Kesse tells you whether a change in your spending is larger than your own period-to-period variation. This page covers how to enter data, how to read the verdict, and the questions it usually raises.
A typical period appears after four; the before-and-after comparison after eight, because each half needs four.
| What it says | What it means |
|---|---|
| Not enough periods yet | Fewer than eight. With fewer than four on a side, a single unusual period would define the spread. |
| Inside your normal swing | The two halves overlap. The difference cannot be told apart from how much you vary anyway. |
| Lower than before | The halves do not overlap and the recent half is lower. |
| Higher than before | The halves do not overlap and the recent half is higher. |
A percentage is written only when the two ranges come apart. When they overlap, the difference is indistinguishable from noise, and putting a number on it would sell a certainty nobody measured. This is the point of the app, not a limitation of it.
It has two sources, kept apart because they behave differently. One is the uncertainty of the estimate itself, which narrows as you enter more periods. The other is how much your spending genuinely varies from period to period, which does not narrow — and should not, because it is real. An app that showed a single shrinking figure would be promising that enough data makes the uncertainty go away. It does not.
The spread is measured from the median absolute deviation. With three periods, one unusual period is the spread — a single month writing its own band. With four, the middle two deviations are averaged: an outlier moves the estimate but does not define it.
Because small samples understate their own spread. Measured across six thousand simulated histories with the level held exactly constant, a naive comparison at four periods calls chance a change 15.4% of the time. Kesse widens its band until that stays under 5%. The multipliers are a measured table (1.65 at four periods, 1.51 at five, 1.16 at six, 1.10 at seven, none from eight on), and they are re-measured on every build.
Either, in Settings. Changing it does not delete anything: entries recorded in the other length are kept but not measured. Mixing four weeks with one month would invent a conversion nobody measured.
Optional. With it, Kesse shows how many periods that amount lasts, as a range: the short end from the top of your spending band, the long end from the bottom. If the band reaches zero, there is no long end, and Kesse says “at least N periods” instead of inventing one. Left empty, nothing is shown — treating a missing amount as zero would say you have nothing.
No, and it never will. That connection is the single largest privacy cost in this category, and this measurement does not need it.
No. An export path is also an exfiltration path, and the build fails if one appears. The data is small by design — one number per period.
No. There is no server and no cloud sync. Entries made on one device stay there; your own device backup is the only copy.
You cannot: Kesse refuses a second entry for the same period and category, because it would silently double the total. Swipe an entry away on the Spending tab to remove it.
Your device’s own region setting. There is nothing to choose and nothing to convert.
No. Kesse reports statistics on numbers you typed in and never recommends what to do. It is not affiliated with any bank, broker or lender.
Anything unclear, anything wrong: hello@norea.studio.