Here come the lawyers

If the lawyers have their way, you won't keep your measurement and calibration data as long as you would prefer. This was one of the points brought up at this year's Measurement Science Conference (February 27–March 3, Anaheim, CA) during a discussion about a new version of NCSLI Recommended Practice 1 (Ref.1). The next version of the document will include a discussion about data retention. Why? Because some lawyers may force the destruction of calibration data that proves product quality in an attempt to dismiss any chance of liability, said Don Wyatt, of Diversified Data Systems, and Charles Motzko, a test consultant.

Calibration standards (on rack) with proven stability may have long cycles between their calibrations referenced to a Josephson junction array (on left). Courtesy of Fluke. Click to read the Dilbert cartoon that ran with the print version of this article (scroll down after clicking).

"I've seen cases where an attorney enters a calibration lab and tells engineers that they must discard data after seven, five, or even three years," added Wyatt. "Lawyers don't understand calibration nor do they understand which data is important to keep, so they advise clients to discard all data as soon as legally possible."

Unfortunately, calibration data is the only means by which you can prove that your company's measurements are traceable to a national standards lab such as NIST. If you discard calibration and traceability data, the measurements used to design and manufacture a product become meaningless.

The problem is even more disheartening because some corporate calibration standards have data histories reaching back 40 or 50 years. Because historical data can prove that a calibration standard is stable over long periods, the standard may have five-year calibration cycles. Government guidelines often specify data-retention periods ranging from three to seven years, which may be short of even one calibration cycle (Ref. 2).

Another issue is measurement uncertainty. Every measurement is an estimate of an actual value, with an associated uncertainty and a probability that the actual value is within the uncertainty limits. Although small, a probability always exists that a measurement falls outside acceptable limits.

Lawyers, though, may not think in terms of uncertainty and probabilities, and they use that small probability that a measurement was out of tolerance to cast doubt. So, if no measurement records exist, prosecutors can't prove that a company manufactured a defective product. But neither can they prove that your product was built correctly or your engineering data is accurate.