A Zerohedge article caught my attention. There is another way to calculate the non-farm payroll numbers and this is called the JOLTS (Job Openings and Labor Turnover) data. When you subtract hires and separations from each other, you get the change in non-farm payroll numbers.
Change in NFP = Hires - Separations.
As you can see here, the charts do match with each other and it looks like the JOLTS data is more accurate than the NFP data (which gets revised a lot). This correlation isn't a very important one, but you can use the JOLTS data to predict the NFP data and with NFP data, you can predict the trend of the unemployment rate.
This brings me to this more interesting chart.
Here you can see the change in NFP data versus the change in unemployment rate. As you can see, whenever the NFP changes to the downside (less people get jobs), the unemployment rate of change edges upwards.
So, the key is to watch these JOLTS and NFP data and predict the curve of the unemployment rate.