MEDIA MIX OPTIMIZATION
FOR MEDIA BUYERS

The objective is to predict the revenue of a customer based on the investments trough various medias.

The media considered are:

  • Investments in TV time
  • Investments in Radio time
  • Investments in Billboards display
  • Investments on the Point-Of-Sales
  • Online Investments in SEO: buying keywords
  • Online Investments: buying displays inside banners
  • Online Investments: buying clicks and paying them “per click”
  • Online Investments: buying clicks and paying them “per conversion”

These kind of predictive models also contains some extrageneous variables such as:

  • Temperature
  • Humidity
  • How many schooldays in the week/month?
  • GDP of the country
  • % of unemployment in the country

Such project is generally composed of 2 phases that are iterated many times: creating the predictive model and assessing the predictive model using many charts. Once you have a predictive model, you can use it to make different simulations. Each simulation test for a different ways of investing your wallet inside different media’s. At the end, using the simulator, you are able to find the investments that generates the highest ROI for you.

The Cofidis case

Let’s give an example on the Codifis case. The exact objective here is to predict the NLR: Number of Loan Requests.

Data transformation

Here is a screenshot of the data transformation required to create the predictive models with Anatella.

graphic1

Automatic reporting

Once you have a model, Anatela automatically create a powerpoint report based on the results of the predictive models.

graphic2

Some results

Here are some results generated by Anatella based on the above data-transformation. The second chart explains the effect of each group of variables inside the model. The variables are grouped into 3 themes: the external drivers such as temperature, humidity, etc (in green); the investments in ads from competitors (in grey), your own investments in ads (in red). The blue layer is the base: it’s the recurrent quantity of sales if there is is no investments in ads from anybody.

Real NLR vs Predictions

graphic3

Advertisement effects

graphic4

Deep insights

The first graph hereunder gives you have an idea on the contribution of each group/theme of variables. You can also “zoom-in” and display the effect of the variables that are only inside the theme “Variables describing investments in ads from competitors”. You can see that, amongst your own investments, the investment that is working the best is “Promo”: in the case of Cofidis, the “Promo” are actually the investments to have “cofidis” banner during sport events: indeed Cofidis is sponsoring The “Codifis Cup”: a very popular football cup in Belgium and The “Cofidis Team” the famous bike team.

Global advertisement effects

graphic5

Own advertisement effects

graphic6

MEDIA CUSTOMERS

cofidis

Cofidis

lotterienationale

Lotterie Nationale

brittishamericantobacco

Brittish American Tobacco

havasmedia

Havas Media

oasis

Oasis

ipgmediabrands

IPG Mediabrands