Hi ​@emiller - wonderful question (ok I maybe just weirdly love rank structures ). I totally get your question and can see how this is going wrong.  A common pitfall in rank structures is to create a gradual path up the ranks utilising several or all of the possible data points.  When you do that you can run into some very strange situations and will find that members get stuck (as is the case here).  

The best practice we normally describe is to ‘segment’ the types of member behaviours and to design ranks accordingly.  For example, we typically have an initial set of ranks for ‘lurkers’ i.e. members who have logged-in but haven’t contributed yet, then a set of ranks for those who contribute - these can be designed so that both topics OR replies will get you moving up the ladder.  If you highly value best answers, as we do here, then a great practice is to have the next set of ranks above that for those starting to help others - which can be ONLY 1 best answer to start with, and no other criteria.  What will then happen is that anyone who gets a best answer, regardless of their post or topic count, will jump straight up to that rank. In your example, that could mean taking your Level 5 and setting topics and replies to 0 for that rank.

As you then go further up into your expert ranks you can weave topic and/or reply creation back in.  You have highlighted something else that’s really important, however, which is that in many communities (particularly support-oriented communities) we tend to value replies (very often means helping someone else) much more than topics (usually means asking a question).  But that can vary from community to community.  Generally, I’d simply say to think about which behaviours are most important in your community and build the ranks around that.

Kenneth’s reply is probably the most helpful, but one thing I wanted to point out on addressing your issue: 

I recently discovered a flaw in our ranking structure which has kept at least one top member “stuck” at a low rank because he is a “responder” rather than a “conversation starter.”  He has posted 58 replies and 16 best answers, but has never posted a new topic. 

Rank criteria uses AND rules, not OR rules. So unless they meet all the minimums, they won’t level up. (If you add an idea for adding OR rules, I’d vote for it!)

One work around to your problem is that you can create duplicate versions of the ranks, if you want to have an “Admin Exception” version. It’ll clutter up your back end rank structure a bit, but if you make the ranks the same in terms of name and styling on the front, nobody will know that your exception users are actually rank “Level 10 - Exception” and not actually “Level 10.” Then, for the exception requirements, you could zero out the areas they don’t have (Topics) and add a custom role requirement to trigger it; then when you add that custom role to the user(s) they get the rank if they meet the other criteria. 

---

And my two cents: 

Alternatively, should we consider removing the new topic requirement entirely from Level 3 and up? 

Yes, the simplest solution to your issue is just to zero out the topic requirement and any others that aren’t actually important to you. Based on what you share in the image, Topics seem way less important (they only go up very modestly in the upper levels). I’d also look closely at likes given to avoid the same issue. 

You could reward/incentivize topics differently via badges. Then you’re still recognizing and celebrating those who start valuable conversations, but the levels depend on what your system seems to value most: those contributing replies that lead to successful outcomes (answers). 

Overall, your challenge highlights something I think is key to gamification and recognition strategies: different personas are going to have different interests and motivations for how they contribute. It’s easiest for community managers to want all types of activities to increase as they move up ranks, but often times the top users just won’t be interested in some types of engagement. Consider if your strategy should adjust to them or if it’s a hard and fast rule that they need to do everything in order to level up. 

I shared my POV on gamification in my session at Pulse in May.

I also recently completed this Gamification course from The Community Strategy Academy that I thought said a lot of the same things better and more concisely than I did, along with some other solid strategies and examples.