The temperature error is proportional to the height above the measuring station. The rule of thumb I use is that a 20 degC cooler day than ISA (say -5 degC at sea level) gives a 7% overread of altitude.
So at 1500 ft on glideslope you'll see 1600 ft on the altimeter on an ISA-20 day. When you get to 300 ft on the glideslope, the altimeter will read 330 ft. So add 30 ft to your normal decision height.
I'd suggest making the correction based on the temperature, not the altimeter reading at OM crossing. 100 ft is only half a dot at the OM. If you see 100 ft off at OM, it may be the glideslope indicator.
It's also worth bearing in mind that the altimeter setting is rounded to the nearest millibar (30 ft), a typical static calibration correction (PEC) is 30 to 50 ft at approach speed, and the barometer itself may be good only to +/- 50 ft anyway.
So don't sweat it! But I suppose on an ISA-20 day that's unlikely...
[This message has been edited by bookworm (edited 29 August 2000).]