Skip to the method

A playful experience. An open explanation.

The story is yours.
The method should be clear.

The 27-question Kolache tradition pilot describes selected food, language, occasion, and family-transmission answers. It offers questions worth asking and a reason to preserve a recipe. It has not been validated as a way to identify a person’s family origins.

“Kolache DNA” is a metaphor for food traditions. This is not a genetic test. No DNA is collected or analyzed. Neither the free reading nor the paid Field Guide establishes ancestry, ethnicity, nationality, biological relationship, or a geographic origin. A purchase adds explanation, interview prompts, and a worksheet; it does not make the model more accurate.

What an answer does

Answer for one baker or one family branch where possible. Each selected option has zero or more small, preassigned integer weights. “Not sure / it varied” adds no points. Some answers describe transmission or select a playful archetype without adding regional points.

The weights were assigned by hand. They are hypotheses, not estimates learned from a representative sample. They have not been calibrated against documented ancestral origins. The presence of a trait in a regional tradition does not establish that a family with that trait came from that region.

Two channels, kept separate

Old-country tradition profiles: Valašsko, Slovácko, Haná, Lašsko and Czech Silesia, South Bohemia, West Bohemia/Chodsko, and Central/East Bohemia.

U.S. diaspora chapter: Texas/Oklahoma, the Midwest lodge belt, and the Eastern Corridor. These broad profiles summarize selected food and language answers along with self-reported settlement. They do not independently discover migration history.

The question about where your family settled in America (D5a) contributes only to the U.S. channel. Changing that answer cannot change any old-country points, shares, or the minimum-signal decision. U.S. points are never included in the denominator for old-country shares.

When a profile is withheld

The pilot needs at least three separate answers that add old-country points and five total old-country points before it shows a leading old-country profile. If either condition fails, it withholds that profile and its percentage shares. This is a conservative product rule, not an empirically validated threshold.

Answering three questions does not necessarily meet the regional threshold. A paid guide can still explain the answers and provide documentation prompts when no old-country profile is reported. The purchase path requires at least three selected, non-skipped answers so you can review whether the available personalization is useful before buying.

What the percentages mean

For each old-country profile, add the weights attached to the selected options. Divide that profile’s points by the sum of all old-country points, then multiply by 100. The result is a share of assigned points, not an ancestry percentage, probability, likelihood, population frequency, or measure of authenticity.

For example, 6 of 20 assigned points would display as 30%. It would not mean a 30% chance of an origin. One answer can add points to multiple profiles, and the profiles do not have equal opportunities to gain points. A zero means no points from those answers, not evidence against a connection.

The Field Guide shows all seven raw totals, the contributing answers, and exact ties at the top. The free reading uses a fixed profile order to select a display label when scores tie; that order adds no evidence. Shares are withheld below the minimum-signal rule and may not add to exactly 100% after rounding.

Shape and archetype are descriptions

These describe the geometry of the assigned points, never confidence in an origin. U.S. display bands are “visible” at 2–3 points and “strong” at 4 or more; those are also hand-set display rules. Archetypes such as “The Keeper of the Card” are playful summaries, not scientific classes.

What the pilot cannot account for

Recipes travel through adoption, intermarriage, neighbors, bakeries, church and lodge communities, commercial products, and migration. Memory can be uncertain. Families can combine several traditions. The model does not account for these processes in a statistically defensible way.

Its broad cultural labels are not biological populations. There has been no representative validation study, calibration study, subgroup fairness assessment, repeatability study, or demonstrated predictive performance. Festival participants are a self-selected convenience sample; their answers cannot represent all Czech, Czech-American, or other families. The Czech translation has not yet received independent native-editor approval.

Testing the code can establish that the English, Czech, server, and report implementations agree. It cannot establish that the cultural hypotheses are true. Before making research findings, the project would need a documented study protocol, appropriate consent and sampling, independently supported labels, held-out evaluation, and qualified historical, genealogical, language, and methods review.

What gets sent or saved

The free reading can be calculated on your device without an email address or submission. Asking for a result email sends the versioned answer IDs to the server so it can independently recompute that result. Separate permission is needed for genealogy follow-up; a result email or purchase does not opt you into marketing.

A paid Field Guide needs your answers and purchase information to produce and provide the purchased guide. Its checkout notice and privacy policy explain that separate fulfillment purpose and retention. It is not Census research consent. Census research collection remains closed. Private report access is available for one year; print or save a personal copy if you want to keep it longer.

Context is not individual proof

Regional food traditions and word differences can be documented without establishing a diagnostic test. For example, Valašský frgál’s protected geographical indication is documented in Commission Implementing Regulation (EU) No 1263/2013. The Czech Linguistic Atlas documents geographic variation in crumb-topping vocabulary. These sources do not validate this pilot’s weights or an inference about an individual family.

Versioned and inspectable

Instrument: KAT-2026-08-pilot-1.
Scoring model: KAT-score-2026-08-pilot-2.

The Field Guide includes each selected answer and its point contributions. The complete English question and weight register is also available to inspect. Question order, option order, weights, thresholds, and archetype rules are versioned; public-facing explanations must match those versions.

Explore a fictional guide before you decide to buy. For questions about the project or your data, write to hello@ajhpress.com.

Start my free readingSee the sample