Measure the path, not the floor plan
The single reliable way to be short of cable is to take the distances off a drawing. A floor plan is a horizontal projection, and a cable run is not horizontal. It leaves the panel, climbs to a joist bay or a loft, travels a route determined by where the framing lets it travel, comes back down inside a partition and arrives at a box at knee height. On a two storey house the vertical component of a typical drop is comfortably a tenth of the run, and on a three storey one it can dominate.
Then there are the loops. A service loop at the panel lets somebody pull the panel forward and work behind it. A loop above the outlet lets somebody reterminate a jack that got crushed, twice, without pulling a new run through a finished wall. Both are cheap while the cable is on the reel and impossible to add afterwards. Together with slack for the route being longer than you thought, they are the difference between a takeoff that lands and one that leaves you three drops short with the drywall going up on Monday.
Why two drops per location instead of one
The marginal cost of the second cable is the cable itself, one more jack and ten minutes. The marginal cost of deciding later that you wanted it is opening a wall. That asymmetry is the whole argument, and it does not depend on predicting what devices will exist in ten years. Anything with a port benefits, and anything wireless still wants a wired feed to the radio that serves it.
The same logic runs the other way at the panel. A drop that has nothing plugged into it costs one port on a panel you already bought. A drop you did not run costs a day. Pull to the awkward places while you can: the far corner of the loft for a future access point, the garage, an eave for a camera, the wall behind wherever the television will end up even though it will be wireless this time.
The length limit and what actually breaches it
Copper twisted pair has a channel length limit, and this page takes it as an input rather than asserting one, because the figure depends on the cabling standard you are working to, the cable and the equipment. The widely used working figure for a copper channel is 100 metres, about 328 feet, conventionally divided into a permanent link from panel to jack and an allowance for patch cords at each end.
What breaches it is almost never a long horizontal run. It is a moderate run plus the things that get left out of the estimate:
| Component | Easy to forget | Typical bite |
|---|---|---|
| Horizontal route | No | The number people quote |
| Vertical rise and drop | Yes | Several feet per storey, at both ends |
| Service loops | Yes | A few feet at each end |
| Routing around obstructions | Yes | Ten percent is a modest allowance |
| Patch cords | Frequently | They count towards the channel and are worse per foot than solid |
The consequence is that a run measured at eighty feet on a plan can be a channel that fails a certification test, and it will fail intermittently and at the worst speeds rather than cleanly. A run that genuinely exceeds the limit is not a cable quality problem. It is a topology problem, answered by relocating the panel, putting an intermediate switch closer to the far end, or running fibre for that one leg.
Ordering in whole boxes and what to do with the remainder
Cable comes in boxes and the last one is always partly full. That leftover is not waste if you plan for it: it is the two drops you will want in six months, the replacement for the run somebody puts a screw through, and the material for making up your own patch cords if you are inclined that way. Rounding down to save a box is the one economy on this list that reliably costs money.
Terminations run ahead of drops for the same reason. The first few jacks anybody terminates are practice, and a crushed one is not repairable in place. Order spares in proportion to how new you are to it rather than in proportion to the job. The tooling matters more than the brand of the parts, and a jack terminated with the wrong pair order will link at a lower speed rather than failing, which is the most annoying possible failure mode.
Adjacent numbers this does not cover
This is a quantity takeoff. It says nothing about whether a given run will carry a given signal, which for audio and video links is a separate question handled by the AV cable run calculator. It says nothing about how much power the switch has to deliver if the drops feed cameras or access points, which is the PoE power budget calculator. And it says nothing about where all this terminates, which is the structured wiring panel planner. If the drops have to share a conduit or a sleeve on the way, size that first with the low-voltage pathway fill calculator, because a pathway that is full is a pathway you cannot add to.
Questions people ask
How much cable does a whole-house install take?
It scales with the route, not the square footage, which is why the average run field matters more than any of the others. As a shape of the answer rather than a number: a drop in a two storey house is commonly half again as long as its horizontal distance once the rise, the loops and the slack are counted, and a house with a couple of dozen drops is usually into a second box. Rather than trusting a rule of thumb, walk the actual route for the nearest drop and the furthest one, average them, and put that in. Fifteen minutes of walking beats any estimate anyone can give you from a description of the building.
Do the patch cords really count towards the length limit?
Yes, and they are worse per foot than the solid cable in the wall, because patch cords are stranded for flexibility and stranded conductors have higher loss. That is exactly why the conventional split reserves part of the channel budget for them rather than letting the permanent link use the whole thing. The practical failure this causes is invisible at install time: the run passes with the short cords used during testing, then somebody swaps in a long cord to reach a device across a desk and the link starts negotiating down. If a channel is near its limit, note that fact at the panel so the next person knows not to lengthen it.
How many spare ports should the patch panel have?
Enough that adding a drop later does not mean buying and mounting a second panel, which in a wall enclosure may not physically be possible. Panels come in fixed port counts, so the practical answer is usually to take the next size up from your drop count rather than the closest fit. The cost difference is small and the alternative is discovering that a 24 port panel holds 24 drops on the day you want the 25th. The same reasoning applies to switch ports, with the extra wrinkle that the router and any uplinks consume ports without being drops.
Can low voltage cable share a hole or a box with mains wiring?
That is governed by the electrical code adopted where you are, and the authority having jurisdiction decides how it is read. There are rules about separation, about divided boxes, and about what may share a pathway, and they are not the same everywhere or in every edition. Anything that shares a box or a pathway with line voltage is licensed work and this page gives no procedure for it. What is worth knowing generally is that cable inside a wall or in an air handling space has rating requirements, that the rating is a real and enforced distinction rather than a marketing tier, and that the time to establish which rating applies is before you buy a box of the wrong one.
Is it worth running fibre alongside the copper?
It is a genuine option and it is cheapest at exactly one moment, which is while the walls are open. The case for it is strongest on the runs where copper is already marginal, on the leg between buildings, and on the backbone between a main panel and a remote switch. The case against is that terminating fibre is a different skill with different tooling, and that pre-terminated assemblies need a pathway large enough to pull a connector through, which changes the conduit sizing. A common middle path is to pull a spare pull string or an empty innerduct on the routes where you might later want it, so the decision stays open without committing to it now.