Some thoughts on rank ranking. If there are two watch faces A and B, with the same design, their click-to-download conversion rates are the same.
However, because watch face A is ranked higher at the beginning, it gets more exposure. As a result, its number of comments and downloads are far ahead, which makes it get a higher ranking, and the cycle repeats. The other watch face disappears forever.
My idea is to give watch faces with different rankings a coefficient. To balance their irreversible advantages.
For example, the coefficients may need to be adjusted，1-3 may be multiplied by a coefficient of 0.7, 4-6 0.8, 7-9 1,0. Those watch faces that are not seen on the homepage may get a bonus of 1.5 to make up for the disadvantage that no one sees them. And the truly excellent and popular watch faces, which are usually downloaded several times more than other watch faces, are not easily affected. And those watch faces with similar popularity are easy to fall when they are in the first place. When they fall to 4-6, or later, the penalty mechanism is reduced. They have a chance to get ahead again, thus dynamically balancing. Make the ranking easier to refresh.
Those with several times the download volume will still firmly occupy the first position. Or when the user has downloaded the watch face, or the homepage pushes it to the user for a long time but the user does not download it, it will automatically be pushed to the back. Otherwise, the user's homepage will never change.
And Garmin's recommendation algorithm is too old.They do not have an effective algorithm like some video websites. They will randomly push to users, and then look at the conversion rate, giving higher traffic to videos with high conversion rates. Instead of simply recommending videos with high views.