Speaker 1: Hey and welcome to Funnels YouTube channel where I talk all things digital marketing, data visualization and analytics. All the great content you need to shortcut your way to being a better marketer. Although I'd like to, I don't actually know everything about digital marketing. Meaning from time to time I enlist the help of experts to go deeper into a topic than I possibly can. So in this video we have Kvalata Lundberg from renowned marketing intelligence agency NEPA who have done over 50 MMM projects and she'll be diving deeper into MMM and actually presenting an MMM case study that they did with a client so you can get a better idea of the outcomes of MMM and how it might be applicable to your business. So take it away Charlie. In this video I will walk you through a
Speaker 2: marketing mixed modeling project we did at NEPA to show you what the outcome of an MMM project is and how the results can be used in order to optimize your marketing and impact the bottom line. Why do companies want to look into MMM? But before I get started I just want to clarify a few terms that I'll be using throughout. When I talk about media investments I'm referring to the money a company spends for a visitability through advertising. These media investments can be aimed at short-term sales activation such as an ad with a link to a specific product or discounts or it could be aimed at long-term brand building such as a commercial with the purpose of improving the brand awareness. Thus increasing sales over time. And lastly by short-term I'm talking about days or weeks and with long-term I'm talking about months or even years. Getting into the case. Okay let's dive into the case. I'm going to refer to the client as the client rather than the specific company name. But for context it's a market leader within retail. The client had historically been focusing a lot on short-term sales activating media investments but recently they shifted to incorporating more long-term upper funnel brand spend as well. So the main question here was that they wanted to assess the change in investment and understand how to balance short-term and long-term communication going forward. The core questions we needed to answer in order to answer the overarching business question was. One should they increase their total media budget or not? Is it worth it doing so? And two is a split between different media channels well balanced or should they reprioritize? What does the process look like for an MMM project? Now I'll walk you through what the process actually looks like for an MMM project. After agreeing on the business questions we need to identify what data is required for the modeling. Then we visualize that data to make sure it is all correct and fit for purpose. Once that's confirmed the modeling can start which takes us to the outcome and implementation. Let's look a bit closer at each of these milestones. What data do you need to execute an MMM project? When the business questions were defined we sat down with the client to identify and discuss which data points we need to use for their specific business situation. One of the core data inputs to the model is the weekly sales data. In this case our client had both online and offline sales. In addition to that the weekly media spend is crucial. The accuracy of the output increases with a number of years included in the model. But there is also a risk of basing the model on data that is now old and outdated. I just want to take a second to give a few variables that could mean all the data is out of date for the purpose of MMM. One variable could be if your company is growing extremely fast because three years ago is an age at fast-growing businesses. Another could be if you have acquired a company or drastically altered your offering. With that said we have found that the best models rely on three years data. For larger companies weekly brand tracking data also brings additional value to the model. It's not necessary to get useful outputs but helps to make a more robust model. If you're not familiar brand tracking data is continuous data collection of brand KPIs such as ad and brand awareness, brand consideration and brand preference of the client. For this MMM case the client had implemented brand tracking with us enabling us to apply additional in-house algorithms. Brand tracking data is and always will be noisy but we reduce the noise by applying our propriety NEPA brand noise reduction algorithm to improve the quality of the brand KPI implemented in a brand modeling by reducing noise from the sample. Also to better estimate the brand effect in the brand modeling and remove arbitrariness from the process we also applied our brand health index. Luckily this clients also had performed several campaign measurements over time where the creative communication was evaluated on KPIs such as level of observation and how consumers liked their content. This allows us to quantify and measure not only the effect from each dollar spent on media investment but also the financial value of a strong creative content in advertising. In order to answer our main questions we want to understand the effect media investment has on sales but to figure out what we need to take lots of other variables into account. What is often overlooked by marketing intelligence teams that we find incredibly valuable is the clients industry knowledge as some variables may be relevant to specific industries but not for others. For instance the weather is a highly relevant variable for a company selling ice cream but wouldn't at all be useful for a company selling toilet paper. Other common external variables to add are discounts, salary weeks and COVID related effects etc. We then need to structure all the media data and other relevant data specific for the clients industry. We then share the structured and visualized data with a client so that we can both get an overview of the inputs to the model. This also gives us an opportunity to discover missing data so don't skip this step better to find out before start running models. The modeling step. When the data is confirmed to be correct the modeling can begin. This might be the most confusing part so stay with me. We built two different models. One focusing on the direct effects from marketing the other on brand. In this way we are able to see how much each media affects the sales in the short term. Both in-store and online and the brand in long term. To clarify MMM should not be confused with attribution modeling. The benefit of using marketing mixed modeling instead of approaches like attribution modeling is also that long-term sales activities can be considered which is difficult when using individual level data and the modeling is not as sensitive to regulatory changes such as GDPR. The output from the modeling phase is the contribution on sales for the different models. With this we calculate the return on investments or ROI per media which we have on a total level for the direct effects and on brand respectively. As you can see based on the client's current media mix it is clear that there's a possibility to reprioritize marketing investments. Investments can be moved from the weaker performing channels to channels with a higher ROI. For this client we gave the recommendation to increase investments in TV by 4%, increase investments in online web TV, banners and search by 3%, decrease investments in print by 3% and maintain the investment levels for outdoor. This showed that we could increase sales by 15% without increasing the media budget and instead optimizing the existing channel mix. This is what we might call a conservative recommendation which may be a common scenario given the current financial situation of decreased media budgets all around. MMM also gives us the possibility to see at what investment level we start getting diminishing returns. It is important to be aware of this level to get out as much as possible and to set the right expectations for each investment made. In this case we identified that it was possible to increase investments by 15% and maintain the accuracy of the model. As you can see a more aggressive approach based on the model would be if they increase the budget by 15% and simultaneously optimize the media mix. We see that the expected increase in sales is 25%. What was the outcome of the MMM project? Based on the output from the modeling our client chose to go with a more conservative recommendation where the total media investment was unchanged but the mix between channels was optimized. The client applied the recommendations from NEPA to their new strategy which helped them to increase their store visits by 17% exceeding our estimated impact of 15%. Thanks for watching if you have any questions you can reach me at NEPA.com and find more information about the great work we do.
Speaker 1: Well thanks Charlotta for coming in and taking us through a real MMM case study. If you found this video useful subscribe where you'll get more content, more videos from me and expert guests about digital marketing best practices, marketing data and analytics and data visualization tips and tricks. All the great content you need to shortcut your way to being a better marketer.
Generate a brief summary highlighting the main points of the transcript.
GenerateGenerate a concise and relevant title for the transcript based on the main themes and content discussed.
GenerateIdentify and highlight the key words or phrases most relevant to the content of the transcript.
GenerateAnalyze the emotional tone of the transcript to determine whether the sentiment is positive, negative, or neutral.
GenerateCreate interactive quizzes based on the content of the transcript to test comprehension or engage users.
GenerateWe’re Ready to Help
Call or Book a Meeting Now