How this works
A cookie is a formula, not a page
Recipe sites treat every cookie as a separate document. That makes
them impossible to compare. We treat each one as a set of ratios, which makes them
directly comparable — and makes it possible to ask what actually causes chew.
Step one: put every recipe in the same units
Recipes are written in cups, sticks, dozens and pinches. We resolve every
ingredient line to grams — preferring the author's own gram weights where they
publish them — then express everything as a percentage of the flour weight. That
is the baker's percentage, and it is what lets a recipe making 48 cookies be
compared directly with one making 12.
99.4% of quantified ingredient
lines resolve to grams across all thirty recipes and twenty-five sites.
The two that don't are "1 bag of chips" and "1 standard ice cube".
Step two: find what actually moves the outcome
Reviews are the hidden asset. When someone writes "I used 1½ cups brown
sugar instead of ½ and they came out chewier", that is a controlled
experiment someone else ran for free. We found undefined of them in
7124 reviews, and used them to measure which ratios change which
outcomes.
Only levers whose 95% confidence interval excludes zero appear below.
Each bar is the interval; the tick is the estimate.
brown-sugar share
→ spread
-7.8111
lowers it · n=17
brown-sugar share
→ chew
+2.6741
raises it · n=285
leavener
→ spread
+2.0951
raises it · n=16
leavener
→ spread
+1.8144
raises it · n=46
leavener
→ browning depth
-1.387
lowers it · n=55
salt
→ sweetness perceived
-0.7509
lowers it · n=21
salt
→ edge texture
-0.7389
lowers it · n=339
leavener
→ spread
-0.7119
lowers it · n=242
leavener
→ center texture
-0.3497
lowers it · n=260
egg
→ structural integrity
+0.2134
raises it · n=12
fat
→ browning depth
+0.0914
raises it · n=55
fat
→ spread
+0.0794
raises it · n=24
fat
→ spread
-0.0637
lowers it · n=46
fat
→ edge texture
-0.06
lowers it · n=29
total sugar
→ edge texture
+0.0599
raises it · n=167 · within-recipe
total sugar
→ sweetness perceived
-0.0573
lowers it · n=79 · within-recipe
egg
→ structural integrity
+0.0449
raises it · n=121
egg
→ center texture
+0.0437
raises it · n=152
total sugar
→ spread
+0.0432
raises it · n=16
fat
→ structural integrity
-0.026
lowers it · n=20
total sugar
→ chew
+0.0258
raises it · n=18
fat
→ spread
+0.0254
raises it · n=242 · within-recipe
total sugar
→ edge texture
-0.0251
lowers it · n=27
total sugar
→ sweetness perceived
+0.0236
raises it · n=19
fat
→ center texture
-0.0222
lowers it · n=152
fat
→ spread
+0.0198
raises it · n=365 · within-recipe
total sugar
→ chew
+0.0121
raises it · n=285
mix-ins
→ sweetness perceived
+0.0116
raises it · n=19
fat
→ structural integrity
+0.0106
raises it · n=12
mix-ins
→ sweetness perceived
+0.0088
raises it · n=156
What we can and cannot claim
This matters more than any feature. We tested the engine two ways.
So this site ranks; it never scores. You will never see
"7.2 out of 10 for chew" here, because that number would not be trustworthy. Every
comparison is relative, and relative is what the evidence supports.
Honest gaps
- No egg-free, dairy-free or gluten-free options. All thirty
recipes contain all three. Those filters correctly return nothing.
- Calories cover 21 of 30. Nine sites publish no nutrition data.
- Reviews skew positive. Only the 7–20 reviews embedded in each
page are readable, and they lean five-star — so anything derived from failure
rates is optimistic.
- Costs are estimates, running about 14% below the one site
that publishes real per-ingredient prices.