Most gingerbread bakes soft and chewy already: 23 of the 41 recipes we measured land there, against just 6 crispy. The one change that actually holds up under scrutiny is how you handle the fat and the mixing bowl, not a specific ingredient. Fat and salt point the same direction, but that's a lead, not proof.
Every gingerbread recipe on the internet answers "soft or crispy" with one recipe and calls it decided. We ran the same texture clustering across all 41 recipes and found the split is not close to even. Two process choices below hold up once checked against everything else we tested on this dish. Two ingredient levers point the same way and don't clear that bar yet, and the page says so rather than rounding up.
This is not a preference we assigned. It's how the 41 recipes actually cluster once you measure their edges and middles instead of reading their titles.
| Texture group | Recipes | Share of the corpus |
|---|---|---|
| Chewy | 23 | 56% |
| Cakey | 12 | 29% |
| Crispy | 6 | 15% |
Only 6 of 41 published recipes bake into something we'd call crispy. That is the real content of "soft or crispy": most authors already picked soft, on purpose or not, and a reader who wants the other thing is not choosing between two equally common cookies. They're going after the one almost nobody writes.
Every popular tip gets checked against the same model, and then against every other lever we tested on this dish at once. Fat and salt clear the first test. Not the second.
More fat tracks with a crisper edge in this dish's fitted model. That's a real pattern on its own. Checked against the other 14 lever tests run on this same dish, it stops clearing the bar for genuinely significant.
A directional signal, not a settled finding
Less salt tracks with a crisper edge in this dish's fitted model. That's a real pattern on its own. Checked against the other 14 lever tests run on this same dish, it stops clearing the bar for genuinely significant.
A directional signal, not a settled finding
The one lever on this whole dish that holds up against every other one we tested is leavening, on spread: more of it makes any gingerbread flatter and wider, chewy or crispy alike. That's a spread lever, not a texture one, so it doesn't answer "how do I make this crispier" even though it's the cleanest result in the dataset. Fat and salt are the closest things we have for that question. They just haven't earned the same confidence yet.
We also checked one more thing and came up empty: Whether egg firms or loosens the middle. Egg is one of the axes these recipes vary the most, and none of that variation lined up with a measurable change there, at any level of confidence. Reported here because "we looked and found nothing" is a different, more useful claim than staying silent about it.
Two process choices, not ingredients, are the strongest real findings on this entire dish. Both survive correction for testing 15 different levers at once, which fat and salt above do not.
Measured against 30 recipes on the other side. This is one of only two comparisons on the whole dish that survives correction for testing every lever at once, ingredient and process alike. It's also decided by a 5-recipe minority arm pulled from a wider pooled comparison, not this dish's recipes alone. Both things are true together: real, and thin.
Measured against 30 recipes on the other side. This is one of only two comparisons on the whole dish that survives correction for testing every lever at once, ingredient and process alike. It's also decided by a 5-recipe minority arm pulled from a wider pooled comparison, not this dish's recipes alone. Both things are true together: real, and thin.
| Process | How the corpus splits |
|---|---|
| Fat handling | creamed 36, melted 2, browned 1, cut_in 2 |
| Rest | short 26, long 15 |
| Mixing | standard 39, reverse 2 |
| Shaping | scoop 15, roll 25, pan 1 |
Cut cold fat into the flour instead of creaming it, or mix the dry ingredients in before the wet ones. Both are the two comparisons on this whole dish that hold up once checked against everything else we tested, even though each rests on a handful of recipes. Salt and fat point crisper too, less salt, more fat, but that's a lead worth trying, not a settled result.
This is what most published gingerbread already does, and it's the majority result for a reason. Nothing here needs fixing if soft is what you wanted in the first place.
Only 6 of 41 recipes cluster as crispy. Every comparison involving that group is working with a thin slice of the corpus, and we'd rather say so than dress it up as more certain than it is.
The fat-handling and mixing comparisons above are the only two comparisons on this entire dish that survive correction for testing every lever at once. They are also each decided by a five-recipe minority arm pulled from a wider pooled comparison, not this dish alone. Both facts are true at the same time, and neither cancels the other out.
Both levers clear a raw confidence interval on their own. Neither survives once checked against the other 14 lever tests run on this dish. That doesn't mean they're wrong. It means we don't have enough confidence yet to call them settled the way we can the two process findings above.
These are published recipes by different authors, not one dough baked several ways and compared. A pattern across 41 recipes is real evidence. It is not a controlled experiment.