Designers do thousands of iterations. Why? Well, the next iteration of a design will be better than the last. Right? This is normally the case, but sometimes an iteration is a failure (most of them are). How does one decide if it is a good iteration or not?
Using software, we can only use objective metrics, and each metric has to be applied to each iteration. This creates a context specific to the problem being solved–that is to say, each iteration solving the same problem will inform all other designed iterations of that problem.
I shared this video on my last article, Building Efficiency 101, but wanted to share it here as well. I did 120 iterations within TestFit on the same site to generate some test data.
We have 120 different site design iterations at our disposal. We need to eliminate 90% of them before even looking at them. Below I have plotted all site plans with two criteria: NRSF vs EFF%.
Having all of these site plan options is awesome, but the pro-forma is calling for 350,000 NRSF minimum to pencil. Over 375,000 NRSF will be cost-prohibitive. This gives us a narrow constraint to work with. See plot below.
Now we are down to 12 iterations that will work with pro-forma. What do those 12 spatially look like? Here they are (in order from lowest NRSF to greatest):
Now, I think it is okay to be subjective, after we have complied with objective metrics. This is what designers are great at. Personally, I like iteration 28 (below). It is highly connective, wraps the garage completely, and has a variety of courtyard sizes. They range from grand to intimate. If I was taking this site even further? I might adjust parameters affecting unit yield (see Boosting Unit Yield 101), adjust the parking demand, or slide amenity space to a different corner.
By the way, this blog post is minimum viable generative design. If we were to automate this whole thing with active goals (rather than manual goals) we might have real generative design. Stay tuned!