One ratio decides it
Strip the comparison down and it is a fixed cost against a per-socket cost. The switch route pays a large amount per fixture — the switch itself, whatever the awkward boxes cost extra, and whatever anybody is paid to be in them — and then a couple of dollars per socket for ordinary lamps. The bulb route pays nothing per fixture and a lot per socket. So the answer turns on sockets per fixture and almost nothing else.
On the defaults here — 14 fixtures, 2 sockets each, 3 multi-location fixtures with one companion apiece, 4 boxes needing a $16 dearer switch, switches at $22, smart bulbs at $13, ordinary bulbs at $3, 15 minutes of paid time per switch at $95 an hour, and a 10-year comparison against an 8-year bulb — the bulb route comes to $728 for 28 bulbs and one replacement set, and the switch route to $925.75 for 17 switches. The bulbs are $197.75 cheaper, and the crossover sits at 2.61 sockets per fixture. Below that the bulbs win; above it the switches do.
Change one input and the whole thing turns over, which is why running it with your own figures matters more than the headline. Set the paid time to zero because you are doing the work yourself: the switch route falls to $522, the switches now win by $206, and the crossover drops to 1.36 sockets — meaning almost any fixture with two lamps in it favours a switch. That is the honest summary. The labour line, not the hardware price, is what usually decides this, and it is the line most comparisons leave out entirely.
The bulb replacement is the other input that swings it. Set the bulb life equal to the 10-year comparison period, so no set is replaced, and the bulb route drops to $364, the bulbs win by $561.75, and the crossover moves out to 6.01 sockets per fixture. Whether smart bulbs see out a decade is a judgement about products this page will not make for you, and the difference between assuming they do and assuming they do not is over $360 on 28 sockets.
The multi-location problem is a switch problem
A stairwell light with a switch at the top and one at the bottom needs a device at both, because leaving one of them ordinary means somebody can put the smart one out of circuit. Three such fixtures on the default list add three switches, taking 14 to 17 — 21 percent more hardware and 21 percent more labour, for no additional light controlled.
The bulb route does not have this problem at all — one bulb serves however many switches there are — but it has the mirror image of it, which is that all of those switches must stay on forever. Neither is strictly better. It depends on whether the room has more sockets or more switches, and that is a question you can answer by walking around with a notepad.
What the money does not capture
Three things sit outside the arithmetic and any of them can outweigh it.
The first is the wall switch. A smart bulb that has been switched off at the wall is not slow, it is gone, and no automation reaches it. Households where that is a daily event are households where the bulb route quietly stops working, and no price comparison predicts which households those are.
The second is what happens to the socket later. A switch controls whatever is plugged into it, so a fixture swapped next year is still automated. A smart bulb is the automation, and it goes in the bin with the fixture.
The third is the device count, which the page shows because it matters for reasons other than money: 28 bulbs against 17 switches is 65 percent more things joining the network, reporting state and needing firmware. Whether that matters depends on the network, and coverage is a separate question worked out on the access point coverage calculator.
Questions people ask
Are smart switches or smart bulbs cheaper?
It depends almost entirely on sockets per fixture and on whether you are paying somebody to fit the switches. On the defaults here — two sockets a fixture and paid labour — the bulbs win by $197.75 across 14 fixtures. Take the paid labour out and the switches win by $206 instead, on exactly the same hardware prices. The page shows the socket count at which it flips for your own figures, which is the number worth having.
Why do I need a bulb in the switch route as well?
Because the socket still needs a lamp. It is only a few dollars each, but it scales with sockets in exactly the same way the smart bulbs do, so it subtracts from the per-socket gap between the two routes and pushes the crossover further out. Leaving it out of a comparison makes the switch route look better than it is.
What is the no-neutral field for?
Some switch locations need a different and usually dearer device than others, and which of yours are like that is something only somebody who has opened the boxes can tell you. The field takes that count and the price difference and adds it to the switch route. It is a price input, nothing more. This page states no wiring rule and makes no judgement about any box or circuit.
Can I mix the two?
That is what most houses end up doing, and the page supports it by running the comparison over whichever fixtures you put in scope. Run it once for the single-socket ceiling fittings and once for the multi-arm fixtures and the answer will usually come out differently for each, which is the correct result rather than an inconsistency.
Should smart bulb replacement be counted?
Over a long enough comparison, yes. A smart bulb is a small electronic device as well as a lamp, and it is replaced as a unit when either half fails. At 8 years a bulb over a 10-year comparison the page counts one extra full set, which on 28 sockets is another $364 and roughly doubles the bulb route. If you expect them to see the period out, set the two figures equal and the line disappears.