[{"data":1,"prerenderedAt":1008},["ShallowReactive",2],{"article-alternates":3,"article-\u002Fes\u002Fmarketing\u002Fprueba-bayesiana-ab-test-toma-de-decisiones-rapida":13},{"i18nKey":4,"paths":5},"marketing-002-2026-08",{"de":6,"en":7,"es":8,"fr":9,"it":10,"ru":11,"tr":12},"\u002Fde\u002Fmarketing\u002Fbayesian-ab-test-schnelle-entscheidungen","\u002Fen\u002Fmarketing\u002Fbayesian-ab-testing-fast-decisions","\u002Fes\u002Fmarketing\u002Fprueba-bayesiana-ab-test-toma-de-decisiones-rapida","\u002Ffr\u002Fmarketing\u002Fdecision-making-with-bayesian-ab-testing","\u002Fit\u002Fmarketing\u002Fbayesian-ab-testing-per-decisioni-rapide","\u002Fru\u002Fmarketing\u002Fbayesian-ab-test-hizli-karar-verme","\u002Ftr\u002Fmarketing\u002Fbayesian-a-b-test-ile-hizli-karar-verme",{"_path":8,"_dir":14,"_draft":15,"_partial":15,"_locale":16,"title":17,"description":18,"publishedAt":19,"modifiedAt":19,"category":14,"i18nKey":4,"tags":20,"readingTime":26,"author":27,"body":28,"_type":1002,"_id":1003,"_source":1004,"_file":1005,"_stem":1006,"_extension":1007},"marketing",false,"","Prueba Bayesiana A\u002FB: Toma de Decisiones Rápida","Más allá del p\u003C0.05 frequentista: muestreo secuencial, parada temprana y cuantificación de incertidumbre. Guía para acelerar decisiones en marketing de performance.","2026-08-13",[21,22,23,24,25],"prueba-bayesiana","ab-test","optimizacion-conversion","estadistica-frequentista","muestreo-secuencial",8,"Roibase",{"type":29,"children":30,"toc":994},"root",[31,39,46,51,56,62,67,81,397,409,415,420,425,441,447,468,473,560,565,571,576,581,586,972,977,983,988],{"type":32,"tag":33,"props":34,"children":35},"element","p",{},[36],{"type":37,"value":38},"text","En marketing de performance, las pruebas A\u002FB todavía siguen la metodología frequentista de 2010: cálculo de tamaño muestral fijo, umbral p\u003C0.05, espera a que sea \"significativo\". Testeas tres creativos en Meta Ads, uno pierda claramente, pero \"sin suficientes datos\" quemas presupuesto dos semanas más. La prueba Bayesiana rompe este ciclo: te da derecho a parada temprana, cuantifica incertidumbre y dice \"la variante B gana con probabilidad 94%\". Con Google Optimize descontinuado, si construyes tu propio stack de testing, la matemática Bayesiana acelera decisiones.",{"type":32,"tag":40,"props":41,"children":43},"h2",{"id":42},"las-reglas-fijas-del-testing-frequentista",[44],{"type":37,"value":45},"Las Reglas Fijas del Testing Frequentista",{"type":32,"tag":33,"props":47,"children":48},{},[49],{"type":37,"value":50},"El test A\u002FB clásico funciona así: precalcula tamaño muestral (análisis de poder: 80% poder, 5% alfa, 10% lift esperado), espera a alcanzar esa cantidad, revisa p-valor, decide. El problema: en la realidad el lift es 3%, no 10%; el tamaño muestral se extiende de 2 semanas a 8 semanas. Durante ese tiempo el creativo se fatiga, cambios estacionales afectan, el CPM sube 40%. En frecuentismo, revisar temprano está prohibido — el \"peeking\" infla el error tipo 1. Aunque hagas testing secuencial, las funciones de gasto alfa (Bonferroni, O'Brien-Fleming) añaden complejidad y exigen umbrales rígidos.",{"type":32,"tag":33,"props":52,"children":53},{},[54],{"type":37,"value":55},"Escenario de e-commerce: el control tiene 2.1% CR, el flow de checkout nuevo 2.3%. Después de 1000 sesiones: 9.5% lift pero p=0.12. Frequentista dice: \"no significativo, continúa\". En 2000 sesiones p=0.08, aún insuficiente. En 3500 sesiones p=0.047, ahora es significativo. Pero para entonces la variante B lleva 3 semanas en vivo, la estación cambió, estimar ganancia es imposible. La matemática frequentista da veredicto binario: significativo o no. Existe un intervalo de confianza, pero se usa solo como \"necesito IC 95% para decidir\", nada más.",{"type":32,"tag":40,"props":57,"children":59},{"id":58},"distribución-de-probabilidad-en-el-enfoque-bayesiano",[60],{"type":37,"value":61},"Distribución de Probabilidad en el Enfoque Bayesiano",{"type":32,"tag":33,"props":63,"children":64},{},[65],{"type":37,"value":66},"Bayesian plantea otra pregunta: \"¿Cuál es la probabilidad de que B sea mejor que A?\" La respuesta es una distribución posterior que se actualiza continuamente. Creencia previa (prior) + datos = posterior. Con cada nueva sesión, el posterior se recalcula. 100 sesiones: probabilidad de ganancia 72%, 500: 88%, 1000: 94%. Sin umbral fijo; tú decides: ¿basta 90%, o espero 95%?",{"type":32,"tag":33,"props":68,"children":69},{},[70,72,79],{"type":37,"value":71},"Matemática: modelo beta-binomial. El prior para conversion rate es Beta(α=1, β=1) (uniforme); cada conversión suma +1 a α, cada no-conversión suma +1 a β. Posterior es Beta(α + conversiones, β + no-conversiones). Para dos variantes tienes dos distribuciones beta; con Monte Carlo extrae 10000 muestras y cuenta \"B > A\". Python: ",{"type":32,"tag":73,"props":74,"children":76},"code",{"className":75},[],[77],{"type":37,"value":78},"scipy.stats.beta.rvs",{"type":37,"value":80},". En BigQuery es con UDF, pero Python es más rápido para sampling.",{"type":32,"tag":82,"props":83,"children":87},"pre",{"className":84,"code":85,"language":86,"meta":16,"style":16},"language-python shiki shiki-themes github-dark","from scipy.stats import beta\n\n# Variante A: 50 conversiones, 2000 impresiones\na_alpha, a_beta = 1 + 50, 1 + (2000 - 50)\n# Variante B: 58 conversiones, 2000 impresiones\nb_alpha, b_beta = 1 + 58, 1 + (2000 - 58)\n\nsamples_a = beta.rvs(a_alpha, a_beta, size=10000)\nsamples_b = beta.rvs(b_alpha, b_beta, size=10000)\n\nprob_b_wins = (samples_b > samples_a).mean()\n# Resultado: 0.847 → probabilidad de ganancia 84.7%\n","python",[88],{"type":32,"tag":73,"props":89,"children":90},{"__ignoreMap":16},[91,119,129,139,207,216,274,282,318,352,360,388],{"type":32,"tag":92,"props":93,"children":96},"span",{"class":94,"line":95},"line",1,[97,103,109,114],{"type":32,"tag":92,"props":98,"children":100},{"style":99},"--shiki-default:#F97583",[101],{"type":37,"value":102},"from",{"type":32,"tag":92,"props":104,"children":106},{"style":105},"--shiki-default:#E1E4E8",[107],{"type":37,"value":108}," scipy.stats ",{"type":32,"tag":92,"props":110,"children":111},{"style":99},[112],{"type":37,"value":113},"import",{"type":32,"tag":92,"props":115,"children":116},{"style":105},[117],{"type":37,"value":118}," beta\n",{"type":32,"tag":92,"props":120,"children":122},{"class":94,"line":121},2,[123],{"type":32,"tag":92,"props":124,"children":126},{"emptyLinePlaceholder":125},true,[127],{"type":37,"value":128},"\n",{"type":32,"tag":92,"props":130,"children":132},{"class":94,"line":131},3,[133],{"type":32,"tag":92,"props":134,"children":136},{"style":135},"--shiki-default:#6A737D",[137],{"type":37,"value":138},"# Variante A: 50 conversiones, 2000 impresiones\n",{"type":32,"tag":92,"props":140,"children":142},{"class":94,"line":141},4,[143,148,153,159,164,169,174,179,183,188,193,198,202],{"type":32,"tag":92,"props":144,"children":145},{"style":105},[146],{"type":37,"value":147},"a_alpha, a_beta ",{"type":32,"tag":92,"props":149,"children":150},{"style":99},[151],{"type":37,"value":152},"=",{"type":32,"tag":92,"props":154,"children":156},{"style":155},"--shiki-default:#79B8FF",[157],{"type":37,"value":158}," 1",{"type":32,"tag":92,"props":160,"children":161},{"style":99},[162],{"type":37,"value":163}," +",{"type":32,"tag":92,"props":165,"children":166},{"style":155},[167],{"type":37,"value":168}," 50",{"type":32,"tag":92,"props":170,"children":171},{"style":105},[172],{"type":37,"value":173},", ",{"type":32,"tag":92,"props":175,"children":176},{"style":155},[177],{"type":37,"value":178},"1",{"type":32,"tag":92,"props":180,"children":181},{"style":99},[182],{"type":37,"value":163},{"type":32,"tag":92,"props":184,"children":185},{"style":105},[186],{"type":37,"value":187}," (",{"type":32,"tag":92,"props":189,"children":190},{"style":155},[191],{"type":37,"value":192},"2000",{"type":32,"tag":92,"props":194,"children":195},{"style":99},[196],{"type":37,"value":197}," -",{"type":32,"tag":92,"props":199,"children":200},{"style":155},[201],{"type":37,"value":168},{"type":32,"tag":92,"props":203,"children":204},{"style":105},[205],{"type":37,"value":206},")\n",{"type":32,"tag":92,"props":208,"children":210},{"class":94,"line":209},5,[211],{"type":32,"tag":92,"props":212,"children":213},{"style":135},[214],{"type":37,"value":215},"# Variante B: 58 conversiones, 2000 impresiones\n",{"type":32,"tag":92,"props":217,"children":219},{"class":94,"line":218},6,[220,225,229,233,237,242,246,250,254,258,262,266,270],{"type":32,"tag":92,"props":221,"children":222},{"style":105},[223],{"type":37,"value":224},"b_alpha, b_beta ",{"type":32,"tag":92,"props":226,"children":227},{"style":99},[228],{"type":37,"value":152},{"type":32,"tag":92,"props":230,"children":231},{"style":155},[232],{"type":37,"value":158},{"type":32,"tag":92,"props":234,"children":235},{"style":99},[236],{"type":37,"value":163},{"type":32,"tag":92,"props":238,"children":239},{"style":155},[240],{"type":37,"value":241}," 58",{"type":32,"tag":92,"props":243,"children":244},{"style":105},[245],{"type":37,"value":173},{"type":32,"tag":92,"props":247,"children":248},{"style":155},[249],{"type":37,"value":178},{"type":32,"tag":92,"props":251,"children":252},{"style":99},[253],{"type":37,"value":163},{"type":32,"tag":92,"props":255,"children":256},{"style":105},[257],{"type":37,"value":187},{"type":32,"tag":92,"props":259,"children":260},{"style":155},[261],{"type":37,"value":192},{"type":32,"tag":92,"props":263,"children":264},{"style":99},[265],{"type":37,"value":197},{"type":32,"tag":92,"props":267,"children":268},{"style":155},[269],{"type":37,"value":241},{"type":32,"tag":92,"props":271,"children":272},{"style":105},[273],{"type":37,"value":206},{"type":32,"tag":92,"props":275,"children":277},{"class":94,"line":276},7,[278],{"type":32,"tag":92,"props":279,"children":280},{"emptyLinePlaceholder":125},[281],{"type":37,"value":128},{"type":32,"tag":92,"props":283,"children":284},{"class":94,"line":26},[285,290,294,299,305,309,314],{"type":32,"tag":92,"props":286,"children":287},{"style":105},[288],{"type":37,"value":289},"samples_a ",{"type":32,"tag":92,"props":291,"children":292},{"style":99},[293],{"type":37,"value":152},{"type":32,"tag":92,"props":295,"children":296},{"style":105},[297],{"type":37,"value":298}," beta.rvs(a_alpha, a_beta, ",{"type":32,"tag":92,"props":300,"children":302},{"style":301},"--shiki-default:#FFAB70",[303],{"type":37,"value":304},"size",{"type":32,"tag":92,"props":306,"children":307},{"style":99},[308],{"type":37,"value":152},{"type":32,"tag":92,"props":310,"children":311},{"style":155},[312],{"type":37,"value":313},"10000",{"type":32,"tag":92,"props":315,"children":316},{"style":105},[317],{"type":37,"value":206},{"type":32,"tag":92,"props":319,"children":321},{"class":94,"line":320},9,[322,327,331,336,340,344,348],{"type":32,"tag":92,"props":323,"children":324},{"style":105},[325],{"type":37,"value":326},"samples_b ",{"type":32,"tag":92,"props":328,"children":329},{"style":99},[330],{"type":37,"value":152},{"type":32,"tag":92,"props":332,"children":333},{"style":105},[334],{"type":37,"value":335}," beta.rvs(b_alpha, b_beta, ",{"type":32,"tag":92,"props":337,"children":338},{"style":301},[339],{"type":37,"value":304},{"type":32,"tag":92,"props":341,"children":342},{"style":99},[343],{"type":37,"value":152},{"type":32,"tag":92,"props":345,"children":346},{"style":155},[347],{"type":37,"value":313},{"type":32,"tag":92,"props":349,"children":350},{"style":105},[351],{"type":37,"value":206},{"type":32,"tag":92,"props":353,"children":355},{"class":94,"line":354},10,[356],{"type":32,"tag":92,"props":357,"children":358},{"emptyLinePlaceholder":125},[359],{"type":37,"value":128},{"type":32,"tag":92,"props":361,"children":363},{"class":94,"line":362},11,[364,369,373,378,383],{"type":32,"tag":92,"props":365,"children":366},{"style":105},[367],{"type":37,"value":368},"prob_b_wins ",{"type":32,"tag":92,"props":370,"children":371},{"style":99},[372],{"type":37,"value":152},{"type":32,"tag":92,"props":374,"children":375},{"style":105},[376],{"type":37,"value":377}," (samples_b ",{"type":32,"tag":92,"props":379,"children":380},{"style":99},[381],{"type":37,"value":382},">",{"type":32,"tag":92,"props":384,"children":385},{"style":105},[386],{"type":37,"value":387}," samples_a).mean()\n",{"type":32,"tag":92,"props":389,"children":391},{"class":94,"line":390},12,[392],{"type":32,"tag":92,"props":393,"children":394},{"style":135},[395],{"type":37,"value":396},"# Resultado: 0.847 → probabilidad de ganancia 84.7%\n",{"type":32,"tag":33,"props":398,"children":399},{},[400,402,407],{"type":37,"value":401},"Este resultado va a tu dashboard diario: \"B gana con probabilidad 84.7%, lift esperado 15.3%, intervalo credible 95% ",{"type":32,"tag":92,"props":403,"children":404},{},[405],{"type":37,"value":406},"2.1%, 29.8%",{"type":37,"value":408},"\". No entras en el dilema \"¿es significativo o no?\", entregas una medida de riesgo. Si 85% de probabilidad es suficiente, detén; si no, continúa. Decisión secuencial — cada día reevaluación.",{"type":32,"tag":40,"props":410,"children":412},{"id":411},"muestreo-secuencial-y-criterio-de-parada-temprana",[413],{"type":37,"value":414},"Muestreo Secuencial y Criterio de Parada Temprana",{"type":32,"tag":33,"props":416,"children":417},{},[418],{"type":37,"value":419},"La verdadera fortaleza Bayesiana: puedes detener el test cuando quieras. En frequentismo, el peeking prohíbe esto porque infla error tipo 1 a largo plazo; en Bayesian no existe ese concepto (actualización de creencia en lugar de frecuencias a largo plazo). El criterio de parada lo estableces tú: \"Si probabilidad de ganancia >95% o \u003C5%, detén\". Con este criterio, el tamaño muestral promedio cae 30-50% (según benchmark de VWO 2024).",{"type":32,"tag":33,"props":421,"children":422},{},[423],{"type":37,"value":424},"Pero atención: revisar muy pronto también engaña. Con 50 sesiones puede haber probabilidad de ganancia 98% por fluctuación aleatoria. Aquí entra la minimización de arrepentimiento Bayesiano: calculas expected value of information (EVOI). EVOI = (ganancia esperada) - (costo de continuar test). Si EVOI es negativo, detén. Enfoque práctico: mantén tamaño muestral mínimo (p.ej. 500 impresiones\u002Fvariante), luego aplica stopping rule Bayesiana.",{"type":32,"tag":33,"props":426,"children":427},{},[428,430,439],{"type":37,"value":429},"En ",{"type":32,"tag":431,"props":432,"children":436},"a",{"href":433,"rel":434},"https:\u002F\u002Fwww.roibase.com.tr\u002Fes\u002Fcro",[435],"nofollow",[437],{"type":37,"value":438},"Optimización de Tasa de Conversión",{"type":37,"value":440}," con test de creativos Meta Ads: 3 variantes, cada una $100\u002Fdía de presupuesto. Día 2: variante C pierde claramente (2.1% CTR vs. 3.8% en A\u002FB), posterior Bayesiano dice \"C pierde con 97% de probabilidad\". Pausas C, redistribuyes presupuesto a A\u002FB. Día 5: A gana con 91% de probabilidad, pausas B, todo a A. Decisión en 7 días; frequentismo esperaría 14.",{"type":32,"tag":40,"props":442,"children":444},{"id":443},"expected-loss-y-gestión-de-riesgo",[445],{"type":37,"value":446},"Expected Loss y Gestión de Riesgo",{"type":32,"tag":33,"props":448,"children":449},{},[450,452,458,460,466],{"type":37,"value":451},"Probabilidad de ganancia no es todo. Variante B gana en 60% de casos pero si pierde, pierde -%8 CR; si gana, +%3 CR. Cambiar a B bajo estas condiciones es riesgoso. La métrica expected loss lo mide: la CR media de pérdida en escenarios donde A supera B. Fórmula: ",{"type":32,"tag":73,"props":453,"children":455},{"className":454},[],[456],{"type":37,"value":457},"E[max(0, A - B)]",{"type":37,"value":459},". En Python: ",{"type":32,"tag":73,"props":461,"children":463},{"className":462},[],[464],{"type":37,"value":465},"numpy.maximum(samples_a - samples_b, 0).mean()",{"type":37,"value":467},". Si expected loss \u003C%1 y probabilidad de ganancia >%70, cambio seguro.",{"type":32,"tag":33,"props":469,"children":470},{},[471],{"type":37,"value":472},"Tabla: matriz de decisión Bayesiana",{"type":32,"tag":474,"props":475,"children":476},"table",{},[477,501],{"type":32,"tag":478,"props":479,"children":480},"thead",{},[481],{"type":32,"tag":482,"props":483,"children":484},"tr",{},[485,491,496],{"type":32,"tag":486,"props":487,"children":488},"th",{},[489],{"type":37,"value":490},"Probabilidad de ganancia",{"type":32,"tag":486,"props":492,"children":493},{},[494],{"type":37,"value":495},"Expected loss (CR)",{"type":32,"tag":486,"props":497,"children":498},{},[499],{"type":37,"value":500},"Acción",{"type":32,"tag":502,"props":503,"children":504},"tbody",{},[505,524,542],{"type":32,"tag":482,"props":506,"children":507},{},[508,514,519],{"type":32,"tag":509,"props":510,"children":511},"td",{},[512],{"type":37,"value":513},"94%",{"type":32,"tag":509,"props":515,"children":516},{},[517],{"type":37,"value":518},"0.3%",{"type":32,"tag":509,"props":520,"children":521},{},[522],{"type":37,"value":523},"Cambiar ahora",{"type":32,"tag":482,"props":525,"children":526},{},[527,532,537],{"type":32,"tag":509,"props":528,"children":529},{},[530],{"type":37,"value":531},"78%",{"type":32,"tag":509,"props":533,"children":534},{},[535],{"type":37,"value":536},"1.2%",{"type":32,"tag":509,"props":538,"children":539},{},[540],{"type":37,"value":541},"Recolectar más datos",{"type":32,"tag":482,"props":543,"children":544},{},[545,550,555],{"type":32,"tag":509,"props":546,"children":547},{},[548],{"type":37,"value":549},"51%",{"type":32,"tag":509,"props":551,"children":552},{},[553],{"type":37,"value":554},"2.8%",{"type":32,"tag":509,"props":556,"children":557},{},[558],{"type":37,"value":559},"Detener, sin diferencia",{"type":32,"tag":33,"props":561,"children":562},{},[563],{"type":37,"value":564},"Esta tabla vive en el dashboard. No preguntas al PM \"¿cambiamos a B?\", dices \"B gana con 78% de probabilidad pero expected loss es 1.2%, necesitamos 200 sesiones más\". Decisión clara, riesgo medido, sin perder tiempo.",{"type":32,"tag":40,"props":566,"children":568},{"id":567},"selección-de-prior-y-análisis-de-sensibilidad",[569],{"type":37,"value":570},"Selección de Prior y Análisis de Sensibilidad",{"type":32,"tag":33,"props":572,"children":573},{},[574],{"type":37,"value":575},"La matemática Bayesiana depende de la selección del prior. Prior uniforme (Beta(1,1)) es neutral; los datos dominan. Pero con conocimiento del dominio, prior informativo es mejor: tests anteriores muestran CR entre 2-3%, entonces Beta(20, 980) (media 2%). Este prior estabiliza el posterior en primeras 100 sesiones, reduce fluctuación aleatoria.",{"type":32,"tag":33,"props":577,"children":578},{},[579],{"type":37,"value":580},"Testea sensibilidad del prior: corre posterior con 3 priors distintos (uniforme, débilmente informativo, altamente informativo); si probabilidad de ganancia varía >5%, los datos son insuficientes. Ejemplo: prior uniforme da 82%, informativo 77%, diferencia \u003C%5%, adelante. Diferencia >%10% → recolecta más datos o recalibra prior (con datos históricos).",{"type":32,"tag":33,"props":582,"children":583},{},[584],{"type":37,"value":585},"Código: sensibilidad de prior",{"type":32,"tag":82,"props":587,"children":589},{"className":84,"code":588,"language":86,"meta":16,"style":16},"priors = [\n    (1, 1),           # uniforme\n    (10, 490),        # débilmente informativo, media=2%\n    (30, 1470)        # altamente informativo, media=2%\n]\n\nfor alpha, beta_prior in priors:\n    a_posterior = beta.rvs(alpha + 50, beta_prior + 1950, size=10000)\n    b_posterior = beta.rvs(alpha + 58, beta_prior + 1942, size=10000)\n    prob = (b_posterior > a_posterior).mean()\n    print(f\"Prior Beta({alpha},{beta_prior}): P(B>A)={prob:.2f}\")\n",[590],{"type":32,"tag":73,"props":591,"children":592},{"__ignoreMap":16},[593,610,640,671,702,710,717,740,800,857,883],{"type":32,"tag":92,"props":594,"children":595},{"class":94,"line":95},[596,601,605],{"type":32,"tag":92,"props":597,"children":598},{"style":105},[599],{"type":37,"value":600},"priors ",{"type":32,"tag":92,"props":602,"children":603},{"style":99},[604],{"type":37,"value":152},{"type":32,"tag":92,"props":606,"children":607},{"style":105},[608],{"type":37,"value":609}," [\n",{"type":32,"tag":92,"props":611,"children":612},{"class":94,"line":121},[613,618,622,626,630,635],{"type":32,"tag":92,"props":614,"children":615},{"style":105},[616],{"type":37,"value":617},"    (",{"type":32,"tag":92,"props":619,"children":620},{"style":155},[621],{"type":37,"value":178},{"type":32,"tag":92,"props":623,"children":624},{"style":105},[625],{"type":37,"value":173},{"type":32,"tag":92,"props":627,"children":628},{"style":155},[629],{"type":37,"value":178},{"type":32,"tag":92,"props":631,"children":632},{"style":105},[633],{"type":37,"value":634},"),           ",{"type":32,"tag":92,"props":636,"children":637},{"style":135},[638],{"type":37,"value":639},"# uniforme\n",{"type":32,"tag":92,"props":641,"children":642},{"class":94,"line":131},[643,647,652,656,661,666],{"type":32,"tag":92,"props":644,"children":645},{"style":105},[646],{"type":37,"value":617},{"type":32,"tag":92,"props":648,"children":649},{"style":155},[650],{"type":37,"value":651},"10",{"type":32,"tag":92,"props":653,"children":654},{"style":105},[655],{"type":37,"value":173},{"type":32,"tag":92,"props":657,"children":658},{"style":155},[659],{"type":37,"value":660},"490",{"type":32,"tag":92,"props":662,"children":663},{"style":105},[664],{"type":37,"value":665},"),        ",{"type":32,"tag":92,"props":667,"children":668},{"style":135},[669],{"type":37,"value":670},"# débilmente informativo, media=2%\n",{"type":32,"tag":92,"props":672,"children":673},{"class":94,"line":141},[674,678,683,687,692,697],{"type":32,"tag":92,"props":675,"children":676},{"style":105},[677],{"type":37,"value":617},{"type":32,"tag":92,"props":679,"children":680},{"style":155},[681],{"type":37,"value":682},"30",{"type":32,"tag":92,"props":684,"children":685},{"style":105},[686],{"type":37,"value":173},{"type":32,"tag":92,"props":688,"children":689},{"style":155},[690],{"type":37,"value":691},"1470",{"type":32,"tag":92,"props":693,"children":694},{"style":105},[695],{"type":37,"value":696},")        ",{"type":32,"tag":92,"props":698,"children":699},{"style":135},[700],{"type":37,"value":701},"# altamente informativo, media=2%\n",{"type":32,"tag":92,"props":703,"children":704},{"class":94,"line":209},[705],{"type":32,"tag":92,"props":706,"children":707},{"style":105},[708],{"type":37,"value":709},"]\n",{"type":32,"tag":92,"props":711,"children":712},{"class":94,"line":218},[713],{"type":32,"tag":92,"props":714,"children":715},{"emptyLinePlaceholder":125},[716],{"type":37,"value":128},{"type":32,"tag":92,"props":718,"children":719},{"class":94,"line":276},[720,725,730,735],{"type":32,"tag":92,"props":721,"children":722},{"style":99},[723],{"type":37,"value":724},"for",{"type":32,"tag":92,"props":726,"children":727},{"style":105},[728],{"type":37,"value":729}," alpha, beta_prior ",{"type":32,"tag":92,"props":731,"children":732},{"style":99},[733],{"type":37,"value":734},"in",{"type":32,"tag":92,"props":736,"children":737},{"style":105},[738],{"type":37,"value":739}," priors:\n",{"type":32,"tag":92,"props":741,"children":742},{"class":94,"line":26},[743,748,752,757,762,766,771,775,780,784,788,792,796],{"type":32,"tag":92,"props":744,"children":745},{"style":105},[746],{"type":37,"value":747},"    a_posterior ",{"type":32,"tag":92,"props":749,"children":750},{"style":99},[751],{"type":37,"value":152},{"type":32,"tag":92,"props":753,"children":754},{"style":105},[755],{"type":37,"value":756}," beta.rvs(alpha ",{"type":32,"tag":92,"props":758,"children":759},{"style":99},[760],{"type":37,"value":761},"+",{"type":32,"tag":92,"props":763,"children":764},{"style":155},[765],{"type":37,"value":168},{"type":32,"tag":92,"props":767,"children":768},{"style":105},[769],{"type":37,"value":770},", beta_prior ",{"type":32,"tag":92,"props":772,"children":773},{"style":99},[774],{"type":37,"value":761},{"type":32,"tag":92,"props":776,"children":777},{"style":155},[778],{"type":37,"value":779}," 1950",{"type":32,"tag":92,"props":781,"children":782},{"style":105},[783],{"type":37,"value":173},{"type":32,"tag":92,"props":785,"children":786},{"style":301},[787],{"type":37,"value":304},{"type":32,"tag":92,"props":789,"children":790},{"style":99},[791],{"type":37,"value":152},{"type":32,"tag":92,"props":793,"children":794},{"style":155},[795],{"type":37,"value":313},{"type":32,"tag":92,"props":797,"children":798},{"style":105},[799],{"type":37,"value":206},{"type":32,"tag":92,"props":801,"children":802},{"class":94,"line":320},[803,808,812,816,820,824,828,832,837,841,845,849,853],{"type":32,"tag":92,"props":804,"children":805},{"style":105},[806],{"type":37,"value":807},"    b_posterior ",{"type":32,"tag":92,"props":809,"children":810},{"style":99},[811],{"type":37,"value":152},{"type":32,"tag":92,"props":813,"children":814},{"style":105},[815],{"type":37,"value":756},{"type":32,"tag":92,"props":817,"children":818},{"style":99},[819],{"type":37,"value":761},{"type":32,"tag":92,"props":821,"children":822},{"style":155},[823],{"type":37,"value":241},{"type":32,"tag":92,"props":825,"children":826},{"style":105},[827],{"type":37,"value":770},{"type":32,"tag":92,"props":829,"children":830},{"style":99},[831],{"type":37,"value":761},{"type":32,"tag":92,"props":833,"children":834},{"style":155},[835],{"type":37,"value":836}," 1942",{"type":32,"tag":92,"props":838,"children":839},{"style":105},[840],{"type":37,"value":173},{"type":32,"tag":92,"props":842,"children":843},{"style":301},[844],{"type":37,"value":304},{"type":32,"tag":92,"props":846,"children":847},{"style":99},[848],{"type":37,"value":152},{"type":32,"tag":92,"props":850,"children":851},{"style":155},[852],{"type":37,"value":313},{"type":32,"tag":92,"props":854,"children":855},{"style":105},[856],{"type":37,"value":206},{"type":32,"tag":92,"props":858,"children":859},{"class":94,"line":354},[860,865,869,874,878],{"type":32,"tag":92,"props":861,"children":862},{"style":105},[863],{"type":37,"value":864},"    prob ",{"type":32,"tag":92,"props":866,"children":867},{"style":99},[868],{"type":37,"value":152},{"type":32,"tag":92,"props":870,"children":871},{"style":105},[872],{"type":37,"value":873}," (b_posterior ",{"type":32,"tag":92,"props":875,"children":876},{"style":99},[877],{"type":37,"value":382},{"type":32,"tag":92,"props":879,"children":880},{"style":105},[881],{"type":37,"value":882}," a_posterior).mean()\n",{"type":32,"tag":92,"props":884,"children":885},{"class":94,"line":362},[886,891,896,901,907,912,917,922,927,931,936,940,945,949,954,959,963,968],{"type":32,"tag":92,"props":887,"children":888},{"style":155},[889],{"type":37,"value":890},"    print",{"type":32,"tag":92,"props":892,"children":893},{"style":105},[894],{"type":37,"value":895},"(",{"type":32,"tag":92,"props":897,"children":898},{"style":99},[899],{"type":37,"value":900},"f",{"type":32,"tag":92,"props":902,"children":904},{"style":903},"--shiki-default:#9ECBFF",[905],{"type":37,"value":906},"\"Prior Beta(",{"type":32,"tag":92,"props":908,"children":909},{"style":155},[910],{"type":37,"value":911},"{",{"type":32,"tag":92,"props":913,"children":914},{"style":105},[915],{"type":37,"value":916},"alpha",{"type":32,"tag":92,"props":918,"children":919},{"style":155},[920],{"type":37,"value":921},"}",{"type":32,"tag":92,"props":923,"children":924},{"style":903},[925],{"type":37,"value":926},",",{"type":32,"tag":92,"props":928,"children":929},{"style":155},[930],{"type":37,"value":911},{"type":32,"tag":92,"props":932,"children":933},{"style":105},[934],{"type":37,"value":935},"beta_prior",{"type":32,"tag":92,"props":937,"children":938},{"style":155},[939],{"type":37,"value":921},{"type":32,"tag":92,"props":941,"children":942},{"style":903},[943],{"type":37,"value":944},"): P(B>A)=",{"type":32,"tag":92,"props":946,"children":947},{"style":155},[948],{"type":37,"value":911},{"type":32,"tag":92,"props":950,"children":951},{"style":105},[952],{"type":37,"value":953},"prob",{"type":32,"tag":92,"props":955,"children":956},{"style":99},[957],{"type":37,"value":958},":.2f",{"type":32,"tag":92,"props":960,"children":961},{"style":155},[962],{"type":37,"value":921},{"type":32,"tag":92,"props":964,"children":965},{"style":903},[966],{"type":37,"value":967},"\"",{"type":32,"tag":92,"props":969,"children":970},{"style":105},[971],{"type":37,"value":206},{"type":32,"tag":33,"props":973,"children":974},{},[975],{"type":37,"value":976},"Resultados consistentes (±3%) = prior robusto.",{"type":32,"tag":40,"props":978,"children":980},{"id":979},"cierre-ganancia-de-velocidad-y-adaptación-organizacional",[981],{"type":37,"value":982},"Cierre: Ganancia de Velocidad y Adaptación Organizacional",{"type":32,"tag":33,"props":984,"children":985},{},[986],{"type":37,"value":987},"El test Bayesiano A\u002FB no basta solo; también necesitas cambiar el proceso de decisión organizacional. Pasar de \"espera a que sea significativo\" a \"avanza con riesgo medido\" requiere cambio cultural. Al CMO entregas probabilidad 90%, no certeza 100%; ese cambio psicológico cuesta. Pero la ganancia es clara: tiempo de test de 14 a 7 días en promedio, costo de variantes perdedoras -50%, velocidad de iteración creativa x2. En Meta Ads, esa velocidad se traduce directo en ROAS — más tests, creativos ganadores mejores, CPA menor. Integra matemática Bayesiana en tu dataflow (BigQuery + dbt + Looker) y eliminas cálculos manuales: actualización automática de posterior cada noche, métricas de decisión frescos cada mañana.",{"type":32,"tag":989,"props":990,"children":991},"style",{},[992],{"type":37,"value":993},"html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}",{"title":16,"searchDepth":131,"depth":131,"links":995},[996,997,998,999,1000,1001],{"id":42,"depth":121,"text":45},{"id":58,"depth":121,"text":61},{"id":411,"depth":121,"text":414},{"id":443,"depth":121,"text":446},{"id":567,"depth":121,"text":570},{"id":979,"depth":121,"text":982},"markdown","content:es:marketing:prueba-bayesiana-ab-test-toma-de-decisiones-rapida.md","content","es\u002Fmarketing\u002Fprueba-bayesiana-ab-test-toma-de-decisiones-rapida.md","es\u002Fmarketing\u002Fprueba-bayesiana-ab-test-toma-de-decisiones-rapida","md",1786860285980]