For example, let's say I've tested at 23andMe, and there's a single Native American segment on chromosome 3. I take the raw data to GEDmatch and use SEVERAL calculators (my favorite is Dodecad World9) to determine whether the same segment is identified on chromosome 3. Identifying the same segment using two or more calculators, which typically use slightly different algorithms and reference populations, vastly improves my confidence in the characterization of that segment as being Native American.
I never use just percentages anywhere. I always visualize the segments on the chromosomes.
[One other trick - look for the segment in multiple generations. For example, is the segment there in a previous or later generation? Is the segment larger in a previous generation? If you see it in the same place in multiple generations, that again increases my confidence.]