Skip to content
English
Contact us
  • There are no suggestions because the search field is empty.

Media Mix Modeling

This article contains the following:

Background

What is Media Mix Modeling (MMM)?

Media Mix Modeling is a top-down attribution method used to assess the profitability of your marketing channels, helping you understand the effectiveness of spend across channels so you can better allocate your advertising budget.

Bottoms up approaches may misattribute the incremental effect of marketing spend on a given channel due to cannibalization (i.e. users who would have come anyway and are attributed to the wrong channel) and/or IP matching problems (i.e. users watch ads across multiple devices and are double-counted). By looking at the data from a higher level, MMM analyses avoid these pitfalls and give you a strategic view of marketing activities. 

Who benefits from MMM analysis?

MMM analysis is valuable for advertisers who are spending significantly across both offline and online channels as it becomes more difficult to isolate the incremental impact on a specific channel.

How does MMM help advertisers?

MMM results can be used to both interpret the past impact—and predict the future impact—of marketing spend on sales. In data-rich settings, these estimated effects can help provide directional suggestions towards more optimal marketing budgets and identification of inefficient channels. Additionally, the estimated effects can be used to corroborate or refute ROAS or CPx attributions from internal multi-touch attribution methodologies.

How is this model constructed? 

Generally, MMMs employ statistical methods on aggregated historical data at the daily or weekly level to estimate the relationship between the incremental effect of advertising budget on sales (or other revenue-tied metrics).

What are the results of the model? 

The main results of the model include the isolated incremental effects by spend channel as well as potential effects of interaction between channels. The isolated incremental effects can be extracted and interpreted as ROAS or CPx, depending on your metric(s) of interest. Estimated interaction effects can be used to show conditional relationships between spend channels (i.e. given a change in one channel what changes should be expected due to spending on another channel?). 

What are the requirements to getting started with an MMM?

Tatari offers media mix modeling for our clients who usually have at least six months (one year preferred) worth of TV campaign data however this is evaluated on a case-by-case basis. Reach out to your Client Services Manager to discuss if MMM is a good option for you.

Getting Started 

To start an MMM analysis, follow the steps below. 

  1. Make a copy of this file template.
  2. Add in all required data (see the requirements below). Note: The template includes dummy data to give you an idea of how to format it. Please overwrite this with your data.
  3. Once it’s filled in, send it to your Client Service Manager for review. If everything looks ok, we’ll provide a report within the next two weeks. 

Data to send Tatari

Please include the following data for at least the past year (2+ years is preferred for better analysis) across all marketing channels.

  • Total sales per day 
  • Daily spend per marketing channel. Include only major spending channels. If you do not see a channel here, please add it as a column at the end of the file template. 
    • AdWords Brand Search
    • AdWords NonBrand Search
    • Other NonBrand Paid Search (e.g. Bing)
    • Other Brand Paid Search (e.g. Bing)
    • Facebook-Prospecting
    • Facebook-Retargeting
    • Pinterest-Retargeting
    • Pinterest-Prospecting
    • GDN (Google Display Network)-Prospecting
    • GDN-Retargeting
    • Affiliate Marketing
    • Clicks from Email Campaigns
    • Other Major Marketing Sources
    • Daily revenue and/or sales data (Optional, but preferred)
      Sales count (and optionally, Total Revenue $) by the same marketing channel with daily breakouts.
    • Visitors, mid-funnel, and lower-funnel data (Optional, but preferred if you are interested in seeing other metrics aside from sales)
      Split by marketing channel and daily breakouts. You may choose to incorporate additional tabs into the template (copying the format of acquisitions_by_day and acquisitions_by_channel_day tab in the CSV template) to reflect visitors, trials, or other metrics that would be of interest to you.
    • Other events that may affect sales patterns (Optional) 
      For example daily promotion, sale, and price adjustment periods, inventory constraints, and competitive forces.

      FAQ

      Can we send weekly spend instead of daily spend?

      Weekly data is ok for use in an MMM analysis. Ultimately, more granular data could lead to more nuanced insights. Likewise, you may receive results aggregated at the weekly level after sending daily data due to delayed response and seasonalities.