The process of derandomization is a method used on random-element levels, where the Melinda time would be extremely difficult to achieve, to determine the Melinda time. In a separate, unofficial level, the random elements are either replaced with nonrandom tiles or are simply removed altogether. The resulting score, assuming that this can still be done with the random elements, is the Melinda time.
In extreme cases such as Monster Lab or Lead Us Not into Temptation, a temporary AVI solution is used as the official solution until it can be scored with the real level; as derandomization was not developed until after CCLP2, such measures have only been utilized for the rare CCLP3 levels with random elements.
There are two types of derandomization, depending on the nature of the random elements. For random force floors, they are typically replaced with other sliding tiles which still allow the same boosting overrides. Often ice is put ahead of each force floor, but sometimes force floors against each other are also used. One historical example is Mads' Rush II:
To reach the bold time of 26, Chip would have to be moved R off [2, 3], R or D off [4, 3], and depending on that, either D off [5, 4] or R/D off [4, 5]/[5, 4] (the latter depending on which override off [4, 4] Chip takes), then after hitting the green button, U off [5, 4] and R off [6, 3]. As the chances of this work out to 1 in 512 ([1/4]^4 * 1/2), this was not scored for a long time. When derandomized, the section will look somewhat like this:
Here, the walkers control nothing in the level unless they run into the center (very unlikely), and destroying them would make determining the Melinda time much easier. It was eventually found to be 671, but derandomization might have made it simpler to write the route.