Be careful reading one parcel as a rule. Two successes build a false model.

My rule, after years of this: if it cannot survive being written down with the numbers attached, it is marketing.

Sleep first. Bad night, bad day, more symptoms. Fix that before changing anything else.

Three questions kill most of the confidence in a thread like this: sample size, blinding, who paid.

I would stop trying to convey how bad it is and start saying what you want them to do.

I would take the narrow conclusion. Broad ones are more interesting and less true.

Split orders. Two small parcels beat one large one for reasons that have nothing to do with price.

The last ten percent of a taper takes as long as the first ninety. Nobody says so until you are in it.