Research
Original Findings
Analyses we ran on our own trait matrix, each one reproducible from the published CSV with the script that generated it.
These are not observations, they are analyses: the underlying facts about each species come from the sources in our source hierarchy, but the structure imposed on them, the questions asked of that structure, and the results below are ours. Each page states its method, its sample size and what the result does not license you to conclude.
Tube Feeders Have a Body-Mass Ceiling
The heaviest bird our matrix records as a tube user weighs about 44 g. A third of feeder birds are heavier, and a tube feeder is functionally invisible to all of them.
AnalysisThree Feeders Reach 93% of Feeder Birds, and a Tube Is Not One
Run as a set-cover problem, the optimal first three feeders are a platform, a suet cage and a nectar feeder. The tube feeder that most beginners buy first ranks fourth.
AnalysisThe Seed Coverage Curve Flattens After Two Foods
Hulled sunflower reaches 59% of feeder birds by itself. Adding insect food takes it to 89%. Every food after that adds single digits.
Reproducing any of this#
Download the CSV. Every finding here is a
group-and-count over it. The exact code lives in the repository that builds this site, in
research/analysis.py, and the extraction rules that turned our species profiles into the
matrix are documented on the methods page. If you get a different
answer, the corrections policy applies to research pages exactly as it
does to guides.