Start underneath, because that is the bit that fails
Decks very rarely fail because the boards wore out. They fail because the frame underneath stayed wet, and everything that decides whether it stays wet happens before a single board goes down.
So the ground below wants clearing of vegetation, covering with a membrane and a layer of stone so water drains away instead of sitting, and the frame wants to sit clear of it with air able to move right through. A deck with soil banked against the joists and no airflow is a damp timber structure regardless of what it was treated with, and the treatment is a delay rather than a cure.
Getting rid of the water and letting it dry is worth more than any specification decision you can make further up.
The connection to the house is where it goes wrong, and differently in different countries
If the deck touches the building, one joint carries almost all the risk on the job. What that risk is depends on where you are, and each market warns about its own and never mentions the other.
In Britain the hazard is the damp proof course. A masonry wall has a barrier built into it near the bottom, and it works only because nothing bridges it. Build a deck whose surface sits above that line, or bolt a ledger across it, and water now has a route around the barrier into the wall. It appears months later as damp inside at skirting level, and nobody connects it with the deck.
In the United States the hazard is the ledger itself letting go. Deck collapses are overwhelmingly ledger failures rather than broken timber, which is why how that board is attached is treated as a structural matter there rather than a fixing preference.
Both disappear if the deck stands on its own posts and simply sits near the wall with a gap. For most domestic decks that is the better answer, and it costs one more row of posts.
Frame first, and get it flat before anything covers it
The frame is the only chance you get at level, square and solid. Once boards are on it, everything is fixed and every error is permanent and visible.
- Set the outer frame and check the diagonals. Equal diagonals means square, and it is the only quick check that actually proves it.
- Work off the height that matters, which is usually a threshold or the damp course, not the ground, because the ground is not level and the door is.
- Joists run across the direction the boards will run, so decide the board direction first. It is the visual decision on the whole job.
- Support the frame properly at its ends and brace it. A frame that racks slightly will telegraph through as movement underfoot forever.
Fixings, because treated timber now eats ordinary screws
This one has changed and plenty of advice has not caught up. The copper-based preservatives that replaced older treatments are considerably more aggressive to steel, so ordinary screws and lightly galvanised fixings corrode in treated joists far faster than people expect.
Use stainless, or fixings explicitly rated for modern treated timber. The corrosion happens inside the joint where nobody can see it, so the first sign is a board going loose years later, and by then it is everywhere at once.
Laying the boards
Two screws per board at every joist it crosses, not every other joist. Boards cup as they dry and the fixings are what restrain them. You will also be told to lay them bark side up: that rule is genuinely disputed, and fixing every crossing properly does more to keep a board flat than which way up it went.
Pilot drill near the ends. A screw driven into the last few centimetres of a board acts as a wedge and splits it along the grain, and there is no timber either side to resist it there. Thirty seconds a board against a split that shows for the life of the deck.
Leave the gap between boards. It is there for drainage and for movement, and on a hot dry day timber that goes down tight will push against itself when it swells.
Quantities, and the gap that changes them
Work out the materials before you start, because the gap between boards is part of what each board covers and leaving it out of the sum is how people end up a pack short. The decking calculator handles that, along with the joist count, which is one more than most people expect, and what diagonal laying really costs in waste.
For the frame, joist span covers how far a section will reach before it needs another support, and timber quantity turns the layout into lengths to order. If anything is being notched or drilled for a service, the notching limits are narrower than they look.
Common questions
Does a deck need a gap under it?
Yes. Air has to move under the frame so that the underside of the joists and boards can dry after rain, and the ground below needs to shed water rather than hold it, which is the difference between a deck that lasts fifteen years and one that goes soft in five. A deck built tight to soil, with vegetation growing up into it and no airflow, is a permanently damp timber structure whatever it was treated with. Membrane and stone underneath, and a clear ventilated void, are worth more than any upgrade to the boards.
Can I fix decking to my house wall?
You can, and in Britain the thing to be careful about is the damp proof course. A masonry wall has a barrier built into it low down that stops ground moisture rising, and it works by having nothing bridging it. A deck whose surface sits above that line, or a ledger fixed across it, gives water a path around the barrier and into the wall, and the result appears indoors as damp at skirting level. Keeping the deck surface clearly below the damp course, with a gap between deck and wall, avoids the whole problem.
Is a free-standing deck better than an attached one?
It is simpler and it removes the two failure modes that matter, so for a lot of domestic decks it is the better answer. Standing the frame on its own posts means nothing is fixed through the wall, nothing can bridge a damp course, and nothing depends on fixings into masonry of unknown quality. The cost is an extra row of posts and a little more timber. Where an attached ledger genuinely is the right answer, how it is fixed is a structural question rather than a DIY preference.
What screws should I use on treated timber?
Stainless, or a fastener specifically rated for contact with modern preservative-treated wood. This changed and a lot of older advice has not caught up. The copper-based preservatives that replaced the old chromated arsenical treatments are considerably more aggressive to ordinary steel and to light galvanised coatings, so plain screws in treated joists corrode far faster than people expect, and they corrode where nobody can see them. It is one of the cheapest places on the job to avoid a problem.
Should the boards go bark side up or down?
You will be told bark side up with great confidence, and it is genuinely argued about, because the shrinkage behaviour of timber points the other way. What is not argued about is that boards cup as they dry and that fixing every board down at every joist it crosses restrains that far more effectively than the orientation does. So screw rather than nail, fix at every joist rather than every other one, and let the fixings do the work instead of hoping the grain does it for you.
Why do I need to pilot drill near the ends?
Because a screw is a wedge, and driving a wedge into the last few centimetres of a board splits it along the grain. Near the middle there is timber either side to resist that; near an end there is not. A pilot hole removes the material the screw would otherwise have to displace. It is thirty seconds a board and the alternative is a split that either shows for the life of the deck or has to be cut out.