The joint is the answer to a question. If you don't know the question, you can't know whether the answer is right.
That's where this arc started. Seven posts later, here's the protocol that closes it.
What is the primary load?
Tension or racking. Those are the two cases that require mechanical geometry.
Tension means pull-apart force — drawer fronts, carcase corners, breadboard ends. Racking means rotation in the joint plane — chair rails, table aprons, door frames. Most furniture joints are carrying one or the other. That's where the decision actually lives.
Housing is the third category: dado, rabbet, groove. You're capturing a panel, not resisting a mechanical load. The joint does its work by geometry, not by glue surface.
Then there's nothing structural. Light load, static connection, redundant load paths in the assembly. That's where adhesive and fasteners are adequate. That's where the butt joint belongs.
Get the primary load wrong and everything else is solving the wrong problem.
Does it need mechanical resistance?
Test it this way: will this connection take repeated, cyclic load? A chair rail does. A drawer front does. A face frame on a kitchen cabinet doesn't — the load is shear at the panel, not tension at the corner.
If cyclic load, you need geometry. Fasteners work loose over time. Glue degrades without compression. The joint shape has to carry the movement across the life of the piece.
If the load is static and light, adhesive and fasteners are correct. A face frame with pocket screws isn't a concession. It's the right answer for the load.
What constraints do you have?
Tools, time, skill. None of these change the structural requirement. They change which joint within the right family you cut.
Your shop has a Domino but no mortiser? Floating tenon. Structurally that's M&T-class. A chair rail still needs M&T-class resistance — you just choose the version your tools can produce accurately.
Time works the same way. Pocket-screwed face frames are not a compromise. They're appropriate. A dovetailed face frame isn't better; it's the wrong application of that time.
Skill affects execution, not selection. The beginner who picks the right joint and cuts it poorly is in better shape than one who picks the wrong joint and cuts it well. The right joint at least has a margin for error. The wrong one fails regardless of how clean the line is.
What's the visual requirement?
A through dovetail at a carcase corner says something about how the piece was made. A pocket-screwed corner says nothing visible and doesn't need to. If the joint is part of the aesthetic argument — if it's a carcase on a hand-tool piece, if the joinery is itself evidence — that affects which joint within the structural family you choose. If it's hidden and utilitarian, the fastest adequate joint wins.
Four questions, in that order. The first one does most of the work.
Seven posts covered four structural families — the mortise and tenon, the dovetail, the dado and its relatives, the butt joint — plus three hybrids that borrow from two families to solve a specific problem. The protocol doesn't replace knowing the joints. It's the frame that makes the knowledge accumulate faster. Every connection you cut now has a reason behind it.
That reason is what you carry into the next piece.
Know the question. Then choose the joint.