A cleat works loose. A shelf starts to move under load. A back panel that was tight in March rattles by the next March. You go back and look, and the screws are still there, still seated, still tight to the touch. Nothing stripped. Nothing broke. The connection just stopped doing its job.
Here is what those three have in common. The screw was driven into end grain.
A screw into end grain holds at about 75 percent of what the same screw holds in face grain. That is the Forest Products Lab number, and it does not depend on your driver, your torque, or how good the screw felt going in. The fibers in end grain run parallel to the threads instead of across them, so there is less wood for the thread to pull against when the joint is loaded in withdrawal. The screw feels just as solid the day you drive it. The penalty only shows up later, under load and seasonal movement, which is the only condition that was ever going to test it.
The trouble is that end grain is exactly where a lot of structural screws end up. Carcase backs screwed into the ends of shelves. Shelf cleats fastened to the end of a vertical panel. Breadboard ends pulled down from above. Drawer runners landed on an end-grain face. All of them load-bearing. All of them places where the screw is the only thing holding the connection, and all of them landing in the 75 percent.
Density makes it worse before it makes it better. The same screw holds three to five times harder in white oak than in white pine, because withdrawal scales steeply with how dense the wood is. That is good news in oak and bad news in pine, and the end-grain penalty stacks on top of either one. A screw into the end grain of pine is a long way from a screw into the face grain of oak.
While you are thinking about species, fix your pilot hole. Most woodworkers drill one size for a given screw and use it in everything. That works in softwood, where the fibers compress and grip the thread. In dense hardwood it does not. Hard maple wants a bigger pilot, around 90 percent of the thread root, against roughly 70 percent in softwood. The bit that seats clean in pine will split the maple, and a split has no holding power at all. So an undersized pilot in dense stock is worse than no pilot. Match the lead hole to the wood, not just to the screw.
The fix for the end-grain problem is not a bigger screw. A bigger screw in end grain is still in end grain. The fix is to put the load somewhere the wood can hold it. Land the fastener in face grain instead, on a different face or a different angle. Glue a block where you can screw into long grain. Use a cross-dowel, a threaded insert, a connector bolt, anything that gives the screw a side-grain or mechanical purchase. Or cut a joint that does not ask the fastener to carry the withdrawal load in the first place. A number 8 into the end grain of a shelf is not a shelf joint. It is a hinge that has not loosened yet.
One more, different from the rest but invisible the same way. A smooth-shank nail in a piece that sees real seasonal humidity swings can lose most of its holding power over time. The wood grips the nail on the way in, then relaxes and re-compresses through a few seasons until the friction that held it is gone. Smooth nails are a temporary hold on anything that moves with the weather, not a permanent one. If the connection has to last, that is the wrong fastener.
None of this means every screw you have ever driven into end grain is about to let go. Plenty hold for years, especially in hardwood under light load. The ones that fail are the ones under stress and movement, and that is the case worth designing for. So before you trust a screw to hold a load, check where it is going in. The screw feels strong on the driver. Whether it stays strong is decided by the grain it lands in.