Day-to-day variation is almost never mysterious. Something is changing, and there are only a handful of candidates.
Your dose is drifting. If you grind by time rather than weight, the dose changes as the beans do. Fresh, moist beans flow differently from dry ones three weeks later, so three seconds of grinding does not deliver a constant weight. Weigh the dose and this disappears.
Your beans are ageing. Coffee changes measurably over its life. In the first few days it is gassy and brews fast; at two to three weeks it settles; past a month it flattens. Re-dialling every week is normal, not a fault.
Your grinder retains coffee. If a gram of yesterday’s grounds sits in the chute, your first dose today is partly yesterday’s coffee. On a single-dose workflow this matters; with a full hopper it does not.
The machine is dirty. Oils accumulate and go rancid gradually, so the change is slow enough that you adapt without noticing. If coffee has been getting worse over months rather than varying day to day, this is the one.
Also worth checking
Temperature. On a machine without a PID, whether you flushed before pulling and how long you waited changes the brew temperature by several degrees.
Water. If you sometimes fill from a filter jug and sometimes from the tap, you are brewing with two different waters.
How to find it
Fix everything you can measure — dose by weight, yield by weight, timing — and change nothing for three days. If the coffee is still varying, the variable is one you are not measuring: temperature, water, or a dirty machine. Clean it first; it is free.
Questions people actually ask
Does humidity really affect espresso?
It has a measurable effect on how coffee grinds and how the puck behaves, and it is much smaller than dose drift or bean age. Chase the big variables before blaming the weather.
Should I purge my grinder before the first dose?
If you single-dose, a small purge or a bellows puff clears yesterday’s retained grounds. With a full hopper, the coffee in the chute is the same coffee, so it does not matter.
Last reviewed 8 August 2026