Reading Your Own Incident Data

The short version

Fire departments often have more data than they realise. CAD alone can contain years of incident locations, unit assignments, timestamps, workload, response patterns and resource use. The first step in becoming a data-informed organisation is not buying an analytics platform — it is learning how to ask useful questions of the information you already have. Do not begin with “what can we graph?” Begin with “what decision are we trying to make?” And be careful with averages: a mean response time can look acceptable while a significant number of incidents take far longer. The mean is the total divided by the number of incidents and is pulled about by unusual values; the median is the middle observation, with half above and half below; the 90th percentile is the point at or below which nine calls in ten fall. The distribution is how the observations are spread, and it is usually where the operation actually shows up. The same caution applies to timestamps: if you do not know where one comes from, who creates it, or whether the process changed, be careful what you conclude from it.