The engine is built to possibly take 100+ ratings to hone in. Maybe ok results by 30. Maybe shit results by 200. The game system is too complex to be predictable in this way. If you expect results by any amount of ratings or level prepare for disappointment or to be fooled by good luck.
When you read someone else gets what they want by X ratings or Y level or whatever, you should completely dismiss it as anecdotal. The game is not consistent in how it behaves. A bad seed on week 0 could set you weeks behind.
One technique to help would be to customize at least a few girls with the save editor in week 1, and rate them 10/10, or load in girls from others' photos using the web editor, and again, rate those 10/10. With a good starting week of a few good genes you are way better off from there on.
Few pointers i found useful:
1) You need to have rating profile so that your ratings can be consistend otherwise you mjight exclude elements that you want or viceversa.
2) Even with complete rating profile there are still variables that are not accounted for towards rating (example in noses might not account for how round or sharp the tip are ) and that might bring forth ugly outcomes anyways. hover over the part you are rating to see which specific parts are rated for reference.
3) Same outcome for rating a part can be achieved with more complex ways. I played with the cheat menu to reshape a girl and saw that by changing sub parts i could achieve the same rating result as long as i hit the mass'?' of the targeted pointer for rating. That means that the rating system is simpler than the actual modular capabilities and the complete creation in aesthetics can only achieved by pure luck..
4) while giving 10 in some part can make it come close more often, giving 0 doesn't seam to work as godd as 1 works so far. also it's faster to straight out change the decision if the girl is suited for modeling to either yes or no. dont use "some" or n"not so much" at all for better/faster models generating pool.
If you hit home with something good choose "a lot".
5) No matter how many groups you have use 1 profile for all until you finally manage to tune 1 on your desired standards, Keep the one that managed to develop best and then turn the rest groups in the next profile you need to tune.
PS: your mentioned way is the hidden 6th pointer but less satisfactory since steals the whole joy and frustruation
PS2: Don't bite me if i have speling errors... I'm too lazy to check back :333
