Marketero · Entrevistas
5 Shocking MTA Revenue Secrets Ben Tiernan Doesn't Want You to Know
20 ideas de marketing sacadas de «5 Shocking MTA Revenue Secrets Ben Tiernan Doesn't Want You to Know», de Professor Charley T. Sobre todo de medición y…
- 20 ideas despiezadas
- Fuente Professor Charley T
- Dura 54 min
- Publicada en 2024
Ver la entrevista entera en YouTube
De qué habla
Las 20 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 20 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.
Medición y experimentosLa atribución multitáctil puede empujar a los equipos a tomar decisiones tácticas minúsculas y reactivas que generan ganancias aparentes a corto plazo pero descuidan la medición de la incrementalidad y los cambios de negocio significativos.
what I'm hearing a lot is multi-touch attribution is a good idea and it should help businesses often the people that install data the downside is it empowers folks to make very minute decisions very actionably and a very low level which might get in the way of the overall like success like stepping over you know just picking up every penny on the ground instead of making money instead of making real significant and one of the issues is nobody the biggest question is well how do we measure incrementality how do we understand that this thing is doing anything so instead of being able to solve…
Escúchalo en el minuto 48:26
Medición y experimentosLa atribución multitáctil puede empujar a los equipos a tomar decisiones tácticas minúsculas y reactivas que generan ganancias aparentes a corto plazo pero descuidan la medición de la incrementalidad y los cambios de negocio significativos.
what I'm hearing a lot is multi-touch attribution is a good idea and it should help businesses often the people that install data the downside is it empowers folks to make very minute decisions very actionably and a very low level which might get in the way of the overall like success like stepping over you know just picking up every penny on the ground instead of making money instead of making real significant and one of the issues is nobody the biggest question is well how do we measure incrementality how do we understand that this thing is doing anything so instead of being able to solve…
Escúchalo en el minuto 48:26
Medición y experimentosPara superar los techos de crecimiento (escalar de 2-3M a 10M+), las empresas deben evaluar el impacto cruzado entre canales y la incrementalidad real en lugar de limitarse a herramientas de atribución multitoque (MTA) que solo asignan crédito retrospectivo a campañas individuales.
mtas the issue that I'm seeing is that people are using it to justifi to give them enough information to execute old thinking that's getting in the way it's really setting them up for scale and understanding the impact of one marketing effort on another there so it's leaning you more into the ability to understand incrementality and less in the hope of just attribution to many sales to one businesses the information that they care about is how do I find incrementality in my business not what worked for me yesterday that can throw more money at today or the successful ones and that media mix…
Escúchalo en el minuto 21:14
Fundamentos de modelado econométrico (MMM) para medir impacto publicitario sin depender de atribución de plataformas
Medición y experimentosPara medir con precisión el impacto real de los medios sin el sesgo de atribución de las plataformas publicitarias, se ingieren métricas de entrega publicitaria (impresiones, gasto, clics) en la granularidad temporal más fina posible (minutos, horas o días) y se correlacionan contra la cifra única de ventas reales del negocio, controlando además por factores exógenos (economía, estacionalidad, actividad de PR y…
so okay so we ingest a lot of data um coming in you know with minute increment is that we can get smaller the increments the uh more the uh um the the faster we can model basically um then and you know more accurate kind of reads we can get uh and we're ignoring sales so we don't want to know what the platforms think in terms of sales um so but we're taking in the Impressions the spend the uh um clicks the engagements sales or the finance data from the in other stuff um if we can bring in um industry data market share data if we're kind of global economic data we can bring in political data…
Escúchalo en el minuto 31:36
Riesgo de sobreconteo al sumar los reportes aislados de cada plataforma publicitaria
Medición y experimentosCada plataforma publicitaria reclama la atribución completa dentro de su ventana de conversión; por ello, sumar directamente los reportes de ingresos o conversiones de cada canal suele inflar las métricas entre un 20% y un 300% respecto a los resultados reales del negocio.
or a search and every platform kind of takes credit with an attribution of what they did and when you add up all the reports of every platform you get sometimes 20% sometimes 300% more transaction volume in Revenue than you actually did so you have to understand know if you're looking at the conversion by platform if you're looking at Facebook and saying it's driving 20 conversions and you look at uh search and it says it's driving 20 conversions and it looks at whatever and it says it's driving 20 conversions you add those up you're going to have you're you're going to see you know 60…
Escúchalo en el minuto 7:39
Prueba de inversión a pérdida total para medir el efecto halo entre canales
Medición y experimentosAl evaluar aumentos de inversión en un canal (como gastar un 10% más), conviene modelar o testear la hipótesis bajo el supuesto de que generará cero conversiones directas adicionales; esto permite verificar si el gasto se justifica por su efecto halo en tráfico directo, orgánico, email y audiencias amplias.
changes dramatically right like and you can't scale bottom funnel search for say what happens if I go 10% more here plan that and be safe is I I do what I call a total loss investment what happens if I were to spend 10% more here at zero more zero more transactions I here and set it on fire would that still be good for my business because I if I do that and assume no additional revenue from that channel I'm now looking at well what happens what's the halo effect on emails and uh organic search and you know and GA the direct organic or all happens if I do less retargeting or interest group…
Escúchalo en el minuto 28:34
Modelado de atribución con ventana retrospectiva para mapear la transición del embudo
Medición y experimentosAnalizar el recorrido del usuario mediante una línea de tiempo retrospectiva permite observar cómo las acciones de la parte alta del embudo aportan correlación y valor a las ventas ocurridas días después, facilitando recomendaciones precisas para redistribuir el porcentaje de inversión publicitaria entre canales.
something like a user Journey except for it's more like a look back timeline that says hey at day eight these things had the highest contribution uh or correlation to your eventual sale which was seven days later at day six these things had a higher contribution to your sale and it actually shifts you can see the the kind of flow of upper funnel into lower then we can also translate it into spend potential which is like hey you can continue to invest here and continue get positive Roi um you should decrease here and get there's a recommendation engine which is literally just change your spend…
Escúchalo en el minuto 35:38
Medición y experimentosLos modelos de atribución multitoque suelen sufrir de manipulación política para justificar presupuestos o proteger agencias y gestores; además, optimizar decisiones futuras basándose en datos retrospectivos falla en cuanto se alteran la oferta o el mix de medios global.
integration and stuff and often I found like well we're going to find a weighted average that makes was what made the manager of each platform look good it was like the balanced waiting of everybody keeps their their job right and I've also seen ad agencies do it where the balanced waiting is basically they put a thumb on the scale to make sure that they can take as much credit for something that happens and I working but it's like yeah but that's only in the medium mix you have retroactively when looking at that customer base and as soon as you change the offer as soon as you change the…
Escúchalo en el minuto 13:10
El peligro de redistribuir presupuesto reactivamente según el crédito de atribución
Canales de adquisiciónVolcar todo el presupuesto de adquisición hacia el canal o punto de contacto que recibe más crédito en la atribución desequilibra el ecosistema de marketing. Al saturar ese canal, la efectividad cambia y se termina persiguiendo métricas en lugar de enfocarse en la rentabilidad y el flujo de caja global del negocio.
MTA is that and especially in the lower end businesses that are in the million to five million range where they're phase they're looking at it to see well what gets credit so I can spend more there and that's kind of the thought process and that makes sense the problem is as soon as you throw everything at what's working the balance shifts so something else looks better and then something and you you end up chasing path I think the biggest liability it presents is that it's moving people away from the business fundamentals of investing in the acquisition of profitable customer Journeys in…
Escúchalo en el minuto 17:52
Modelado predictivo de Media Mix (MMM) con Big Data para planificar la reasignación de presupuestos
Medición y experimentosEn lugar de usar la atribución multitáctil sólo como un reporte estático del pasado, aplicar modelos de Media Mix (MMM) procesados con Big Data y Machine Learning permite proyectar incrementalidad y predecir resultados futuros antes de redistribuir el presupuesto publicitario entre canales digitales y tradicionales.
does um essentially they're big data media mix models MTA like insights um and the reason why kind of move across platforms and across uh channels so we move into linear and Big Data mmm um but we're able to granular data um and then run it through machine learning right uh generally is modeling more specifically is Big Data modeling um with uh machine learning um as essentially way to make that in in order to make predictions on on what will work in the future so the of build um reasonable predictions on should reallocate your budget and then
Escúchalo en el minuto 25:13
Mayor granularidad temporal en los datos acelera y precisa el modelado de marketing
Medición y experimentosA menor incremento temporal en la captura de datos (minuto u hora frente a semanas o meses), más rápido se pueden entrenar los modelos estadísticos de marketing y lecturas más precisas de correlación e impacto se obtienen.
platforms right uh and ideally it's coming in you know with minute increments hour increments day increments whatever the smallest increment is that we can get smaller the increments the uh more the uh um the the faster we can model basically um then and you know more accurate kind of reads
Escúchalo en el minuto 31:52
El fin del rastreo granular obliga a volver a la medición por incrementabilidad agregada
Medición y experimentosCon la eliminación progresiva de las cookies y el endurecimiento de la privacidad en sistemas operativos móviles (como iOS y Android), los especialistas en marketing ya no podrán depender del seguimiento mediante píxeles a nivel individual; la medición digital deberá evolucionar hacia el análisis de impacto agregado e incrementabilidad, tal como se medían históricamente canales tradicionales como la radio o la vía…
mean we're going to wake up and the cookie will be gone right we're going to wake up and what happened to iOS 14 on Apple devices is going to happen on Android devices like we're going to reach a spot where you cannot rely on the crutches that people have been really using for the last five or 10 years and you're going to really have to go back to Bare fundamentals of like what happens when I put up a billboard what happens when I run radio like what happens when I put coupons in the newspaper like if you for what it's worth as shocking as it may be billion dooll companies were built before…
Escúchalo en el minuto 51:00
Medición y experimentosLos modelos de atribución multitáctil registran qué puntos de contacto ocurrieron antes de una conversión, pero no demuestran causalidad. Basar decisiones de inversión en estos modelos suele generar narrativas subjetivas sobre el recorrido del usuario, apoyándose en tecnologías de cookies y etiquetas cada vez más limitadas.
that you don't actually see um causation and the thing you assume a correlation you assume causation when you see a lot what you're doing is you're just before deciding and you and you you tell yourself a story and the story makes sense and you've brought meaning to the data um but the meaning is uh storytelling which makes sense um vision of what's happening but it doesn't actually tell you why and then we don't realize that we're bringing the why to it and then making uh uh investment decisions accordingly um so it relies on tagging and cookies and you
Escúchalo en el minuto 16:00
Medición y experimentosLas restricciones de privacidad (como las actualizaciones de iOS y la eliminación de cookies de terceros) han transformado la medición publicitaria, forzando a plataformas como Meta y Google a sustituir el rastreo determinista por la atribución modelada.
except um the deprecation of cookies right and you know iOS 15 update and like like these things have been like major um shifts for um how these systems optimize and how these systems run and you know and I think that at least two major W Garden platforms uh kind of Facebook and and um Google have done a pretty good job of kind of making the adjustments on the back end and they're moving to modeling and kind of uh uh the Google analytics uh moved um they have kind of their new model version for kind attribution um and I think that's correct I think it's the only possible uh Way Forward
Escúchalo en el minuto 24:05
Medición y experimentosLa atribución multitáctil rastrea el recorrido secuencial del comprador mediante etiquetado exhaustivo de impresiones y clics en cada canal y creatividad, asignando un valor proporcional de la transacción final a cada punto de contacto para no atribuir todo el éxito únicamente al último clic.
you know uh heavy tagging um in which the tagging get everything kind of Tagg tagged correctly so that um the MTA system itself can count every impression every click Etc that happens um and then and then look at the journeys um of individual buyers um or converters and kind of see all right so you know in sequence what what were what were people kind of assigned back uh a value to each of those touches according to how many people that did those touches converted um but it's a counting up of sequential Impressions and sequential engagements with advertising as well as owned media um and then…
Escúchalo en el minuto 5:14
Medición y experimentosUn gasto publicitario elevado en un único canal no justifica necesariamente herramientas complejas de modelado; la verdadera necesidad de atribución y análisis avanzado surge cuando el presupuesto se distribuye entre múltiples canales y cientos de campañas simultáneas.
think having those minimums is also making sure that the quality of the insights is there and the action ability is there I mean you might be running you know a million a month which is great but you're on like one channel and it's really simply you're just crushing it this probably isn't the right fit I yeah I'm running two million a month across like five channels with maybe a 100 campaigns it's like okay this is a mess like that's the kind of person that really needs it right because I don't think I don't think people realize how difficult the decision making
Escúchalo en el minuto 46:10
Umbrales mínimos de inversión y datos para modelado de atribución avanzado
Medición y experimentosPara construir modelos econométricos y de atribución robustos se requiere suficiente volumen y dispersión de datos: con un eCPM promedio de aproximadamente 10 USD, se necesita una inversión publicitaria de unos 5 millones de dólares distribuida en 3 o 4 plataformas y un año de histórico de datos de ventas y conversiones.
basically um and that tends to be like around if your average like just ecpm across your total marketing spend you know everything from P PV to um digital um to whatever search uh is about $10 then maybe about $5 million kind of and across three or four platforms um will give you enough um versatility uh in your in your data for us to be able to build models basically and we should be able to look back a year ideally um in your platforms you spend about $5 million over a few different platforms and we go back a Year's worth of data and we get sales data and we can align everything um we've…
Escúchalo en el minuto 44:11
Medición y experimentosEl principal valor de la atribución multitoque (MTA) para las marcas radica en medir cómo la inversión en etapas superiores y medias del embudo (upper y mid funnel) contribuye y se traduce finalmente en conversiones directas.
think it's actually a nice uh service for Brands and brands have been adopting um mtas more regularly because they're looking for more measurement on kind of what's you know what's contributing uh in the end to Brand development and you know how does our brand work ultimately convert into sales um and MP has proposed to answer that at least in lot of the main adoption which is trying to answer the question for brand advertisers how does our upper funnel mid funnel uh end up translating into some sort of a online conversion uh and I think it's most helpful there and
Escúchalo en el minuto 12:02
Medición y experimentosAnalizar los resultados del negocio aislados de la dinámica del mercado impide entender por qué fallan o funcionan las campañas; con frecuencia, las fluctuaciones de rendimiento se deben a variaciones macro o de comportamiento del consumidor que no pueden detectarse dentro de las plataformas de anuncios.
and so many people look at their business in isolation and they have no idea of understanding what other people what is the impact of you know how much money is being dumped in by the market sometimes the reason things aren't working the way that you planned it is because for two weeks the world Works differently in a way that you couldn't possibly perceive inside of the ad
Escúchalo en el minuto 40:28
Medición y experimentosEl Media Mix Modeling (MMM) tradicional elaborado por firmas de consultoría suele entregarse con desfases de meses en reportes estáticos, resultando insuficiente para operaciones digitales que requieren ajustes continuos de inversión basados en flujos de datos en tiempo real.
why uh the mmm media mix models were done by large consulting firms who came out with like fairly kind of like chunky but useful kind of uh insights uh but they weren't like real time decision making running through kind of like you know data streams they were just like every six months you get like a new like this is TVs working for you at this m
Escúchalo en el minuto 38:47
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.