It starts with simply observing the plants. You can vary the dosage from crop to crop or between plants in the same crop. Plants with leaf tip burn will generally have too high of a concentration and plants with pale leaves will generally not have enough nutrients. This assumes other factors like pH are correct.
Another option is leaf/flower/fruit analysis. For example, if you're growing spinach, you may view iron content as an indicator of quality, healthy plants. Determining the iron content of the leaves from plants dosed at different rates will help zero in on the optimum fertilizer program.
The best way, IMO, is to test the nutrient solution before and after it has been "fed" to the plants. This requires a recirculating system for best accuracy, but runoff can also be collected in a drain to waist system. As the author mentioned, unused nutrients will build up in the solution. An ideal solution will have the nutrients used by the plants at the same ratio they are found in the virgin solution. Any skew from that over time will indicate which concentrations should be adjusted and by how much. It can be a tricky beast, however, as the uptake rate of any individual nutrient is influenced by the presence of other nutrients. Change one and you may create an imbalance elsewhere. Like most things, it's a process.