Gain changes in Low Voltage Amplifier over the time

Gain changes in simple inverting-Amplifier over time

I built up simple inverting-Amplifier with LM358 which Gain is R2/R1 = 26.
input voltage is 20mV fixed and hence output voltage should be approximately 520mV. the problem is that the output voltage changes over time while the temperature(25°C) and other circumstances are stable. for instance, if output is 520mV today, next week it changes to 530mV.
any solutions?

Consider drift of 2% changes that affect output level and then verify by measurement.

Input level, temperature effects, R stability, measurement method

Thanks for your reply
Measurement tools are calibrated and verified. Resistors are Axial 1/4W, 1%. and input level is permanently fixed over time.
How is it possible to compensate drift ?

Either the resistor values are changing or you built it on a solderless breadboard that has intermittent contacts.

If resistor values are high and board is uncoated, even relative
humidity could make that kind of change.

At such low input level, Vio drift and things like charge pumping
of ambient 60Hz/120Hz hum could be at play. A copper box and
a known-stable voltage meter might be a good thing to try.

I see "520mV" and "530mV" as being a bit short of resolution to
call anything about stability or accuracy really, I have cheap
meters that flicker about that badly and good ones that don't.
You have to be able to believe the instruments before you get
excited about what they tell you. And of course you have
roughly zero prayer of getting a clean input difference voltage
reading that doesn't mess up the output, by hanging a meter
directly onto the amplifier - especially if Zmeter approaches
Znetwork, even 100:1 since you're fussing about 2% whole-
assembly stability.

You do not mention R3, which would be applied to the +
input in a value of R1||R2 to first-order-null input bias
current related offset errors. Without this, IIB becomes
an input offset voltage term (to be gained up 26:1). The
input bias current can be quite temperature and even
strain sensitive (depending on packaging).

1- Elements have been soldered on PCB so intermittent and uncoated contacts is not possible.
2- Volt meter is calibrated KYORITSU multimeter with 0.1 mV accuracy.
3- Compensated Resistor R3 is also applied by paralleling R1=15K and R2=390K.

I also bulit up similar application with AD620 according to figure 32 (in Datasheet). this application worked fine for nearly 3 weeks, but after that, I observed output voltage increased approximately 10mV. (similar to LM358)
I doubt if it is because some inner parameters of OpAmp changes over the time.

There are many possible reasons why measurements with a multimeter may be wrong. You didn't even tell about the signal type(DC or AC, frequency?), so any further guesses are void.

Thanks for your reply
Signal is DC and as I said in the first comment, input voltage is 20mV fixed.

THe best way to isolate a faulty result is isolate the root cause. THis may require that you measure each component for Resistance and OA for input offset.

Is the Input impedance matched for both (+) and (-) so that changes in input bias current does not affect the result?

If 2% drift of 30mV input occurred, the equiv. input offset is only 600 uV. What are the specs for Iio and Vio ?
What are the R values for each part before and after?