Waiting is a trade, not a sin
The usual framing is that deferred maintenance is always more expensive, which is not quite true and being sloppy about it makes the genuinely urgent cases harder to argue for. Three things move while you wait. The price of the work climbs at whatever rate that trade is climbing. The money you did not spend earns something. And the fault has some chance of causing damage that costs more than the fault does.
The first two are ordinary arithmetic and they often cancel or even favour waiting. On the opening figures the repair costs $1,800 today and $2,025.92 in four years, but four years of the cash earning 5% turn that into $1,666.73 in today money — cheaper than fixing it now, by $133.27. That is the honest starting position before any risk is counted, and it is why a blanket rule about deferral gets the easy cases wrong.
The third term swamps the other two
Then the damage goes in. At the opening 8% a year, the expected consequential damage comes to $3,250.77 in today money, which turns a $133.27 saving into $3,117.50 of cost. One term dominates the whole page, and it is the one made of a number nobody can look up.
So the page solves it backwards. Holding everything else where it is, the chance of the damage starting would have to be about 0.29% a year for waiting to break even — roughly one year in three hundred and forty. Almost nobody would defend a number that low for a fault they can already see. That is a far more useful output than the total, because it does not depend on getting the probability right. It only asks whether the true figure is above or below a threshold, and here the threshold is so low that the estimate stops mattering.
What to put in the damage box
The damage figure is the consequential damage only, on top of the repair you were going to do anyway. For anything involving water that means the parts of the building that get wet and are not visible: framing, sheathing, subfloor, insulation, the finishes that have to come off to reach them, and the contents. For anything electrical or mechanical it is what the failure takes with it. The reason estimates come in low is that people price what they can see, and the thing that makes water damage expensive is precisely that it happens where nobody is looking.
Two inputs deserve a second pass as well. If the repair would go on a credit card rather than come out of savings, the earnings rate is not a savings rate — enter the card rate as a negative number and the comparison flips hard. And the per-year chance is held constant across the whole delay, which flatters waiting, because most faults get likelier the longer they are left.
Related pages
If the item is a whole system rather than a fault — a roof at the end of its life rather than a leak in it — the question is a schedule question and the capital reserve timing calculator puts it in a year. At the moment something fails, repair or replace compares a repair quote to a replacement on cost per remaining year. If several jobs are waiting on the same scaffold or lift, the access bundling calculator prices doing them in one visit.
For what a leak actually does before anyone notices, the water leak guide is the plain version, and the maintenance schedule covers the checks that find these things while they are still cheap.
Questions people ask
Where does the chance figure come from?
From you, and the page says so as plainly as it can. Nothing here supplies a failure rate for anything, because failure rates depend on the component, the installation, the exposure and the climate, and a general figure would be worse than no figure. What the page does instead is solve for the break-even chance, so you can ask whether the true number is above or below a threshold rather than trying to name it.
Why discount the future repair at all?
Because otherwise you are comparing money today with money in four years as though they were the same thing, and they are not. If you fix it now, the cash leaves today. If you wait, it leaves in four years and does something in between. Discounting puts both on the same footing. Set the earnings rate to zero if you would rather see plain nominal amounts, and the comparison becomes the raw price growth.
What if I would put the repair on a card?
Enter the card rate as a negative earnings rate. That is the honest version: waiting does not save you a return, it costs you interest on the balance you carry. The effect is large, and it usually turns a delay that looked marginal into a clear cost even before the damage term is counted.
Does a small leak really get worse in a straight line?
Usually not, and the page is deliberately conservative about it. The per-year chance you enter is held flat across the whole delay, which means the model assumes a fault is no likelier to start damaging things in year four than in year one. Most are, so the real cost of waiting is generally higher than what comes out here. Treat the output as a floor rather than a central estimate.
Is deferring maintenance a problem for insurance or a warranty?
It can be, and that is not something a calculator can settle. Policies commonly distinguish sudden damage from damage that developed over time, warranties commonly require documented upkeep, and both are written in the documents themselves. Read the policy and the warranty terms, and ask the insurer or the manufacturer directly. None of that appears anywhere in this arithmetic.