Experimental marketing is probably cheaper than failing