When we prepare direct marketing, in order to enable calculations of incrementality and return on investment, we often use the principle of a control group - that is a randomly selected sample of the population chosen using exactly the same criteria as used in the main selection. However as targeting improves then target populations get smaller, and then the control group can become statistically insignificant. It is a significant problem.
There are various statistical methods that can help to estimate result reliability more accurately, from simple chi-square tests to complex Fisher analysis. Every day we need to make decisions about balancing the time and effort required for further predictive analysis with the expected impact on statistical result validity and campaign execution timescales.
Sometimes these same issues crop up in far more critical arenas than in the marketing of optional consumer goods. I have suggested that you are more likely to die in a hospital than anywhere else. I still suggest that the marketing of patient choice is over-hyped.
I recently learned of research published in an American medical journal that compared mortality rates in different hospitals. The article looked at very similar operations, and questioned whether those hospitals with a zero mortality rate did better than those with a non-zero mortality rate. Even just judged on that single measure, the subsequent evidence concluded that hospitals with a history of zero mortality subsequently experience mortality rates that are the same or higher than those of other hospitals. Yes, choose a hospital with historically zero mortality and you are more likely to die there.
This is relevant to public health policy in England too. There is regular shouting for more patient choice, particularly by the main political opposition party, but the usefulness of those choices needs to be considered in the light of this kind of statistical evidence. More importantly, you personally may need to make life changing decisions if you ever have to choose a school or hospital.
And a very similar issue affects how we interpret football results :)