Paralysis by Analysis

Ever been faced with a mountain of data and expected to form useful conclusions based on analysis of it? The first time I saw this play out was while working as a summer intern in the stock control office of a major Swiss chemical company. Back then data was usually just dumped onto massive computer print-outs on that continuous-feed paper you no longer see very often. Well, one day a guy was presented with such a foot-high print out of all dye-stuff orders and stock levels in the last month - called the "M25M Report" and asked to interpret. Admittedly he looked kind of grey, or even a bit white sitting at his desk, then he went out for lunch and never returned, just absconded. Talk about being over-faced. His only option of course was manual interrogation, no Lotus 123, or Excel in those days and maybe some people are just put off by a jumble of numbers and seemingly insurmountable tasks - prone to error as well, being manual interpretation.

A few years later another example - this time it was reconciliation of lots of premium data from two separate sources. The staff charged with doing the task every month sat in a closed room for two whole days, taking it in turn to shout out a number from one list, the other then finding it and ticking it off the second list. No excuses with this example though, as they had the data in electronic format and could easily have run a 'find duplicates' analysis in either Excel or Access. Red faces all round when I pointed that one out.

Have things improved with the advent of data now always being available in databases and spreadsheets, with plenty of software tools to assist with interrogation? I have to say things have really moved on at pace and nobody should ever be sitting in a room ticking off figures manually any more. But, the bar has been raised: much more is expected in terms of ability to handle the data and the thoroughness of analysis required. That in a way though is a problem in itself, because with data so easy to both get and analyse, it's the quality of output and conclusions drawn from analysis, that are the real differentiator. Graphing output might help, but too often I just see a graph, without annotations to help highlight what story it's actually telling. Also conclusions can often still be a bit wishy-washy. I prefer conclusions to be stated clearly, backed up with good logic and under-pinned by associated levels of confidence, based upon application of a realistic model and assumption-set. All too often I have to ask for this level of additional rigour to be added though. I hope the millennials coming through the business schools right now don't forget how to use MatLab and other statistical analysis tools once they've graduated and joined the world of work, because we need good data analysis and interpretation skills right now and in the future with Big Data on its way, we'll need them more than ever before.

Howard is a senior Insurance and Risk Management practitioner with over 30 years in the profession, he's a Fellow of the Chartered Insurance Institute, Chartered Insurer title-holder, he also holds an MBA and a Master's Degree in Natural Science from Oxford University. His professional interests span the Corporate Insurance, risk management and financing sectors.