Working in the insurance industry there are numerous places where data can be misinterpreted, fortunately it doesn't happen often.
On one occassion, campaign code analysis resulted in a business partner targeting a specific marketing campaign with much greater success than expected, providing our Contact Centre with 150% of the predicted call volumes. As a result, the longer waiting times made that campaign's performance look poor on the following month because the sales rates were lower than usual for that campaign code (customers who have been waiting on hold can be a tough sale). The fact of the matter was that the marketting campaign was hugely successful delivering many potential clients but some of the statistics looked bad a result of the challenges involved in handling call volume.
As an industry though we are accustomed to interpreting the granular detail of data and looking past the face of a printed stats report. This particular marketing campaign is still a great source of business for us, it's just used more carefully these days to ensure our call volumes are managable and we can provide a good service to all of our customers.
Either that or the results of our mistakes are well hidden. When a premium is calculated on 50+ different rules, all driven by statistics and data interpretation, the impact 1 mistake will have on a quoted premium is generally going to be fairly small. Or if an edge case is triggered and we quote someone £2000 instead of £200, well that won't happen to many customers so again the overall impact to the business will be small.
We often have big challenges, such as EU law making us take gender out of our premium calculation - this was probably the single biggest risk factor involved in insurance calculations (especially life, medical and motor) as statistically it has one of the biggest impact on claim costs and frequency. The fact that the entire UK insurance market was able to take this out of the calculations without bringing the market to it's knees and ruining companies is testament to how well data is interpreted in this industry.
I guess it's just a matter of practise and training. If a small company asks your regular techy DBA to provide them with in-depth statistics and information analysis to base their business plans on you're going to have more mistakes than when a big company with a team of business intelligence analysts and actuarial statisticians are on the case.
^ Thats me!
01010111011010000110000101110100 01100001 0110001101101111011011010111000001101100011001010111010001100101 01110100011010010110110101100101 011101110110000101110011011101000110010101110010