In the RIS News article “Rethinking Retail Traffic Measurement” John explains the difference between accuracy and variance. The essence of his piece is that the traffic counting industry has a habit of measuring variance in counts, specifically the difference between counts of who's coming in and who's going out. The idea is that over time the number of people entering and leaving should be the same, and the closer a counting system comes to that, the better.
However, John goes on to point out, the industry has a bad habit of referring to this metric as accuracy. Accuracy is a completely different thing. Accuracy is the proximity of the machine-counted metrics to the genuine traffic that went through the store (after all, there is a real number of people who went through the store, and that's what we're seeking to determine). Variance is simply of an indicator of whether or not the system behaves consistently. It might be highly inaccurate, but so long as it's consistently inaccurate, the variance will be very little.
(Above, John provides an illustration that shows how accuracy can be poor and variance can still be very low.)
But that's just my capsule summary. To get the full explaination, make sure you read the article.