Laboratory conditions don’t exist Even when we don’t rely on our own prejudices, belief systems and confidence levels and instead do some research into what may be the best course of action, we can still come undone by the environment we choose to test in. Entire industries exist to help mitigate the mistakes we may make in our endeavours. Researchers, social scientists and strategists of all sorts test hypotheses, conduct double blind experiments and enlist carefully selected polling of ‘typical’ subjects, producing reams of data … even big data (the corporate world’s new security blanket). And yet, failure is everywhere. We have often been wary of the true intentions of much research and testing, suggesting (perhaps unfairly) that this work is largely used as ‘screw-up insurance’ — in other words, research conducted not to inform, but as a defence should things go horribly awry. An employee or consultant can hardly be held responsible for failure if the research suggested success was a more likely outcome. However, even when the aspirations and the participants involved in the research are noble and rigorous, errors still persist. Part of the reason for this is the choice of environment in which research is conducted and the margins for error agreed upon. So much of the research people do isn’t conducted in the real world and the artificiality of the environments we create can’t help but skew the results. For example, if you ask someone about their political ideals in a polling questionnaire, they are likely to want to appear more caring, more intelligent and more interested than they may a ctually be. As a result, a lot of research suffers from much of the same over-confidence in its results as our own best estimates. To be fair, big data has started to go a long way to improving this process, given its real-world sourcing, although, like all data, big data is only as powerful as its interpretation and application. Another possible solution lies in a more scientific rather than corporate view of research; that is, research that’s designed to generate information, not conclusions. In other words, rather than looking only to prove a hypothesis, we should also use research as a way of identifying the threats to our hypotheses and the conditions under which this proof may come undone. So instead of focusing on an outcome, we should be focused on generating outcomes.