Some things I would do differently:
1. I noticed the first paintings quickly became somewhat unnecessary, not helped by the fact that my main concern at the phase was to test the overall direction, which was easily checkable in the next phase. One basic house concept would have done just as well.
2. I would spend more time on the initial sketching phase, not only finding interesting elements, but also solving how to use them in the architecture. This would have allowed me to incorporate them earlier in the base models. Even if the initial building would look very cluttered, the unnecessary elements would be quicker to paint away when doing paintovers, compared to adding them to the scenes later. Philosophy would be: Prepare the base models for maximum amount of happy accidents, and deal the unhappy ones later.
3. Another time saver: I would spend a bit more time on the models that were on the generic side (tables, pots, people…), as the possible design changes in them wouldn't affect the scene much, but did now produce extra cleanup work later.
4. My initial plan for final images was too vague: "Get a feeling of the culture's look". Rather, I would set specific scenes that would test the needed variety of locations for the culture.
5. Pay more attention to scale in the beginning. Don't rush: measure twice, boolean once.