I've been off helping a friend and studying other tech. Rust C, good language. Did you miss me? My cat didn't.
So we all know the existence of news publishers rests on engagement with their customers on some level. The business case is a given. So this topic must dealing with specifics based on general observations.
So, respectfully, I find the view and understanding of the comments issue by most industry stakeholders (ie. Atlantic, TI, Coral, etc, etc) to be myopic. The reasons seem to vary (old-school thinking, intellectualism, distinguishing UI weakness from commenter problems, leveraging monetization foolishly, poorly thought out great new features and strategies, positions based on development schedules.) The list of these weaknesses is long.
On the flip side, there also exists some deep insight, great studies, a good (needs features, early version) of software, lots of talent, lots of enthusiasm, lots of smart people. But there seems to be a disconnect between deep insight on any one focused subject and how that should translate into a business plan. You actually come off as not being completely mercenary and corporatist, although I'll reserve judgement on that one. In balance I think your efforts here might succeed using your own metrics and there will be some wider benefits that include the commenters. The cost of failure being continued growing irrelevance.
In the mean time, TI is losing commenters. You would expect this, it is part of the business plan. But they are losing the best commenters. The issues being simple things. No automatic scrolling, no search, your position on search features, short edit times, comments closed too quickly, confusing nesting[, replacing a crude comment system with simplistic comment system having bugs].
New versions of Talk are being quickly developed. The agile methodology is working smoothly. This is how you develop software and you won't improve on it much. Users will tolerate a lack of features in a product if you communicate how and when you will respond to their complaints [and involve them in the process], which you don't. [If I am an exception to that it is because I doggedly chased you down.] However, note how many of the above complaints are features TI already has control of. What I view as misguided policy is always going to be the main problem in MSM and lesser players[, your plan does not reflect the market place].
While they are (and TI is) addressing theses issues their comments section is changing. Reduced numbers and variety seems to be occurring lately. I don't have access to their stats so I really can't confirm (cite) this. This is a problem they can address though. They can start by directly contacting these people, in the comments section if they like or privately. Prove you don't want to lose them. Easy. Why am I mediating this issue on TI's site?
As a counter example: I am seeing a similar drop for completely different reasons on RT. This resulting [from the] deliberate biasing in US search result ranking. They are limited in strategies that address this effectively. Similar results but a much more difficult problem.
That's a long rant so I will get to the point:
My (premature, it's pressing) recommendation should be one sentence: It's all about the view of the data, as in you want every and all commenters regardless of their nature because the view only shows the best commenters with clickable side (nested) threads where the (labelled) banter, dialogue, off-topic, toxic, irrelevant, (define a category) less prefered comments are alway available. (Incidentally) plugins and moderators manage this nested best view dynamically and users also vote.
You want money right? You want a big community? You want to show the best comments representative of your organization and objectives right? Well that does mean you want a big community, and not a bunch of intellectually snobbery. Dear goodness, The Atlantic wants us to write letters to the editor that they just might post because their current commenter community doesn't suit their tastes. Lunacy...
Regards, David G. Horsman