This is yet another example of the abuse of numbers in reporting. The relative rankings of various companies tell you nothing about customer's satisfaction with those companies since few people will have rated more than one insurance provider in a given year. All three of the companies could have improved their satisfaction scores, but as long as some other company improved more, their relative ranks would have slipped by one spot. Nor is a 4 point change on a 1000 point scale particularly meaningful.
Sometimes I want to mandate statistics classes in elementary school.