When dealing with estimated data you will often get numbers with way more precision than you need or even trust. Perhaps your cost estimator says a task will cost $81,944,401.0347 but really anything to the left of that first “9” is pretty meaningless as the numbers are just estimates anyway. Dynamic Rounding (or a Floating Round) will take that $81,944,401.0347 and just keep the part we think is useful $81,900,000.0000. These numbers are for one, easier to look at; but more importantly, rounding them in this way can create gigantic gains in both compression and performance, while causing model size to absolutely plummet. The process is simple, but it isn’t entirely intuitive. Luckily, Brian Grant will walk you through the whole thing, step by step.
