We see this in technical domains too. For example, I think Clojure succeeded this way. (We can argue about how much Clojure has succeeded, but it's a Lisp—we have to grade on a curve.) I don't think any attempt at improving on Common Lisp, no matter how technically solid, could have achieved that without "naming a new game", to use the article's vocabulary. Hot newness dominates improved oldness: every element of the set of hot newness beats every element of the set of improved oldness. Elixir/Erlang is another example. I say this article is one of the rarest things, a piece of marketing literature (like Crossing the Chasm and maybe the Innovator's Dilemma) that is useful to engineers.
We offer form, phone number, email, and chat. I would guess 40% chat, 40% direct email, 15% form, 5% phone?
Personally, I think forms suck on a phone and probably go to a black void I can’t even follow up on.
It's usually a chatbot that handles the repetitive questions when they're worded properly, and when it can't, asks you to hold on a sec while it forwards the question to the customer-service-farm.
Also, unrelated: I hate those "conversational" customer service bubbles. I don't know what I expect to happen when I click them, but I know that "Jennifer H", whose picture is shown next to it, is not sitting at a keyboard eagerly awaiting my questions. Maybe I expect a chatbot, maybe I expect outsourced customer-service-farms, but I don't click buttons when I don't know what they really do, especially when I know they're avoiding being up-front about it.