# Breeders in the east, finishers in the west **Answer.** **Maldonado** is the most breeding-heavy department: 71.7% of its beef herd is breeding cows and calves. **Soriano** leans hardest to finishing: 29.8% steers and fattening cows. The east breeds; the west finishes. Notebook: https://cow-tools.sophia-labs.com/notebooks/herd-structure/herd-structure.ipynb · Page: https://cow-tools.sophia-labs.com/notebooks/herd-structure/index.html · Data: Uruguay MGAP/SNIG DICOSE 2024–2025 (graph `uruguay-cows` at https://cow-tools.sophia-labs.com/mcp). ## The notebook A beef herd is a pipeline: breeding cows have calves, calves become yearlings, steers are fattened and sold. Regions specialise. Where the land is poor, farmers keep cows and sell the calves (*cría*); where it is better, they buy those calves and finish them (*invernada*). The declaration's nine beef categories show which is which, department by department. The categories (`codigoscategoria.csv`, species 1): 1 bulls, 2 breeding cows (mated), 3 cows being fattened, 4–6 steers by age, 7–8 heifers, 9 calves. We fold them into four groups so the chart stays legible: **breeding** (2 breeding cows + 9 calves), **steers and fattening cows** (3, 4, 5, 6), **heifers** (7, 8) and **bulls** (1). ```python # Setup. import altair as alt import pandas as pd import cowtools_client as cw import cowtools_charts as cc names = cc.department_names() GROUP = {"1": "bulls", "2": "breeding", "9": "breeding", "3": "steers & fattening cows", "4": "steers & fattening cows", "5": "steers & fattening cows", "6": "steers & fattening cows", "7": "heifers", "8": "heifers"} ``` ```python # The query: cattle (beef and dairy) by category and department, 2025, with the beef category labels. allcat = cw.sparql(""" SELECT ?department ?species ?category (SUM(?n) AS ?animals) WHERE { GRAPH ?g { ?obs a cow:AnimalCategoryStockObservation ; cow:year "2025"^^xsd:gYear ; cow:department ?department ; cow:species ?species ; cow:animalCategory ?category ; cow:animalsCount ?n . VALUES ?species { } } } GROUP BY ?department ?species ?category """) cat = allcat[allcat["species"] == "urn:sophia:cow:species/1"].copy() labels = cw.sparql(""" SELECT DISTINCT ?category ?label WHERE { GRAPH ?g { ?category skos:prefLabel ?label } FILTER(STRSTARTS(STR(?category), "urn:sophia:cow:animal-category/1/")) } """) labels["code"] = labels["category"].map(cw.dept_code) labels = labels.drop_duplicates("code").sort_values("code", key=lambda s: s.astype(int)) cc.show_table(labels[["code", "label"]], name="beef-categories") ``` | code | label | |-------:|:-------------------------------------------| | 1 | 1) TOROS | | 2 | 2) VACAS DE CRIA (ENTORADAS) | | 3 | 3) VACAS DE INVERNADA | | 4 | 4) NOVILLOS DE MAS DE 3 AÑOS | | 5 | 5) NOVILLOS DE 2 A 3 AÑOS | | 6 | 6) NOVILLOS DE 1 A 2 AÑOS | | 7 | 7) VAQUILLONA DE MAS DE 2 AÑOS SIN ENTORAR | | 8 | 8) VAQUILLONAS DE 1 A 2 AÑOS | | 9 | 9) TERNEROS/TERNERAS | ```python cat["code"] = cat["department"].map(cw.dept_code) cat["cat_code"] = cat["category"].map(cw.dept_code) cat["group"] = cat["cat_code"].map(GROUP) cw.check("every beef category maps to a group", int(cat["group"].isna().sum()), 0, source="the 9 beef categories") wide = cat.pivot_table(index="code", columns="group", values="animals", aggfunc="sum").fillna(0) share = wide.div(wide.sum(axis=1), axis=0) share["name"] = share.index.map(names) share = share.sort_values("breeding", ascending=False) cc.show_table(share[["name", "breeding", "steers & fattening cows", "heifers", "bulls"]].reset_index(drop=True), floatfmt=".3f", name="group-shares") ``` ```text PASS every beef category maps to a group: 0 vs 0 (exactly) — the 9 beef categories ``` | name | breeding | steers & fattening cows | heifers | bulls | |:---------------|-----------:|--------------------------:|----------:|--------:| | Maldonado | 0.717 | 0.132 | 0.131 | 0.021 | | Treinta y Tres | 0.698 | 0.153 | 0.132 | 0.017 | | Rocha | 0.678 | 0.175 | 0.129 | 0.018 | | Florida | 0.677 | 0.149 | 0.157 | 0.017 | | Lavalleja | 0.673 | 0.178 | 0.131 | 0.018 | | Rivera | 0.67 | 0.167 | 0.146 | 0.016 | | Colonia | 0.657 | 0.183 | 0.149 | 0.012 | | San José | 0.652 | 0.171 | 0.165 | 0.011 | | Cerro Largo | 0.649 | 0.205 | 0.13 | 0.016 | | Durazno | 0.649 | 0.205 | 0.129 | 0.017 | | Salto | 0.646 | 0.196 | 0.14 | 0.018 | | Artigas | 0.644 | 0.211 | 0.129 | 0.015 | | Paysandú | 0.637 | 0.213 | 0.133 | 0.017 | | Tacuarembó | 0.62 | 0.219 | 0.145 | 0.016 | | Flores | 0.606 | 0.232 | 0.148 | 0.015 | | Canelones | 0.601 | 0.226 | 0.159 | 0.015 | | Río Negro | 0.583 | 0.262 | 0.142 | 0.013 | | Montevideo | 0.581 | 0.208 | 0.193 | 0.018 | | Soriano | 0.527 | 0.298 | 0.16 | 0.014 | ### Cross-check against the source ```python csv = cw.source_csv(2025, "datosanimalesdetallados.csv") beef = csv[csv["EspecieCodigo"] == "1"] src = cw.num(beef["AnimalesPorCategoriaPD_AD"]).groupby([beef["DepartamentoCodigo"], beef["CategoriaCodigo"]]).sum() cube = cat.set_index(["code", "cat_code"])["animals"] diff = (cube - src.reindex(cube.index)).abs().max() cw.check("2025 beef cattle by department × category: cube = CSV (largest difference)", int(diff), 0, source="datosanimalesdetallados.csv 2025, AnimalesPorCategoriaPD_AD") cw.check("2025 beef categories sum to the beef total in datosanimales", int(cat["animals"].sum()), 11_583_321, source="notebook national-herd: beef cattle 2025") ``` ```text PASS 2025 beef cattle by department × category: cube = CSV (largest difference): 0 vs 0 (exactly) — datosanimalesdetallados.csv 2025, AnimalesPorCategoriaPD_AD PASS 2025 beef categories sum to the beef total in datosanimales: 11,583,321 vs 11,583,321 (exactly) — notebook national-herd: beef cattle 2025 ``` The cube also materialises a composition at ingest, the derived metric `category_share_of_bovine`. Its denominator is a department's **whole** cattle herd, beef and dairy together (its shares sum to 1 across both species' categories), where ours is the beef herd alone. Put on its denominator, our numbers must reproduce it for every department and category, to the six decimals it keeps. ```python derived = cw.sparql(""" SELECT ?department ?category ?value WHERE { GRAPH ?g { ?d a cow:DerivedMetricObservation ; cow:year "2025"^^xsd:gYear ; cow:metric cow:metric_category_share_of_bovine ; cow:department ?department ; cow:animalCategory ?category ; cow:metricValue ?value . } } """) whole = allcat.assign(code=allcat["department"].map(cw.dept_code)).groupby("code")["animals"].sum() ours_all = allcat.assign(code=allcat["department"].map(cw.dept_code)) ours_all["share"] = ours_all["animals"] / ours_all["code"].map(whole) mine = ours_all.set_index(["code", "category"])["share"] theirs = derived.assign(code=derived["department"].map(cw.dept_code)).set_index(["code", "category"])["value"].astype(float) cw.check("derived rows cover every department × category", len(theirs), len(mine), source="cube derived metric") gap = float((theirs - mine.reindex(theirs.index)).abs().max()) cw.check("category_share_of_bovine (derived, 6 dp) = our shares on its denominator (largest gap)", round(gap, 7), 0.0, tol=5e-7, source="cow:metric_category_share_of_bovine, materialised at ingest") ``` ```text PASS derived rows cover every department × category: 304 vs 304 (exactly) — cube derived metric PASS category_share_of_bovine (derived, 6 dp) = our shares on its denominator (largest gap): 5e-07 vs 0 (within ±5e-07) — cow:metric_category_share_of_bovine, materialised at ingest ``` ```python east = share.iloc[0] west = share.sort_values("steers & fattening cows", ascending=False).iloc[0] # The sentence below says "east" and "west": hold it to that. EAST = {"Maldonado", "Rocha", "Treinta y Tres", "Lavalleja", "Cerro Largo"} WEST = {"Soriano", "Río Negro", "Colonia", "Paysandú", "San José"} cw.check("the most breeding-heavy department is in the east", east["name"] in EAST, True, source=f"{sorted(EAST)}") cw.check("the most finishing-heavy department is in the west", west["name"] in WEST, True, source=f"{sorted(WEST)}") cw.finding("headline", f"**{east['name']}** is the most breeding-heavy department: {east['breeding']:.1%} of its beef herd is " f"breeding cows and calves. **{west['name']}** leans hardest to finishing: " f"{west['steers & fattening cows']:.1%} steers and fattening cows. The east breeds; the west finishes.") ``` ```text PASS the most breeding-heavy department is in the east: True vs True (exactly) — ['Cerro Largo', 'Lavalleja', 'Maldonado', 'Rocha', 'Treinta y Tres'] PASS the most finishing-heavy department is in the west: True vs True (exactly) — ['Colonia', 'Paysandú', 'Río Negro', 'San José', 'Soriano'] ``` > **Finding.** **Maldonado** is the most breeding-heavy department: 71.7% of its beef herd is breeding cows and calves. **Soriano** leans hardest to finishing: 29.8% steers and fattening cows. The east breeds; the west finishes. ```python long = share.reset_index().melt(id_vars=["code", "name"], value_vars=["breeding", "heifers", "bulls", "steers & fattening cows"], var_name="group", value_name="share") order = share["name"].tolist() stack = alt.Chart(long, title=alt.Title("Beef herd composition by department, 2025", subtitle="share of beef cattle (PD_AD); departments ordered by breeding share")).mark_bar(height=14).encode( y=alt.Y("name:N", sort=order, title=None), x=alt.X("share:Q", stack="normalize", title=None, axis=alt.Axis(format=".0%")), color=alt.Color("group:N", scale=alt.Scale(domain=["breeding", "heifers", "bulls", "steers & fattening cows"], range=[cc.CATTLE, cc.SERIES[3], cc.MUTED, cc.SHEEP]), title=None), order=alt.Order("group_order:Q"), tooltip=["name:N", "group:N", alt.Tooltip("share:Q", format=".1%")], ).transform_calculate(group_order="indexof(['breeding','heifers','bulls','steers & fattening cows'], datum.group)").properties(width=500, height=380) cc.show(stack, "composition", alt_text="Stacked bars per department: eastern departments (Maldonado, Treinta y Tres, Rocha) have the largest breeding share; Soriano and Río Negro the largest share of steers.") ``` ![Stacked bars per department: eastern departments (Maldonado, Treinta y Tres, Rocha) have the largest breeding share; Soriano and Río Negro the largest share of steers.](charts/composition.svg) Chart data and encoding: [charts/composition.vl.json](charts/composition.vl.json) (Vega-Lite). ```python share["steers_per_breeding"] = share["steers & fattening cows"] / share["breeding"] cc.show(cc.choropleth(share.reset_index(), "steers_per_breeding", title="Steers and fattening cows per breeding animal, 2025", subtitle="(categories 3–6) ÷ (breeding cows + calves); higher = more finishing", legend="ratio", fmt=".2f"), "finishing-map", thumbnail=True, alt_text="Map: the finishing ratio is highest in the littoral west (Soriano, Río Negro) and lowest in Maldonado.") print(cc.map_credit()) ``` ![Map: the finishing ratio is highest in the littoral west (Soriano, Río Negro) and lowest in Maldonado.](charts/finishing-map.svg) Chart data and encoding: [charts/finishing-map.vl.json](charts/finishing-map.vl.json) (Vega-Lite). ```text Map: Natural Earth 1:10m admin-1 (public domain, v5.1.2), joined on the DICOSE department code. Made with Natural Earth. ``` ## Where every number came from ```python cc.show_table(cw.checks()[["check", "got", "expected", "rule", "source"]], name="checks") cw.provenance() ``` | check | got | expected | rule | source | |:---------------------------------------------------------------------------------------|--------------:|--------------:|:--------------|:---------------------------------------------------------------------| | every beef category maps to a group | 0 | 0 | exactly | the 9 beef categories | | 2025 beef cattle by department × category: cube = CSV (largest difference) | 0 | 0 | exactly | datosanimalesdetallados.csv 2025, AnimalesPorCategoriaPD_AD | | 2025 beef categories sum to the beef total in datosanimales | 1.15833e+07 | 1.15833e+07 | exactly | notebook national-herd: beef cattle 2025 | | derived rows cover every department × category | 304 | 304 | exactly | cube derived metric | | category_share_of_bovine (derived, 6 dp) = our shares on its denominator (largest gap) | 5e-07 | 0 | within ±5e-07 | cow:metric_category_share_of_bovine, materialised at ingest | | the most breeding-heavy department is in the east | 1 | 1 | exactly | ['Cerro Largo', 'Lavalleja', 'Maldonado', 'Rocha', 'Treinta y Tres'] | | the most finishing-heavy department is in the west | 1 | 1 | exactly | ['Colonia', 'Paysandú', 'Río Negro', 'San José', 'Soriano'] | ```text kind target rows \ 0 sparql uruguay-cows 304 1 sparql uruguay-cows 9 2 sparql uruguay-cows 304 first line 0 SELECT ?department ?species ?category (SUM(?n)... 1 SELECT DISTINCT ?category ?label 2 SELECT ?department ?category ?value ``` ```python cw.save() ```