TER scoring is essentially a survey.
Now, in a professional survey - political polling, for example - sophisticated techniques are used to statistically attenuate individual survey replies so you'll get more "usable" aggregate results; sadly, TER doesn't use these techniques.
My average response for a "good" session, looking back, it 6.92 - let's call it "7". So, I'm a "7" kind of guy - assuming most hobbyists enjoy the hobby, and safely assuming that I'm at least somewhat like most hobbyists, you can assume if I rate a girl a "7" she's worth seeing.
For another reviewer, that metric might be an "8", or even a "9".
When I'm researching a girl in TER, I have a quick look at all her reviewers: if they have one review, I disregard their review; if they've submitted pairs of reviews monthly, I disregard their reviews (they're posting fakes for membership); the remainder I look at their scoring trends to see if it leans "hi" or "lo", and then amend the review of my girl of interest. No, I don't use a spreadsheet or anything - it's just a quick mental assessment of what's going on statistically so I can make a (more) informed selection.
(Over time, I'll become aware of certain reviewers predilections and weather they're similar or different from mine - if a reviewer likes "thick" girls, for example, I'll consider his high reviews for appearance to be a negative, so I might replace "9" with, say, "4".)
TER would be immensely more useful if it would implement two tools: (1) smoothing scoring results with analytics and (2) allowing users to "tag" other reviewers as (+) or (-) if those reviewers have tastes very similar to (+) or different from (-) mine; this latter function would improve the analytics, as well.
So if JoeThickLvr reviews a girl with a giant ass and huge tits, he's going to give her a "9" for appearance - but if I've tagged JoeThickLvr as a (-) reviewer (because I like skinny model-types), TER will "correct" the scoring results I see per my tastes.
...
I need to hire a programmer.