Every design review eventually produces the same fight. Somebody wants the thing findable and somebody wants the thing to feel earned. Put the entry point in the top level navigation and it converts, but it flattens, because nothing in a main menu has ever felt like a discovery. Bury it behind a gesture, an easter egg, or a members area and the people who find it care more, while most users never learn it exists at all. The usability heuristics have an answer to this. The answer is boring, and clients keep asking for the other thing regardless.
The argument settles faster against a physical example, because buildings have been running this experiment much longer than software has. A speakeasy is a hidden entry point with a century of field data behind it, and the pattern turns out to be strict. A real door, deliberately underspecified, plus one marker legible only to somebody already looking for it. Not zero signal. Low signal, aimed.
One working instance sits downtown: a craft cocktail bar with a hidden room called the Captain's Quarters. Whale stencils on the brick, a door easy to miss, and both of those facts stated plainly on its own website. Read the whole thing as an interface and it resolves cleanly. The public site carries the discoverability, so the information is retrievable by anyone who bothers to look. The physical layer stays unmarked, so the finding still happens in the room. Documentation does one job. Space does the other.
That split is the transferable part, and it is cheap to implement. Hide the affordance in the product and put the explanation in the docs, the onboarding email, or the release notes. Command palettes have worked this way for thirty years. Nothing in the chrome advertises the shortcut, the shortcut is written down where a curious person will hit it, and the first time a user fires it the moment feels private even though it was public the entire time.
The other half of that room is worth a harder look, because it is the pattern most clients actually ask for. Behind the door there is no menu. A bartender asks what you like and builds to it. In interface terms that is a conversational front end replacing a catalog, and it fails on the web the same way it succeeds in a bar. It needs an expert on the other end. Swap the expert for a lookup table and it collapses immediately, which is the honest reason so many chat widgets underperform the search box they were installed to replace. The pattern is not the problem. The staffing is.
So a rule for the next build. Count who genuinely needs to find the feature. If the answer is everyone, put it in the navigation and stop litigating it. If the answer is the fraction of users who will care, then make the door real, keep the stencil small, and write the whole thing down somewhere a curious person can reach it.