aakk9999 - 10:20 pm on Dec 12, 2013 (gmt 0)
They look for patterns. If your site matches a pattern of link buying or selling site, it will be probably labelled as such.
So in the above example, if a link purchase is one off, and the seller sells one link only to you (does not sell links to others), you could almost certainly get away with it. But if this seller sells links to many sites, or you are buying links from many sites, this shows the footprint.
Unfortunately, some sites may show footprint even when they are not buying or selling links - they are collateral damage because otherwise Google cannot scale.
To try to draw some parallels:
I worked with ePOS (point of sale) and this was the way Loss Prevention in stores/shops worked. The averages and an acceptable deviation from these average were calculated for various metrics. And if a checkout operator's metrics were outside this acceptable corridor, they got investigated.
For example, one of metrics may be an average number and total value of returned items supervisor performed in a day/week. So if someone was doing fake returns and pocketing money, this would get highlighted and this supervisor would get investigated or a detective would be sent in the store.
Staff never knew where the limits were - they were kept secret (and they were slightly different depending on a store type).
So when someone starts to cheat, at first bogus returns are still "hidden" because at first they are very careful. They do one bogus return and pocket the money. Nobody noticed so after a few days they do another. And another. And they get braver, the amounts get larger and the number of bogus returns get higher - it is a human nature. Until they break the limit. And this limit breach is not one-off, it soon happens day after day, week after week over the x period.
So the ePOS system highlights this pattern and the retailer investigates. And in most cases supervisors were found guilty. When found guilty - this is publicised within the company - to serve as deterrent to others.
See the similarity?
But Google cannot afford to investigate - it does not scale. They penalise. (They may take a closer look if there is a reconsideration request. Eventually, taking a closer look over many websites over the time may perhaps result in some adjustments to the metric as they *will* learn something from the write-ups of webmasters' many reconsideration requests.)