Marketing Mix Modeling for Scenario Planning - Recast

Quoted on this wiki

Every place a page here uses this source, in the order the words come in it.

1. A value greater than 1 implies orders exceed demand, a value of less than 1 implies demand exceeds orders, while a value of exactly 1 implies demand and order are in equilibrium. “While advanced time series forecasting techniques manage to reduce the bullwhip effect, errors persist.” For instance, consider a classic supply chain consisting of a retailer, wholesaler, and a manufacturer. The retailer may respond to a sudden demand spike by placing a larger order with the wholesaler. It calculates the quantum of this order by inputting demand signals into a time series-based forecasting model to arrive at a number. At this point, a small forecasting error is likely to creep in due to the inaccuracy of time series forecasting. To avoid being caught out of stock, the retailer may even order a little over this number as a buffer.

   In Funnel scenario modeling

2. Scenario Optimization Using Solver “What we’re interested in next is finding out how we should allocate our marketing budget to each of our marketing channels to maximize this predicted sales value.” Here’s how we can do it in Excel.

   In Channel mix allocation

3. This defines the maximum marketing budget we have at our disposal. Solver will show us the most optimal marketing mix based on this value. If we don’t enter this constraint, there is no limit to what the maximum amount we can spend on each marketing channel, and thus the optimal value would tend to be as high as is numerically possible. “We’ve also added a new column Total Spend which is the sum of the marketing spend across the three channels, and which must be kept below the total marketing budget.” Step 3: Launch Solver

   In Channel mix allocation

4. Solver then presents to us the most optimal marketing mix allocation which would maximize our predicted sales under the given conditions. “We’ve built a simple scenario planning model using MMM and optimized it for maximum sales.” It bears repeating once again that this is a very simple model. We could add several layers of complexity to make it more accurate and more powerful so that it becomes capable of delivering forecasts even when working with data the model hasn’t seen. For instance, we could implement neural networks within this Excel model, or execute it using R, or combine the two to have a neural network based forecasting model with MMM in R, and so on.

   In Channel mix allocation

Quotes are verbatim and belong to their publishers. Where a quote has been translated, the original wording is kept below it.