A pay raise is the easiest money to lose track of. The salary goes up, the month ends the same way it always did, and a few months later nobody can point to where the difference went. It was not spent on anything memorable. It dissolved into the baseline: a subscription upgraded here, a delivery habit there, a slightly larger grocery order each week. Small increases are almost invisible one at a time, which is exactly why they absorb a raise so completely.
Building a net worth tracker kept running into that pattern. People would log a higher income, watch their spending rise to meet it within two or three cycles, and conclude the tool had nothing useful to say. The interesting design problem was not tracking the raise. It was catching the moment before the raise stopped being a distinct number.
What was built
Introworth at https://introworth.com tracks assets, debts and the value of time in one picture, and the part of the build this post is about is the baseline snapshot that sits underneath that picture.
The snapshot is deliberately plain. It records what a normal month actually costs right now: housing, regular bills, subscriptions, food, transport, and the irregular costs averaged honestly rather than ignored. When income changes, in either direction, the flow asks for one decision while both figures are still known: split the difference on purpose. A portion to debt or savings, a portion kept for genuine quality of life, and the remainder left as breathing room. The split does not need to be strict to work. It needs to happen once, deliberately, before the higher pay settles in and becomes the new normal.
Three choices mattered most. First, the baseline is captured before the change, not reconstructed after it. A baseline written from memory is always kinder than the real one. Second, the additions are kept as a list in the order they arrived, so if income later drops, the household can lower the baseline by decision, cutting the newest additions first, instead of by drift. Third, the snapshot lives next to the trend line rather than replacing it. Net worth over time shows whether the chosen baseline is actually holding; the snapshot shows what holding it is supposed to look like month to month.
What broke
The first break was timing. The earliest version prompted people to set the split after the first higher paycheck had already landed. By then the money had mentally merged with the old salary and every proposed portion felt like a loss. Moving the prompt to the moment the change is recorded, before payday, fixed most of that resistance. The same split, offered a week earlier, read as a plan instead of a cut.
The second break was granularity. The first snapshot wanted every recurring cost itemised, down to small subscriptions most people could not recall without opening a bank statement. Completion fell off a cliff. The fix was to ask for the few large lines exactly and let the small recurring costs be entered as one honest lump figure, editable later. A slightly approximate baseline that exists beats a precise one that was abandoned halfway through.
The third break was tone around the quality of life portion. Early copy treated any spending increase as leakage, which quietly shamed the part of a raise people had genuinely earned and looked forward to. People stopped recording raises at all. The rewrite names that portion plainly as part of the plan, not a failure of it. A raise split on purpose includes enjoying some of it. Recording went back up once the tool stopped arguing with that.
What the numbers look like
The honest measure here is not the size of anyone's raise. It is whether the difference is still visible three months later. In early use without the snapshot, the pattern was consistent: spending rose to meet new income within about two monthly cycles, and the net worth trend line barely moved. With the baseline chosen at the time of the change, the split mostly held, not perfectly, but visibly. The savings and debt portions survived because they were moved first, as part of the plan, rather than left to whatever remained at month end, which was usually nothing.
There is a second number worth naming: how often the baseline gets revisited at all. A household that knows its real baseline can answer a drop in income in one sitting, because the additions are listed in order and the cuts are a decision, not an investigation. That turned out to matter as much as the raise case, and it was not the use the feature was originally built for.
What stayed
Capture the baseline before the change, split the difference while it is still a distinct number, and keep the additions listed in the order they arrived so the baseline can move down by decision as cleanly as it moved up. None of that requires discipline as a personality trait. It requires one deliberate moment, taken early, with both figures on the screen at the same time.
The lesson that survived every revision is simple: a raise kept on purpose is worth more than the same raise absorbed by accident, and the difference is almost never the amount. It is whether anyone chose the new baseline, or whether the baseline chose itself.