Empirically, they're usually not nearly as useful as people think. People are on average unreliable observers of their own experiences.

Much better than individual anecdotes are collections of anecdotes that are searchable and somewhat categorized. A lot of old school drug forums provide this.

Much better still is systematically varying data and taking statistics to separate signal from noise.