Adhering to research standards would not make cases such as the following happen: 

“My own results are — literally and figuratively — all over the map. I’m German with some companies, not German at all with another. My personal reported Scandinavian ancestry ranges from a low of 0% — literally none — with one company all the way up to 66.9% with another”

If AncestryDNA adhered the minimum acceptable sample rules, random sampling, and deliberate avoidance of selection and confirmation bias for starters, Ancestry could be one of those companies which would have more accurate results.