ABSTRACT The two authors’ company (IGS) is a contractor that undertakes in situ testing and direct push sampling by a number of methods. One of their tests undertaken repeatedly, approximately 40% of the company’s business, is cone penetration testing (CPT), most commonly using piezocones. The CPT cones they use are good quality commercial units supplied by the Dutch company Geomil. Cone capacities range from 3MPa to 100MPa, and both compression-type cones and subtraction-type cones are used. IGS undertakes their own in-house calibrations on all new cones, using calibrated load cells, and a combination of dead weights and hydraulic load application, and using each cone’s true (measured) tip and sleeve dimensions. This calibration process is then reiterated almost continuously during a cone’s life, on an unusually frequent basis. This is part of the IGS company culture and typically involves cones being (a) purpose-calibrated at the beginning of every job, then (b) recalibrated again at the end of the job (or after only a few days of use on longer-duration projects). At every calibration, reference values or zero offsets of each load sensor (tip qc, sleeve fs, and pore pressure U) are recorded, and the slope of the applied-load-vs-cone-readout for each of these sensors is measured and adjusted to give as close as reasonably possible 100% accurate output; all of this is recorded for each cone. Thus the authors have a significant database of reference-value-drift that can be compared to slope adjustments that have been needed to achieve the desired cone output accuracy. This paper reports the data for a few CPT cones of each type used by the company, declared to be typical, and suggests as a conclusion that relying on reference value drift to inform on CPT correctness might be convenient but might not be a reliable procedure in practice.