I also filled the memory once and have taken to dumping the ride list more often. The unit is still reliable but when analysing the data I suspect (and I cannot prove this) that it:
- undervalues flat-land efforts by 20-50W (ie shows 150-180W when my manual calculations suggest 200W is closer)
- overvalues sprints by a considerable amount - as much as 50% higher (ie shows 1500W when manual calculations point to maybe 1000W) but only for a second or 2
- is most accurate at sustained high or medium-effort climbs, where the output is often within 2-5W of manual calculations based on speed, time and inclination
- lags the actual effort by 10-30 seconds
- loses it's pretty little head in corners and over bad bumps.
- smooth your data and remove 'outliers' such as spurious high-Watt readings
- normalise your data in a spreadsheet or online tool
- fine tune your friction and aero values on the provided USB-link software - this is better than re-doing the "coast" setup, I reckon, but it just may be that I've never done the 'coast' correctly (hmmmm...)
- ride on smooth roads and never go around corners.
Some other quirks are:
- It alters altitude overnight - presumably as the barometer rises and falls - so adjusting it is a good idea fi you want your data to be consistent
- It adds 100kg (or maybe just defaults to a really high weight) when you swap batteries - make sure you check your setup after changing batteries!