Serial numbers intelligence

Fred Cisin cisin at xenosoft.com
Sun Jan 31 17:21:36 CST 2021


> > I'm curious to what degree people have used serial number
> > intelligence gathering and countermeasures in the industry. Like
> > were/are there market research firms that would go to Fry's and
> > record numbers off of boxes to try to extrapolate sales for things
> > like printer consumables, and whether companies like HP ever took
> > measures to try to obfuscate the potential information content of
> > their product serial numbers.

On Sun, 31 Jan 2021, Dennis Boone via cctalk wrote:
> http://www.guardian.co.uk/world/2006/jul/20/secondworldwar.tvandradio

A minor glitch, common to MOST studies is the assumption that the sample 
is representative of the whole.

But, consider that the tanks were presumably put into service immediately, 
as soon as they could leave the factory.

Therefore, #20 was in use for a while longer before #92.  Unstated was 
whether they had any idea how long that had been.
It is likely that the tanks that were captured were probably captured 
because they broke down, or were no longer working as well as the fresh 
ones.  That would tend to bias the sample as tending to be the oldest.
Therefore, it would be likely (not a certainty) that the newest ones, 
being in better condition, and POSSIBLY including some improvements were 
more likely to have been able to get away.  That would call for detailed 
comparison of #20 with #92, looking for any differences, AND looking to 
see how well they had been maintained.  (Loss of efficiency over time can 
be a function of maintenance)
Admittedly,within their tiny sample, the numbers are fairly well 
distributed, although there is a definite increasing gap between numbers 
between.

The formula assumes that production is at a constant rate, rather than 
speeding up over time (early production is usually not as fast as later), 
or slowing down with parts shortages.

Their sample was too small to be completely trusted as being 
representative.  Realistically, there were numerous factors that could 
make such a tiny sample unrepresentative.  And therefore, the estimate 
of 109 should have been plus or minus a LARGE number.

It is not clear nor stated how many had been captured by the time that 
they made the GUESS of 246 for an actual number of 245.
Perhaps they had captured hundreds by then, and could more reliably 
estimate.  If they had still captured only a few, then their guess was 
extraordinarily lucky.  The reliability of the calculation increases 
dramatically as the sample size grows.



More information about the cctalk mailing list