[{"data":1,"prerenderedAt":833},["ShallowReactive",2],{"article-alternates":3,"article-\u002Fes\u002Fmarketing\u002Fmmm-incrementality-setup-2026":13},{"i18nKey":4,"paths":5},"marketing-004-2026-06",{"de":6,"en":7,"es":8,"fr":9,"it":10,"ru":11,"tr":12},"\u002Fde\u002Fmarketing\u002Fmmm-incrementalitaet-2026-attribution-setup","\u002Fen\u002Fmarketing\u002Fmmm-incrementality-attribution-setup-2026","\u002Fes\u002Fmarketing\u002Fmmm-incrementality-setup-2026","\u002Ffr\u002Fmarketing\u002Fmmm-incrementalite-2026-attribution-setup","\u002Fit\u002Fmarketing\u002Fmmm-incrementality-2026-attribution-setup","\u002Fru\u002Fmarketing\u002Fmmm-incrementalnost-2026-attribution-setup","\u002Ftr\u002Fmarketing\u002Fmmm-incrementality-2026nin-attribution-setupi",{"_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":827,"_id":828,"_source":829,"_file":830,"_stem":831,"_extension":832},"marketing",false,"","MMM + Incrementality: El setup de atribución de 2026","Robyn, Meta Lift, geo experiments — ¿cuándo usar cada uno? Árbol de decisión práctico para atribución post-cookie.","2026-06-04",[21,22,23,24,25],"mmm","incrementality","atribución","robyn","geo-test",8,"Roibase",{"type":29,"children":30,"toc":813},"root",[31,39,46,51,56,61,67,72,81,126,134,153,161,179,195,201,206,213,225,233,256,266,271,279,297,305,323,329,334,342,365,373,391,396,404,422,429,447,453,458,611,619,662,667,673,683,693,703,713,723,729,734,797,802,808],{"type":32,"tag":33,"props":34,"children":35},"element","p",{},[36],{"type":37,"value":38},"text","El 80% del tracking de cookies desapareció, Multi-Touch Attribution (MTA) ya no es fiable, los dashboards de plataformas se contradicen entre sí. En 2026, los marketers miden la \"contribución\" combinando dos métodos: Marketing Mix Modeling (MMM) y tests de incrementalidad. El problema: pocos saben cuándo usar cada uno. Este artículo muestra dónde encajan Robyn (la librería MMM open source de Meta), la Meta Lift API y los tests basados en geo dentro del mismo setup.",{"type":32,"tag":40,"props":41,"children":43},"h2",{"id":42},"last-touch-attribution-está-muerto-pero-qué-lo-reemplaza",[44],{"type":37,"value":45},"Last-touch attribution está muerto — pero ¿qué lo reemplaza?",{"type":32,"tag":33,"props":47,"children":48},{},[49],{"type":37,"value":50},"Google Analytics 4 dice \"data-driven attribution\", Meta dice \"modeled conversions\", TikTok da su propio número. Los tres reportan cifras diferentes. En 2025, un ecommerce que gasta 100 dólares ve 8 conversiones en GA4, 12 en Meta y 6 en TikTok. ¿Cuál canal realmente funciona? Last-touch no puede responder porque un usuario pasa por múltiples touchpoints y cada plataforma se da crédito a sí misma.",{"type":32,"tag":33,"props":52,"children":53},{},[54],{"type":37,"value":55},"Marketing Mix Modeling resuelve esto de otro ángulo: toma los canales como variables independientes, la venta o revenue como variable dependiente, calcula la contribución marginal de cada canal mediante regresión. Los tests de incrementalidad van más al grano: expones un grupo a un canal, otro grupo no, y mides la diferencia. Ambos rompen la ilusión del último clic, pero sus escenarios de uso no se solapan.",{"type":32,"tag":33,"props":57,"children":58},{},[59],{"type":37,"value":60},"La diferencia radica aquí: MMM es macro (largo plazo, todos los canales), incrementalidad es micro (corto plazo, canal o campaña específica). Un setup que combine ambos es estándar en 2026.",{"type":32,"tag":40,"props":62,"children":64},{"id":63},"mmm-setup-de-regresión-semanal-con-robyn",[65],{"type":37,"value":66},"MMM: setup de regresión semanal con Robyn",{"type":32,"tag":33,"props":68,"children":69},{},[70],{"type":37,"value":71},"Robyn es el framework MMM open source del equipo Facebook Marketing Science de Meta. Usa R, regresión Bayesian ridge, y ajusta automáticamente curvas de adstock (efecto retardado) y saturation (rendimientos decrecientes). Con granularidad semanal, devuelve el porcentaje de contribución de cada canal (TV, display, paid social, SEO, email) a las ventas.",{"type":32,"tag":33,"props":73,"children":74},{},[75],{"type":32,"tag":76,"props":77,"children":78},"strong",{},[79],{"type":37,"value":80},"Las 4 capas del setup Robyn:",{"type":32,"tag":82,"props":83,"children":84},"ol",{},[85,96,106,116],{"type":32,"tag":86,"props":87,"children":88},"li",{},[89,94],{"type":32,"tag":76,"props":90,"children":91},{},[92],{"type":37,"value":93},"Recopilación de datos:",{"type":37,"value":95}," Mínimo 1,5 años de datos semanales. Cada fila = una semana. Columnas: gasto por canal, impresiones o clicks; variables independientes (precio, stock, estacionalidad); variable dependiente (revenue, órdenes, conversiones). Sin datos, el modelo no funciona.",{"type":32,"tag":86,"props":97,"children":98},{},[99,104],{"type":32,"tag":76,"props":100,"children":101},{},[102],{"type":37,"value":103},"Tuning de hiperparámetros:",{"type":37,"value":105}," Robyn busca para cada canal los parámetros de adstock decay (α) y saturation shape (γ). Ejecuta 2000+ combinaciones de modelos y sugiere los 5-10 mejores de la frontera de Pareto. Esta fase tarda 10-30 minutos (en 64 cores).",{"type":32,"tag":86,"props":107,"children":108},{},[109,114],{"type":32,"tag":76,"props":110,"children":111},{},[112],{"type":37,"value":113},"Selección del modelo:",{"type":37,"value":115}," Tomas el modelo con el NRMSE más bajo (Normalized Root Mean Squared Error) + la decomp.rssd más alta (estabilidad de descomposición). El output: porcentaje de contribución de cada canal a las ventas totales, estimación de ROI, asignación óptima de presupuesto.",{"type":32,"tag":86,"props":117,"children":118},{},[119,124],{"type":32,"tag":76,"props":120,"children":121},{},[122],{"type":37,"value":123},"Asignación de presupuesto:",{"type":37,"value":125}," La función \"budget allocator\" de Robyn redistribuye el presupuesto total — igualando el ROI marginal de cada canal. Este output te sirve para planificar el próximo trimestre.",{"type":32,"tag":33,"props":127,"children":128},{},[129],{"type":32,"tag":76,"props":130,"children":131},{},[132],{"type":37,"value":133},"Cuándo usar Robyn:",{"type":32,"tag":135,"props":136,"children":137},"ul",{},[138,143,148],{"type":32,"tag":86,"props":139,"children":140},{},[141],{"type":37,"value":142},"Decisiones de asignación presupuestaria intercanal (por ej., plan Q3)",{"type":32,"tag":86,"props":144,"children":145},{},[146],{"type":37,"value":147},"Simulación de agregar\u002Feliminar un canal nuevo",{"type":32,"tag":86,"props":149,"children":150},{},[151],{"type":37,"value":152},"Análisis de tendencias a largo plazo (6+ meses)",{"type":32,"tag":33,"props":154,"children":155},{},[156],{"type":32,"tag":76,"props":157,"children":158},{},[159],{"type":37,"value":160},"Cuándo NO usar Robyn:",{"type":32,"tag":135,"props":162,"children":163},{},[164,169,174],{"type":32,"tag":86,"props":165,"children":166},{},[167],{"type":37,"value":168},"Optimizar dentro de una campaña (periodos \u003C 2 semanas)",{"type":32,"tag":86,"props":170,"children":171},{},[172],{"type":37,"value":173},"Decisiones de creative test específico de plataforma (MMM no ve diferencias de creative)",{"type":32,"tag":86,"props":175,"children":176},{},[177],{"type":37,"value":178},"Feedback en tiempo real (hay 1 semana de lag)",{"type":32,"tag":33,"props":180,"children":181},{},[182,184,193],{"type":37,"value":183},"Roibase integra Robyn en nuestro servicio de ",{"type":32,"tag":185,"props":186,"children":190},"a",{"href":187,"rel":188},"https:\u002F\u002Fwww.roibase.com.tr\u002Fes\u002Fdijitalpazarlama",[189],"nofollow",[191],{"type":37,"value":192},"Dijital Pazarlama",{"type":37,"value":194},": conectamos GA4, GTM server-side, Meta CAPI y BigQuery, construimos un pipeline ETL semanal, y visualizamos el output en Data Studio.",{"type":32,"tag":40,"props":196,"children":198},{"id":197},"tests-de-incrementalidad-meta-lift-y-geo-based-holdout",[199],{"type":37,"value":200},"Tests de incrementalidad: Meta Lift y geo-based holdout",{"type":32,"tag":33,"props":202,"children":203},{},[204],{"type":37,"value":205},"MMM responde \"cuánto\", los tests de incrementalidad responden \"¿realmente funciona?\" Dos preguntas distintas. Si gastas 100k TL en Meta y obtienes 120 conversiones, ¿está \"bien\"? MMM te dice \"Meta toma el 15% de tu presupuesto, genera el 12% de las ventas totales\". ¿Pero cuántas conversiones habrían ocurrido de todas formas (organic)? Para eso necesitas un test de incrementalidad.",{"type":32,"tag":207,"props":208,"children":210},"h3",{"id":209},"meta-conversion-lift",[211],{"type":37,"value":212},"Meta Conversion Lift",{"type":32,"tag":33,"props":214,"children":215},{},[216,218,223],{"type":37,"value":217},"Meta Lift API mide el ",{"type":32,"tag":76,"props":219,"children":220},{},[221],{"type":37,"value":222},"impacto real",{"type":37,"value":224}," de tus anuncios en Facebook e Instagram. ¿Cómo? Muestra la campaña a un grupo holdout pequeño, a otro no, y mide la diferencia 7-14 días después. La diferencia = conversiones incrementales.",{"type":32,"tag":33,"props":226,"children":227},{},[228],{"type":32,"tag":76,"props":229,"children":230},{},[231],{"type":37,"value":232},"Setup:",{"type":32,"tag":135,"props":234,"children":235},{},[236,241,246,251],{"type":32,"tag":86,"props":237,"children":238},{},[239],{"type":37,"value":240},"Antes de lanzar, abres un Lift study (Ads Manager > Measure & Report > Conversion Lift)",{"type":32,"tag":86,"props":242,"children":243},{},[244],{"type":37,"value":245},"El ratio de holdout es 5-10% (muy pequeño = ruido, muy grande = pérdida de impresiones)",{"type":32,"tag":86,"props":247,"children":248},{},[249],{"type":37,"value":250},"Duración mínima: 7 días (más corto = poder estadístico bajo)",{"type":32,"tag":86,"props":252,"children":253},{},[254],{"type":37,"value":255},"Resultado: conversiones incrementales, CPA incremental, intervalo de confianza",{"type":32,"tag":33,"props":257,"children":258},{},[259,264],{"type":32,"tag":76,"props":260,"children":261},{},[262],{"type":37,"value":263},"Ejemplo de interpretación:",{"type":37,"value":265},"\nControl group: 1000 personas, 40 conversiones\nTest group: 9000 personas, 450 conversiones\nConversión incremental = (450\u002F9000 - 40\u002F1000) × 9000 = 90 conversiones\nLift = 90 \u002F (450 - 90) = 25%",{"type":32,"tag":33,"props":267,"children":268},{},[269],{"type":37,"value":270},"Entonces, de las 450 conversiones vistas por la campaña, solo 90 vinieron realmente del anuncio. El resto habría ocurrido de todos modos. CPA incremental = (gasto) \u002F 90. Este número es 30-60% más alto que MTA — porque es real.",{"type":32,"tag":33,"props":272,"children":273},{},[274],{"type":32,"tag":76,"props":275,"children":276},{},[277],{"type":37,"value":278},"Cuándo usar Meta Lift:",{"type":32,"tag":135,"props":280,"children":281},{},[282,287,292],{"type":32,"tag":86,"props":283,"children":284},{},[285],{"type":37,"value":286},"A\u002FB test de nuevas campañas o creatividades",{"type":32,"tag":86,"props":288,"children":289},{},[290],{"type":37,"value":291},"Decisión de plataforma (¿Meta vs. Google vs. TikTok cuál es más incremental?)",{"type":32,"tag":86,"props":293,"children":294},{},[295],{"type":37,"value":296},"Medir el impacto real del retargeting (problema común: retargeting siempre aparece con CPA bajo, pero 80% habría comprado igual)",{"type":32,"tag":33,"props":298,"children":299},{},[300],{"type":32,"tag":76,"props":301,"children":302},{},[303],{"type":37,"value":304},"Desventaja:",{"type":32,"tag":135,"props":306,"children":307},{},[308,313,318],{"type":32,"tag":86,"props":309,"children":310},{},[311],{"type":37,"value":312},"Solo funciona en Meta (Google Display & Video 360 tiene equivalente limitado)",{"type":32,"tag":86,"props":314,"children":315},{},[316],{"type":37,"value":317},"Crear un grupo holdout cuesta impresiones (revenue cae a corto plazo)",{"type":32,"tag":86,"props":319,"children":320},{},[321],{"type":37,"value":322},"Test mínimo 1 semana — no sirve para decisiones diarias",{"type":32,"tag":207,"props":324,"children":326},{"id":325},"geo-based-experiments-holdout-geográfico",[327],{"type":37,"value":328},"Geo-based experiments (holdout geográfico)",{"type":32,"tag":33,"props":330,"children":331},{},[332],{"type":37,"value":333},"Para canales fuera de Meta (Google, TikTok, TV), corres tests basados en geografía: abres campaña en algunas ciudades, la cierras en otras, observas la diferencia en ventas. Es el método más limpio académicamente porque no hay manipulación a nivel de usuario.",{"type":32,"tag":33,"props":335,"children":336},{},[337],{"type":32,"tag":76,"props":338,"children":339},{},[340],{"type":37,"value":341},"Ejemplo de setup:",{"type":32,"tag":135,"props":343,"children":344},{},[345,350,355,360],{"type":32,"tag":86,"props":346,"children":347},{},[348],{"type":37,"value":349},"Selecciona 30 ciudades (población y nivel económico similares)",{"type":32,"tag":86,"props":351,"children":352},{},[353],{"type":37,"value":354},"Abre campaña de Google Ads en 15, mantenla cerrada en 15 (aleatorizadas)",{"type":32,"tag":86,"props":356,"children":357},{},[358],{"type":37,"value":359},"Espera 4 semanas",{"type":32,"tag":86,"props":361,"children":362},{},[363],{"type":37,"value":364},"Compara conversiones por ciudad en Google Analytics 4",{"type":32,"tag":33,"props":366,"children":367},{},[368],{"type":32,"tag":76,"props":369,"children":370},{},[371],{"type":37,"value":372},"Análisis:",{"type":32,"tag":135,"props":374,"children":375},{},[376,381,386],{"type":32,"tag":86,"props":377,"children":378},{},[379],{"type":37,"value":380},"Ciudades tratadas: promedio 120 conversiones\u002Fciudad",{"type":32,"tag":86,"props":382,"children":383},{},[384],{"type":37,"value":385},"Ciudades control: promedio 95 conversiones\u002Fciudad",{"type":32,"tag":86,"props":387,"children":388},{},[389],{"type":37,"value":390},"Lift incremental: (120 - 95) \u002F 95 = 26.3%",{"type":32,"tag":33,"props":392,"children":393},{},[394],{"type":37,"value":395},"Extrapolas este 26.3% de lift a todo el país. Con presupuesto de Google Ads de 200k TL, calculas revenue incremental e iROAS incremental.",{"type":32,"tag":33,"props":397,"children":398},{},[399],{"type":32,"tag":76,"props":400,"children":401},{},[402],{"type":37,"value":403},"Cuándo usar geo test:",{"type":32,"tag":135,"props":405,"children":406},{},[407,412,417],{"type":32,"tag":86,"props":408,"children":409},{},[410],{"type":37,"value":411},"Medir la contribución neta de cada canal en setup multicanal",{"type":32,"tag":86,"props":413,"children":414},{},[415],{"type":37,"value":416},"Ver impacto de canales no digitales (TV, OOH, podcast)",{"type":32,"tag":86,"props":418,"children":419},{},[420],{"type":37,"value":421},"Cuando no confías en los dashboards de plataformas",{"type":32,"tag":33,"props":423,"children":424},{},[425],{"type":32,"tag":76,"props":426,"children":427},{},[428],{"type":37,"value":304},{"type":32,"tag":135,"props":430,"children":431},{},[432,437,442],{"type":32,"tag":86,"props":433,"children":434},{},[435],{"type":37,"value":436},"Pocas ciudades = poder estadístico bajo (mínimo 20 ciudades)",{"type":32,"tag":86,"props":438,"children":439},{},[440],{"type":37,"value":441},"Heterogeneidad geográfica sesgada (İstanbul ≠ Şanlıurfa, no se puede meter en el mismo bucket)",{"type":32,"tag":86,"props":443,"children":444},{},[445],{"type":37,"value":446},"Largo: 4-8 semanas",{"type":32,"tag":40,"props":448,"children":450},{"id":449},"árbol-de-decisión-cuándo-usar-cada-método",[451],{"type":37,"value":452},"Árbol de decisión: cuándo usar cada método",{"type":32,"tag":33,"props":454,"children":455},{},[456],{"type":37,"value":457},"En el mismo setup, organizamos los tres métodos así:",{"type":32,"tag":459,"props":460,"children":461},"table",{},[462,491],{"type":32,"tag":463,"props":464,"children":465},"thead",{},[466],{"type":32,"tag":467,"props":468,"children":469},"tr",{},[470,476,481,486],{"type":32,"tag":471,"props":472,"children":473},"th",{},[474],{"type":37,"value":475},"Escenario",{"type":32,"tag":471,"props":477,"children":478},{},[479],{"type":37,"value":480},"Método",{"type":32,"tag":471,"props":482,"children":483},{},[484],{"type":37,"value":485},"Frecuencia",{"type":32,"tag":471,"props":487,"children":488},{},[489],{"type":37,"value":490},"Output",{"type":32,"tag":492,"props":493,"children":494},"tbody",{},[495,519,542,565,588],{"type":32,"tag":467,"props":496,"children":497},{},[498,504,509,514],{"type":32,"tag":499,"props":500,"children":501},"td",{},[502],{"type":37,"value":503},"Asignación presupuestaria quarterly",{"type":32,"tag":499,"props":505,"children":506},{},[507],{"type":37,"value":508},"Robyn MMM",{"type":32,"tag":499,"props":510,"children":511},{},[512],{"type":37,"value":513},"Cada 3 meses",{"type":32,"tag":499,"props":515,"children":516},{},[517],{"type":37,"value":518},"ROI por canal, asignación óptima",{"type":32,"tag":467,"props":520,"children":521},{},[522,527,532,537],{"type":32,"tag":499,"props":523,"children":524},{},[525],{"type":37,"value":526},"Test de nueva campaña (Meta\u002FInstagram)",{"type":32,"tag":499,"props":528,"children":529},{},[530],{"type":37,"value":531},"Meta Lift",{"type":32,"tag":499,"props":533,"children":534},{},[535],{"type":37,"value":536},"Cada campaña grande",{"type":32,"tag":499,"props":538,"children":539},{},[540],{"type":37,"value":541},"CPA incremental",{"type":32,"tag":467,"props":543,"children":544},{},[545,550,555,560],{"type":32,"tag":499,"props":546,"children":547},{},[548],{"type":37,"value":549},"Incrementalidad cross-channel",{"type":32,"tag":499,"props":551,"children":552},{},[553],{"type":37,"value":554},"Geo-based holdout",{"type":32,"tag":499,"props":556,"children":557},{},[558],{"type":37,"value":559},"2 veces\u002Faño",{"type":32,"tag":499,"props":561,"children":562},{},[563],{"type":37,"value":564},"Lift real por canal",{"type":32,"tag":467,"props":566,"children":567},{},[568,573,578,583],{"type":32,"tag":499,"props":569,"children":570},{},[571],{"type":37,"value":572},"Decisión de refresh creativo",{"type":32,"tag":499,"props":574,"children":575},{},[576],{"type":37,"value":577},"Meta Lift + CRO",{"type":32,"tag":499,"props":579,"children":580},{},[581],{"type":37,"value":582},"1 vez\u002Fmes",{"type":32,"tag":499,"props":584,"children":585},{},[586],{"type":37,"value":587},"Qué creativo es incremental",{"type":32,"tag":467,"props":589,"children":590},{},[591,596,601,606],{"type":32,"tag":499,"props":592,"children":593},{},[594],{"type":37,"value":595},"Ajuste en tiempo real",{"type":32,"tag":499,"props":597,"children":598},{},[599],{"type":37,"value":600},"API de plataforma (ROAS feedback)",{"type":32,"tag":499,"props":602,"children":603},{},[604],{"type":37,"value":605},"Diario",{"type":32,"tag":499,"props":607,"children":608},{},[609],{"type":37,"value":610},"Optimización táctica",{"type":32,"tag":33,"props":612,"children":613},{},[614],{"type":32,"tag":76,"props":615,"children":616},{},[617],{"type":37,"value":618},"Flujo práctico:",{"type":32,"tag":82,"props":620,"children":621},{},[622,632,642,652],{"type":32,"tag":86,"props":623,"children":624},{},[625,630],{"type":32,"tag":76,"props":626,"children":627},{},[628],{"type":37,"value":629},"Semanal:",{"type":37,"value":631}," Monitorea dashboards de plataforma (parecido a MTA, pero sin confiar ciegamente)",{"type":32,"tag":86,"props":633,"children":634},{},[635,640],{"type":32,"tag":76,"props":636,"children":637},{},[638],{"type":37,"value":639},"Mensual:",{"type":37,"value":641}," Prueba campañas grandes con Meta Lift",{"type":32,"tag":86,"props":643,"children":644},{},[645,650],{"type":32,"tag":76,"props":646,"children":647},{},[648],{"type":37,"value":649},"Quarterly:",{"type":37,"value":651}," Ejecuta Robyn sobre todos los canales, realoca presupuesto basado en largo plazo",{"type":32,"tag":86,"props":653,"children":654},{},[655,660],{"type":32,"tag":76,"props":656,"children":657},{},[658],{"type":37,"value":659},"2 veces\u002Faño:",{"type":37,"value":661}," Valida el lift real de cada canal con geo test",{"type":32,"tag":33,"props":663,"children":664},{},[665],{"type":37,"value":666},"Este setup de 3 capas te permite tomar decisiones tácticas (qué creativo funciona) y estratégicas (cuánto presupuesto por canal) con datos.",{"type":32,"tag":40,"props":668,"children":670},{"id":669},"malentendidos-comunes-y-tradeoffs",[671],{"type":37,"value":672},"Malentendidos comunes y tradeoffs",{"type":32,"tag":33,"props":674,"children":675},{},[676,681],{"type":32,"tag":76,"props":677,"children":678},{},[679],{"type":37,"value":680},"Malentendido 1:",{"type":37,"value":682}," \"Si haces MMM no necesitas test de incrementalidad\"\nFalso. MMM muestra correlación, asume causalidad. El test de incrementalidad mide causalidad. Se complementan. Ejemplo: MMM dice \"Instagram contribuye 15%\", pero Lift test muestra que 40% de eso sería orgánico. La contribución real es 9%.",{"type":32,"tag":33,"props":684,"children":685},{},[686,691],{"type":32,"tag":76,"props":687,"children":688},{},[689],{"type":37,"value":690},"Malentendido 2:",{"type":37,"value":692}," \"Todo test de incrementalidad se hace en cada campaña\"\nFalso. Crear un holdout cuesta impresiones. Solo lo haces para decisiones grandes (nuevo canal, nueva dirección creativa, estrategia de retargeting). Las micro-optimizaciones usan A\u002FB test.",{"type":32,"tag":33,"props":694,"children":695},{},[696,701],{"type":32,"tag":76,"props":697,"children":698},{},[699],{"type":37,"value":700},"Malentendido 3:",{"type":37,"value":702}," \"Robyn se configura una vez y luego es automático\"\nFalso. El modelo se reentren cada trimestre. Si agregas canal, cambia precio, o varía estacionalidad, el modelo se actualiza. Robyn es mantenimiento continuo.",{"type":32,"tag":33,"props":704,"children":705},{},[706,711],{"type":32,"tag":76,"props":707,"children":708},{},[709],{"type":37,"value":710},"Tradeoff 1: Velocidad vs. precisión",{"type":37,"value":712},"\nMMM requiere 1,5 años de datos, resultado con 1 semana de lag. Geo test tarda 4-8 semanas. Si necesitas decisión rápida, confiarás en dashboards de plataforma pero aceptarás 30-50% de margen de error.",{"type":32,"tag":33,"props":714,"children":715},{},[716,721],{"type":32,"tag":76,"props":717,"children":718},{},[719],{"type":37,"value":720},"Tradeoff 2: Granularidad vs. tamaño de muestra",{"type":37,"value":722},"\nGeo test por ciudad = pequeño sample size, IC amplio. Por municipio = heterogeneidad aumenta. MMM semanal no responde preguntas diarias. Cada método tiene límite de resolución.",{"type":32,"tag":40,"props":724,"children":726},{"id":725},"cómo-se-construye-el-stack-de-atribución-en-2026",[727],{"type":37,"value":728},"Cómo se construye el stack de atribución en 2026",{"type":32,"tag":33,"props":730,"children":731},{},[732],{"type":37,"value":733},"El setup técnico consta de:",{"type":32,"tag":82,"props":735,"children":736},{},[737,747,757,767,777,787],{"type":32,"tag":86,"props":738,"children":739},{},[740,745],{"type":32,"tag":76,"props":741,"children":742},{},[743],{"type":37,"value":744},"GTM server-side + first-party cookie:",{"type":37,"value":746}," Envía señales limpias a GA4 y Meta CAPI (no ATT bypass, sino data enrichment basada en consentimiento)",{"type":32,"tag":86,"props":748,"children":749},{},[750,755],{"type":32,"tag":76,"props":751,"children":752},{},[753],{"type":37,"value":754},"Data warehouse BigQuery:",{"type":37,"value":756}," Centraliza todos los datos de plataformas (GA4, Meta Ads API, Google Ads API, TikTok Ads API, CRM)",{"type":32,"tag":86,"props":758,"children":759},{},[760,765],{"type":32,"tag":76,"props":761,"children":762},{},[763],{"type":37,"value":764},"dbt transformation:",{"type":37,"value":766}," Genera tablas semanales agregadas (cada fila = 1 semana, cada columna = gasto de 1 canal + 1 outcome)",{"type":32,"tag":86,"props":768,"children":769},{},[770,775],{"type":32,"tag":76,"props":771,"children":772},{},[773],{"type":37,"value":774},"Pipeline Robyn:",{"type":37,"value":776}," Script R en Cloud Run ejecutado semanalmente, output en BigQuery",{"type":32,"tag":86,"props":778,"children":779},{},[780,785],{"type":32,"tag":76,"props":781,"children":782},{},[783],{"type":37,"value":784},"Dashboard Looker Studio:",{"type":37,"value":786}," MMM output + MTA de plataformas + resultados de tests de incrementalidad lado a lado",{"type":32,"tag":86,"props":788,"children":789},{},[790,795],{"type":32,"tag":76,"props":791,"children":792},{},[793],{"type":37,"value":794},"Alertas Slack:",{"type":37,"value":796}," Si NRMSE de modelo sube >10%, aviso de anomalía en datos",{"type":32,"tag":33,"props":798,"children":799},{},[800],{"type":37,"value":801},"Armar este stack tarda 4-6 semanas. Después, mantenimiento semanal de 2-3 horas. ROI: asignación presupuestaria 15-25% más eficiente (Robyn reporta 18% mejora en su benchmark).",{"type":32,"tag":40,"props":803,"children":805},{"id":804},"qué-hacer-ahora",[806],{"type":37,"value":807},"Qué hacer ahora",{"type":32,"tag":33,"props":809,"children":810},{},[811],{"type":37,"value":812},"Si aún decides basado en last-touch attribution, no competirás en 2026. Primer paso: envía datos de plataformas a BigQuery, crea tabla semanal con 1,5 años de historial. Segundo paso: configura Robyn, entrena el primer modelo. Tercer paso: en la próxima campaña grande, abre Meta Lift study. Cuarto paso: en 6 meses, valida con geo test. Estos 4 pasos transforman tu attribution stack desde la ilusión MTA hacia fundamentación en incrementalidad real.",{"title":16,"searchDepth":814,"depth":814,"links":815},3,[816,818,819,823,824,825,826],{"id":42,"depth":817,"text":45},2,{"id":63,"depth":817,"text":66},{"id":197,"depth":817,"text":200,"children":820},[821,822],{"id":209,"depth":814,"text":212},{"id":325,"depth":814,"text":328},{"id":449,"depth":817,"text":452},{"id":669,"depth":817,"text":672},{"id":725,"depth":817,"text":728},{"id":804,"depth":817,"text":807},"markdown","content:es:marketing:mmm-incrementality-setup-2026.md","content","es\u002Fmarketing\u002Fmmm-incrementality-setup-2026.md","es\u002Fmarketing\u002Fmmm-incrementality-setup-2026","md",1782050761646]