another example might bring this closer to home for companies in the current economic downturn. Two years ago or so, many HR departments were working furiously on retention programmes. In 2009, many of those same HR people were bragging about how brilliant their retention rates are now. The assumption they have made is that the retention programmes and plans they put in place had the desired effect, and that they deserve the credit for the improvement in staff turnover rates.

When the data is actually examined, however, at many organisations workers feel no happier, no more engaged, and are actually more motivated to leave their current role sometime soon. What changed (as I am sure you noticed) was the economy. With unemployment rising every where, and daily news stories about recession - even a Depression - the current workforce as a whole is less likely to want to leave.

In this case fear is the causation. HR strategies are just a correlation. A quick test of this statement would be to analyse the turnover of your 10% of performers and see if this has reduced in line with the rest of your staff. My observation is that your top talent is still almost as mobile as ever - thus telling you that your strategy is not as good as you thought it was.

Fluke vs planning

The second issue is related to the issue of correlation vs causation. You might not find this one in your stats textbooks as I have stated it (it might be more technically defined as the problem of medians and outliers), but I think it's hugely important.

Let's say you were asked to guess the height of a random person. Go on - guess how tall I am? I can safely assume that your guess is within 1m of the actual answer (I am 6'3'', or 192cm tall to be precise). The reason you were within 1m of being correct is that the range of possibilities is quite narrowly defined (how many adults do you know who are over 3m or under 92cms tall?).

Now try to guess how many copies of my books have been sold. This is tougher, is it not? Since I am not a world-famous author, you'd be right to guess on the lower side, but the range of possible answers is from three (I have three books, and you can be sure at least my mother would buy a copy of each) to multiple millions. (The actual answer is somewhere around 40 000.) You'd be much luckier to get close to the actual number in this example.

When tracking trends, most of the issues we're watching fall into the latter category: there are many, many possible futures and outcomes, and most of them have a wide range of possibilities. Even when we look backwards, we might fool ourselves into thinking that we've spotted an obvious trend, when in fact, we've simply spotted something that happened by chance.

A good example of this might be found in Jim Collins' work on "Good to Great". I've lost count of the number of conferences I've spoken at that have used this phrase as their theme, and how many companies have used it as a mantra. They've dipped into Collins' book, and cherry-picked some (never all) of the principles he outlines ("Level 5 Leadership", "Get the right people on the bus", "The Hedgehog principle" and others). And now they're confident in their ability to go from good to great.

Yet, recent research into how the 10 companies he researched have fared during the last year's economic downturn should be sobering. You can read more detail and see a graph here , but the sad reality is that two of them went bust, and the others have performed worse (on average) than the badly performing S&P 500 index on the NYSE! Good to great to gone?

In some analyses I have seen, there is a strong argument that these companies were not in fact "great", but rather they were lucky. They were not necessarily headed by brilliant leaders, just lucky leaders. This is a form of the causality issue I spoke about above, but it's even more dangerous, because we actually define the wrong thing as the cause of the outcome, and therefore end up looking at the wrong trend altogether.

Recommended reading

• The Halo Effect, Phil Rosenzweig
• The Black Swan, Nassim Nicholas Taleb
• False Prophets, James Hooper