Hehe, yeah it's funny there must be quite a few members who never logged out and saw the new landing page
The "kanji mnemonics" picture is trying to convey how mnemonics work so it's fine if the stroke order is not exact, as the goal is to introduce the concept to visitors who may have not heard of RTK. At this stage the stroke order is something I wouldn't even mention on the landing page.
What I was wary of is to give the impression that the site actually features this color coded system. It's pretty cool... but that would be quite complicated to do, plus lots of data for it. And then you'd still have to demonstrate that color coded components in the kanji make a significant difference to the memorizing anyway (which I doubt).
In fact I experimented with just coloring primitives in the stories. I closed [this github issue](https://github.com/fabd/kanji-koohii/issues/68
) because I wasn't satisfied with the solution but dicussion is still open!
Another issue to be totally honest is that I think the rate of "conversion" is pretty low. I think people who find their way to the page expect something that really takes them by the hand. I can tell because of 20-35 ish sign ups per day I look at creation date vs last login date and you can see 80% of accounts created say, a week ago, haven't signed in since. Still, that may be relatively normal across web apps.. there's always going to be sign ups where people check out a service, and decide it's not what they were looking for.
Here the main improvements required imho is Improve the flow for new users
(github issue). I think that first experience for new users is quite bad. If I can guide users on their first steps chances are the "conversion" of sign ups would increase.