Marketero · Entrevistas
How to Test Meta Ads and Scale with Confidence
31 ideas de marketing sacadas de «How to Test Meta Ads and Scale with Confidence», de Professor Charley T. Sobre todo de canales de adquisición, medición…
- 31 ideas despiezadas
- Fuente Professor Charley T
- Dura 59 min
- Publicada en 2025
Ver la entrevista entera en YouTube
De qué habla
Las 31 ideas de esta entrevista se reparten así entre las etapas del marketing. El reparto es lo que distingue una entrevista de otra: dice en qué se detuvo de verdad quien hablaba.
Las 31 ideas, con su cita y su minuto
Cada una está sacada de un punto concreto de la entrevista. El enlace del titular lleva a
la página de esa idea, donde se ve quién más la dice por separado; el del pie salta al
segundo exacto del vídeo para que la compruebes tú.
Canales de adquisiciónNo existe un anuncio 'perfecto' único; las campañas deben estructurarse como un portafolio donde distintas piezas desempeñan roles especializados (prospección en la parte alta, consideración intermedia, cierre y recompra en la baja), evaluando el éxito según la función de cada elemento.
And again, the misconception around creative testing is that you do it to find the winning ad. And my supposition here, backed by hundreds of millions of dollars of mistakes and well over a billion in revenue driven, is that there's no such thing as a best ad. You have an investment portfolio. Some ads are top of funnel. Some ads are mid. Some are bottom. Some are helping really get people they've already bought for me to buy again. Some are getting brand new people in. Everything has a role. So what's best when the definition of success for everything is different? Like a soccer or football…
Escúchalo en el minuto 33:53
Simplificación de la estructura de campañas para medir impacto real en ingresos
Canales de adquisiciónMinimizar las variables y partes móviles en la cuenta publicitaria permite evaluar directamente si aumentar el gasto en un anuncio mejora o empeora el rendimiento global del negocio y los ingresos netos, evitando el ruido de infinitas combinaciones algorítmicas.
The point is there are fewer grains of sand on the planet Earth than there are in possible ways for your 50 ads to be shown to any one individual on all of the platforms on any given day. So instead of trying to get lucky by seeing what works and what doesn't and then taking very out of context information, what you instead need is the fewest amount of moving parts so that you can simply say, when I spend more money on this ad, my overall campaign performance goes better or goes worse. And as a result of that campaign performance getting better or worse, what happens to my bank account?
Escúchalo en el minuto 20:27
Estabilización de cuenta y poda de anuncios por beneficio bruto por transacción
Medición y experimentosAntes de realizar nuevos tests creativos, se debe consolidar la cuenta reduciendo el exceso de anuncios para estabilizar el reparto del gasto y apagar de inmediato las piezas cuyo beneficio bruto por transacción (AOV menos CPA) esté por debajo del promedio.
money on all of our business objectives and optimize down to as few moving parts as possible So instead, like if we understand that if you have an account with dozens or even hundreds of ads running, it's completely impossible to understand the real impact of any one of your ads. Because also the thing is, if you stop running new ads, the distribution between all of your ads is going to start to stabilize. And to be fair, until you know what's going to happen, you have no ability to test with confidence because you don't know if the thing that you're doing is any better or worse understand…
Escúchalo en el minuto 36:43
Escalabilidad mediante mayor tolerancia de adquisición frente a obsesión por optimizar el costo por compra
Canales de adquisiciónEscalar un negocio a través de pauta publicitaria es viable cuando se está dispuesto a pagar un mayor costo por adquisición de cliente, mientras que intentar escalar forzando permanentemente una reducción del costo por compra frena el crecimiento exponencial.
It's very easy to scale a business if you're willing to spend more money to acquire a customer. It's very hard to scale a business if you're constantly trying to improve your cost per purchase. You can get your money to work harder, but the amount of money you get back won't be exponentially greater.
Escúchalo en el minuto 21:20
Canales de adquisiciónTener menos anuncios activos pero mejor alimentados con presupuesto permite que los algoritmos acumulen más datos por pieza, logrando un rendimiento más predecible, estable y resistente a la fatiga que lanzar continuamente lotes de anuncios nuevos.
And yeah, even if it stays the same, having fewer pieces that are more informed means that those ads have more data, which means their output is going to be more projectable, more predictable, and more stable. If you stop launching a bunch of new ads, those ads aren't going to fatigue. In fact, they're going to stabilize.
Escúchalo en el minuto 40:48
Optimizar por beneficio bruto por transacción (GPT) y volumen absoluto en vez de ROAS o porcentaje de margen
Medición y experimentosLas cuentas se pagan con dinero real y no con márgenes porcentuales. El ROAS aislado es engañoso: un ROAS bajo con gran volumen de gasto puede generar sustancialmente más beneficio neto absoluto que un ROAS alto sobre un presupuesto mínimo.
GPT. If you do not have this metric inside of your ad account, you do not understand whether or not you are making any money. Margins aren't money. You can't pay your bills on margin. You pay your bills with currency. GPT or gross profit per transaction is real money. Simply put, gross profit margin is very different than actual gross profit volume. Now, to explain this very simply, 100% profit margin. A ROAS of 2 on $100 is $100 profit. A ROAS of 1.01 on a million dollars is 100 times more actual money. And for that matter, ROAS obviously needs to have context and volume.
Escúchalo en el minuto 7:32
Canibalización de canales: un ROAS alto en Meta puede perjudicar la rentabilidad global
Medición y experimentosUn anuncio nuevo que consume presupuesto y muestra un ROAS elevado en Meta puede estar perjudicando las finanzas globales del negocio si solo está canibalizando ventas del fondo del embudo que habrían entrado orgánicamente por Google o email. El impacto real de los anuncios debe medirse contra el crecimiento neto del negocio y la cuenta bancaria, no exclusivamente por las métricas aisladas de la plataforma…
However, if that ad does get spent, what happens? If that ad gets spent and my overall campaign performance looks better and my bank account looks worse, it's a bad ad. And often that happens because your ROAS goes up because you have more ads at the bottom of the funnel and you're just stealing volume from Google and email. And after a few days, you've just dried up the funnel by making your least efficient marketing channel that is designed to amplify your overall business model more effective at stealing opportunity from where you are far more profitable. Conversely, it's also completely…
Escúchalo en el minuto 51:13
Aplicación del método científico en publicidad mediante una línea base de control
Medición y experimentosPara experimentar con rigor en publicidad pagada, es imprescindible simplificar la cuenta hasta tener un grupo de campañas con comportamiento predecible y medible que funcione como control; lanzar cientos de anuncios simultáneos impide aislar variables y vuelve imposible atribuir mejoras reales.
You have a control and a variable. If you have 100 ads that all behave wildly differently every single day, showing their ads to completely different people all the time, and you're constantly launching new ads with a bunch of different campaigns that all have different definitions of success, you don't have a control. if you don't know how your ads are going to behave, then you don't have a control. If you don't have a control, then you cannot test. So the first thing you have to do is simplify your account to the point where you have your single best investment optimized around different…
Escúchalo en el minuto 45:39
Medición y experimentosLa optimización de anuncios pagados debe gestionarse bajo el método científico y no mediante tácticas de 'hackeo' o lanzamientos masivos aleatorios (como 100 anuncios por semana), ya que las plataformas publicitarias son sistemas de machine learning impulsados por datos estructurados.
The ads manager is a piece of software designed by some very smart engineers. That's a machine learning algorithm with more data than literally anything else in the world. And its sole purpose is effectively designed to give people positive user experience while conditioning human behavior. Why not use that as your factory and you worry about the raw materials going in and selling what comes out? So if that's the case, it's science. Let's use the scientific method. When somebody says launch 100 ads a week, ask them how does this fit into the scientific method? And if you're testing outside…
Escúchalo en el minuto 43:30
Medición y experimentosLa optimización de anuncios pagados debe gestionarse bajo el método científico y no mediante tácticas de 'hackeo' o lanzamientos masivos aleatorios (como 100 anuncios por semana), ya que las plataformas publicitarias son sistemas de machine learning impulsados por datos estructurados.
The ads manager is a piece of software designed by some very smart engineers. That's a machine learning algorithm with more data than literally anything else in the world. And its sole purpose is effectively designed to give people positive user experience while conditioning human behavior. Why not use that as your factory and you worry about the raw materials going in and selling what comes out? So if that's the case, it's science. Let's use the scientific method. When somebody says launch 100 ads a week, ask them how does this fit into the scientific method? And if you're testing outside…
Escúchalo en el minuto 43:30
Evitar campañas de tráfico para testear creativos publicitarios
Canales de adquisiciónUtilizar campañas con objetivo de tráfico para validar creativos contamina la data del píxel y la cuenta publicitaria con clics de baja intención de compra, distorsionando el aprendizaje del algoritmo.
Please stop creative testing with traffic campaigns that only pollutes your entire existing marketing pool with people that just want to click on ads that have no intent in a way that does not improve your relationship with the platform at all. Like getting bad data to pollute your marketing pool, to make poor decisions, to solve, to not
Escúchalo en el minuto 35:35
Medición y experimentosLanzar decenas de anuncios simultáneamente en presupuestos fragmentados (ABO) genera canibalización en el fondo del embudo: el algoritmo los prueba primero contra la audiencia más templada, provocando que los 'ganadores' triunfen por mera casualidad de atribución de último contacto y no por capacidad real de prospección.
Now, with all that being said, if you launch a bunch of ads in the first few days, those ads are going to be launched against the warmest audience. If you launch 80 ads at a hundred bucks a pop, you're basically saying, if I let all of these things dramatically cannibalize, compete against each other, which one's lucky enough to be the last touch point before somebody does something? Well, inherently they have two problems right away. First off, the reason a lot of those tests won or lost is random distribution and very, very low levels of signal of data. A simple 4PI analysis debunks that…
Escúchalo en el minuto 26:53
Reasignación de presupuesto publicitario hacia anuncios de mayor rendimiento
Canales de adquisiciónMover presupuesto de creatividades con bajo ROAS hacia aquellas con mejor rendimiento incrementa la rentabilidad global y acelera el aprendizaje del algoritmo, incluso si el ROAS individual disminuye ligeramente por el efecto de escala.
Just simply take some of the money from the 1.2 ROAS and move it to the 1.8. If you are able to shift all of that money from the 1.2 to the 1.8 and your blended is still at 1.5 or any better, you're in a phenomenally better place. Because all of your ads are going to learn and they're going to get better. And your output and your profitability is going to be better too. Odds are maybe you double your budget and your 1.8 goes to a 1.7 or a 1.6. Great. You're making substantially more money.
Escúchalo en el minuto 15:19
El volumen excesivo de creatividades no genera un escalado publicitario sostenible
Canales de adquisiciónTratar la creación y lanzamiento de anuncios masivos como boletos de lotería incrementa la complejidad del sistema y no constituye una estrategia sostenible de escalado; en paid media, lo simple escala y lo complejo fracasa.
The path to success in the ads platform is not to do low confidence efforts that make things more complicated. Simple scales and complex fails. Creative volume is not the path forward for more confidence or more scale. You're going to hear a lot of people say the way to be successful is to just launch more and more ads, take more and more shots, try to get more and more lottery tickets. Can that be a path for you to see sustainable results? Absolutely not.
Escúchalo en el minuto 1:50
Preguntas de cualificación para embudos de aplicación en programas o servicios avanzados
ConversiónPara filtrar prospectos calificados en programas formativos o servicios de alto valor, utiliza un formulario de aplicación que recoja el nivel de gasto publicitario mensual, el modelo de negocio, el principal obstáculo en resultados, la experiencia previa con otras ofertas del mercado y la visión concreta de éxito.
How much are you spending? Which bucket are you in? Let's say you're spending $25,000 to $100,000 a month. What's your biggest struggle? My results are inconsistent and unpredictable. What's your business model? I run a brand or I run ads for clients. What's your experience with ad trainings? I bought courses that are over-promised and undelivered. What does winning after it look like? I'm going to review these every day. And I'm going to reach out to people that are serious.
Escúchalo en el minuto 56:59
Canales de adquisiciónLanzar constantemente anuncios nuevos interrumpe la fase de aprendizaje del algoritmo y canibaliza las conversiones fáciles ya generadas por creatividades existentes, creando una falsa fatiga publicitaria provocada por el propio gestor de la cuenta.
So if you launch new ads, what that basically does is say, your ad got lucky. I'm going to push it out into the real world. And then as it begins to learn, instead of letting it make mistakes and then get smarter, I'm just going to steal all the low hanging fruit that takes advantage of all the data that I paid for it to learn with. And just give that all to something new. You're manufacturing ad fatigue. It's operator error. So what you're ultimately doing is saying, I'm going to test a whole bunch of things with very low integrity data at a very low volume based on the bottom of the funnel…
Escúchalo en el minuto 31:20
Canales de adquisiciónLanzar constantemente anuncios nuevos interrumpe la fase de aprendizaje del algoritmo y canibaliza las conversiones fáciles ya generadas por creatividades existentes, creando una falsa fatiga publicitaria provocada por el propio gestor de la cuenta.
So if you launch new ads, what that basically does is say, your ad got lucky. I'm going to push it out into the real world. And then as it begins to learn, instead of letting it make mistakes and then get smarter, I'm just going to steal all the low hanging fruit that takes advantage of all the data that I paid for it to learn with. And just give that all to something new. You're manufacturing ad fatigue. It's operator error. So what you're ultimately doing is saying, I'm going to test a whole bunch of things with very low integrity data at a very low volume based on the bottom of the funnel…
Escúchalo en el minuto 31:20
Optimización presupuestaria reasignando gasto hacia anuncios por encima de la media
Medición y experimentosPara mejorar la rentabilidad de una cuenta publicitaria es más efectivo reasignar el presupuesto aumentando la inversión donde el CPA está por debajo del promedio y el GPT por encima (y recortando lo inverso) que crear y lanzar continuamente nuevos anuncios.
The simplest way of making things better is to spend more money where you're better than your target or better than average and spend less money where you want. So the point here is where is your CPA below average and your GPT above average? Where is the inverse of that true? How do you make your ad account better? It's not launching more ads. In fact, launching more ads just makes the decision of how to make it better more difficult.
Escúchalo en el minuto 14:36
Uso del indicador GPT (AOV menos CPA) para comparar anuncios
Medición y experimentosEvaluar los anuncios mediante el indicador GPT (valor promedio del pedido menos el costo de adquisición: AOV - CPA) neutraliza las distorsiones de atribución externa, ya que el ruido de mercado afecta por igual a todos los anuncios dentro de la misma cuenta publicitaria.
When you look at GPT, what you're actually looking at is AOV minus CPA. Of all the attributed transactions, what is the value of that transaction minus what did it cost to get that transaction within the platform? Now, why this is incredibly more valuable information is because it assumes all of the noise as a constant. If you have a bunch of marketing channels, sure, your CPA might be lower or higher, but how two ads compare against each other, they're both impacted by that outside force in the same way. The AOV is not going to be impacted by those outside forces in nearly the same way. But…
Escúchalo en el minuto 12:52
Optimización de cuentas saturadas mediante la desconexión progresiva de anuncios deficientes
Canales de adquisiciónEn cuentas publicitarias saturadas de creativos, la optimización inicial debe consistir en apagar de forma paulatina los anuncios de bajo rendimiento en lugar de crear nuevos, permitiendo concentrar el presupuesto y estabilizar los resultados durante semanas.
first thing you have to do is simplify your ad account to the point where every single campaign has a knowable and projectable Then you launch a few elements and a controlled effort to try to replace whatever piece is doing its job most poorly. And they had 300 ads going. You don't have to launch another ad till June, July. So turn off the ads that suck. If you turn off the ads that suck, and maybe every day you turn off one or two, you have a month and a half where every single day gets better. And if you don't do that, you'll have no idea what happens.
Escúchalo en el minuto 47:56
Canales de adquisiciónLanzar un volumen desmedido de anuncios simultáneos oscurece el análisis de rendimiento: la plataforma entrega las creatividades en órdenes y combinaciones impredecibles junto a la competencia, haciendo imposible atribuir con certeza el impacto real de cada pieza individual.
And when trying to make money, one of the biggest lies being perpetrated by basically everybody is more ads is good. What I'm saying is instead of worrying about launching more ads, what you should be worried about is the quality of the decisions that you're making. The problem is that completely ignores how the platform actually works. All of your ads are shown to people in different orders at different times on different platforms. And the delivery of all of the ads from every marketer to the actual person plus the organic content plus their experience with everything else that you will see…
Escúchalo en el minuto 17:45
Canales de adquisiciónEl inventario publicitario de alta calidad se gana optimizando la experiencia del usuario y la tasa de acción estimada; utilizar topes de coste (cost caps) o de puja (bid caps) no mejora la calidad de la audiencia, sino que fuerza al sistema a competir por inventario residual de baja calidad.
And by the way, before we get too far down the road here, I just want to say, if you're doing cost caps to prevent your bids, what you're ultimately doing is saying, can I more effectively buy absolute shit inventory? Quality inventory comes at a higher, uh, quality inventory is earned and comes at a lower cost. Shit inventory is more expensive. You cannot cost cap or bid cap your way into better quality inventory. You're just outpricing other people competing for the dregs. So you have to win the quality game of the user experience.
Escúchalo en el minuto 26:06
El ciclo de vida del anuncio requiere permitir la caída temporal de rendimiento para salir de la audiencia templada
Canales de adquisiciónUn anuncio nuevo suele funcionar bien al inicio con la audiencia fácil, para luego empeorar mientras el algoritmo explora y testea nuevos nichos de audiencia fría. Lanzar anuncios nuevos continuamente interrumpe este bucle de datos antes de que el algoritmo encuentre su segmento rentable y se estabilice.
When an ad is launched, that ad is deployed against your low hanging fruit, your warmest The machine wants to understand how your audience would receive that piece of content. What you see is ads tend to work for the first few days and then they get a little bit worse, And then if you let them run, they can get a little bit better. Unless, unless you do the brilliant thing of launching more ads. Because when those go ads go out and they get worse, it's because they're trying to get away from the low hanging fruit to see where they can win and where they can lose. And odds are they might lose…
Escúchalo en el minuto 25:37
Oferta y precioLa primera transacción es la más costosa en todo el embudo; por ello, en modelos como ecommerce o SaaS el objetivo real no es vender un producto aislado, sino adquirir y rentabilizar todo el recorrido del cliente a través de transacciones posteriores.
The most expensive transaction you will get in your entire customer journey is the first one. If you're in e-commerce or SaaS, you absolutely see the upside. Your entire business is actually predicated on generating that upside of additional transactions within the customer journey. You are not selling products. You are buying customer journeys.
Escúchalo en el minuto 5:49
Medición y experimentosEl ROAS diario resulta engañoso para la toma de decisiones operativas porque divide ingresos desfasados (producidos por impactos acumulados de varios canales a lo largo del tiempo) entre el gasto del día, el cual no es la causa directa de esas ventas inmediatas.
And in fact, if you turn off a bunch of ads and you, like, go into it later, it's completely possible for you to say, well, I turned off all my budget today on this. I turned off this ad and the next day a sale happened. Yeah, because somebody that saw that ad bought. The revenue today is neither accurate because it's not inclusive of all of the money, nor is it because of your spend today. Because, again, if you stop spending today, you might see revenue tomorrow from an ad that was shown today or yesterday or last week. So if we know the revenue number is neither accurate or directly caused…
Escúchalo en el minuto 10:18
Medición y experimentosIntentar atribuir con exactitud el último impacto en recorridos de compra multicanal y no lineales es costoso e inútil, ya que los datos de atribución solo tienen valor en la medida en que sean directamente accionables.
And what attribution model are you choosing? But that's also subjective because every customer journey is completely different. I have literally been in businesses that spent over $100,000 to figuring out modeling. Every single time we have that conversation, it's whatever attribution model makes the vice So if you're working on an attribution model to say, well, this is the ad that got somebody to buy, you're spending money and wasting time trying to extrapolate the very last touch point of a customer journey that is decades long, where you have no control over 99.999% And if you don't have…
Escúchalo en el minuto 37:16
El balance bancario es la métrica definitiva sobre las métricas de plataforma
Medición y experimentosEl éxito de los esfuerzos de marketing debe evaluarse por el incremento de dinero real en la cuenta bancaria y no por definiciones miopes o métricas aisladas dentro de las plataformas de anuncios.
The fundamental gap that I'm seeing more and more people address or more and more people bring to the table, unfortunately, is that amplifying their myopic definition of success is somehow a way for you to consistently move your business forward. We have one definition of success in our business. We have one scorecard. That is the bank account. And you have one simple job. With your entire marketing effort, are you good enough to take more advantage of amplifying your bank account?
Escúchalo en el minuto 2:27
Canales de adquisiciónLa plataforma publicitaria de Meta no está diseñada para maximizar el retorno de inversión del negocio, sino para ofrecer la experiencia de mayor calidad y retención a sus usuarios finales, buscando consistencia en las conversiones antes que picos de rendimiento.
Because remember also, the machine does not care about your performance. The meta ads platform is not optimized to give you the best results. It's optimized to give their end users the highest quality experience. Your job is to marry that with the best results. And it in conversion campaigns tries to give you the most consistent result.
Escúchalo en el minuto 30:42
Medición y experimentosNo se debe evaluar el rendimiento de anuncios en prueba contra el control día por día antes de que el experimento se estabilice estadísticamente y en volumen de gasto, ya que las fluctuaciones iniciales distorsionan la toma de decisiones.
Would you compare the performance of the control against the other ads by day? No. Until the test stabilizes, we don't actually have any idea. We don't have any idea how the performance is going to go.
Escúchalo en el minuto 52:28
Canales de adquisiciónModificar continuamente lo que funciona en marketing impide desarrollar maestría en canales de adquisición alternativos, llevando al error común de asumir que otros canales no sirven y atrapando al negocio en una dependencia forzada de más publicidad pagada.
And then you change what works on a regular basis. How good do you think you're going to get at every other marketing channel? And then you say, well, the other stuff doesn't work. I just got to take more ads. Don't make everything worse. And then be surprised when your best, most valuable excuse is to be a victim.
Escúchalo en el minuto 55:52
Oferta y precioPrograma intensivo de mentoría de 12 semanas limitado a 12 participantes con más de 20 horas de acceso directo comercializado a un precio de 15.000 USD.
Again, 12 people are going through it. It's $15,000. And I'm going to hold your hand over 12 weeks where you get over 20 hours of my time to show
Escúchalo en el minuto 24:46
Seguir leyendo
Esta es una de las 1872 entrevistas despiezadas. Puedes
ver todas, leer las ideas por tema o
empezar por la ruta guiada si no sabes por dónde.