Question your data
I remember a lesson in questioning your preconceptions. As a junior hydrogeologist, I visited Sua Pan in Botswana. A mining operation there pumped potassium-rich brine from boreholes drilled beneath the salt crust. The brine was allowed to flow through evaporation ponds to obtain a solid product. The problem (and the reason for our visit) was that the mass of salt recovered at the end of the process did not match the simple calculation: volume of brine x concentration of brine. The Client suspected leaks in the evaporation ponds and had commissioned several investigations before our arrival.
The hydrogeology of the ponds, as indicated from the reports we reviewed, made significant leakage rather unlikely. We suspected there was an inconsistency in the monitoring data of brine volume and brine concentration. We offered to review the data, but the Client refused. Years later, I had the opportunity of discussing the matter with one of the Client's former hydrogeologists. It turned our our suspicions were correct. The brine volume had been estimated by multiplying the theoretical capacity of the borehole pumps by the number of hours they were in operation. The pan hydrogeology and borehole construction meant that the pumps did not actually deliver to their full capacity. So the estimated borehole production was significantly overestimated.
This experience emphasised to me the importance of understanding the source of your data. Too often, we assume that columns of numbers represent clear, defined values. We should not forget that every measurement is an estimate of a value and that estimate is subject to possible distortions: operators misreading gauges, poorly-calibrated instruments, assumptions regarding borehole pumping volumes, and an infinite number of others.
Here is the take-home: Don't take the reliability of your data for granted.