Marketing Optimization Explained

Paul Maiste Optimization

Marketing Optimization refers to understanding and analyzing your data to make better marketing decisions across channels. Doing this properly can result in:

  • Improved ROI
  • Increased customer response rate
  • Increased annual revenue
  • Increased website conversions
  • Decreased cost per order

The challenge is to determine how to optimize your data for improved results. The amount and significance of information organizations have can be overwhelming. Think hard and define the true question you’re trying to answer. Then find the data that will answer that question.

To put this into context, consider the following example:

We had a client who wanted to focus their media efforts to maximize campaign responses. The data we had consisted of:

  • Multiple Channels
    • Newspaper inserts
    • Shared mail
  • Large Potential Audience
    • Over 30,000 zip codes

Based on the data available to us, we considered which combination of zip codes should be targeted in order to maximize campaign responses.

The client’s budget for this allowed targeting for 2,000 of the 30,000 zip codes in the campaign. But which of these thousands of zip codes should be targeted? There are a lot of possibilities. Often, businesses simplify problems like this by introducing rules that reduce options down to a relatively small number of possibilities. In some instances, a business might only consider certain geographies or segments, or simply re-target well-performing zip codes from previous campaigns.

The reality is, it is unlikely that you will luckily hit upon the truly best solution, or even the top best solutions, by simplifying business rules. 

That’s where Constrained Optimization comes in. Constrained optimization techniques help identify better-performing targeting, like choosing the zip codes that will have the best results for the organization, instead of just the most convenient choice. These techniques iteratively reduce the possible set of “best” solutions down intelligently while accounting for appropriate constraints.

Below, you’ll find steps to leveraging constrained optimization to solve problems similar to the one above.

Step One: Decide what you want to optimize. This is the ‘Objective Function’ and will drive the rest of your work. Often, people consider their Key Performance Indicators, as these numbers guide how decisions are made at their organization. This can include:

  • Average Order Amount
  • Response Rate
  • Cost per Click/Order/Conversion
  • Churn
  • Spend
  • ROI
  • Lifetime Value

For our example, the KPI we focused on in the example above was the number of respondents, which we wanted to maximize.

Step Two: Identify which levers and constraints exist in the universe of the campaign. These levers are crucial to consider. This could include marketing levers as well as constraints.

Marketing Levers:

  • Audience Segment
  • Offer
  • Channel
  • Creative
  • Message
  • Contacts
  • Budget

Constraints:

  • Budget
  • Production
  • Business Requirements

For our example, the main lever was the zip code. Future improvements may overlay Creative or Offer on top of zip codes.  Many general business constraints were involved as well, including budget. 

Step Three: Collect and organize the data to support the optimization decision. To do this, we need supporting data related to all of our marketing levers. Some examples include:

  • Historical results
  • Forecasts
  • Predictive Model Outputs
  • Expected Trends
  • Costs

This information allows organizations to consider their levers and constraints in the appropriate context.

For our example, key supporting data included: zip code targeting history, circulation, predictive model scores that estimated future response rate, and media costs.

Step Four: Select the decision variable. The process will assess all possible values of the decision variable to find the best solution. Examples include target, price, and budget.

For our example, the decision variable was whether to target, or not target, each zip code.  It was a 0/1 variable, to be decided for each of 30,000 codes.

The primary goal is to positively impact the KPI’s. We can directly compare before and after results, conduct back-tests, or run A/B in-market tests.

For our example, our work resulted in an 87% response rate, an increased revenue of $1.8mm, and a decrease in planning time of 90%.

This process and these tools have the potential to unlock significant business growth. The standard approach to decision making in the marketing context may seem less risky at face-value, but the reality is organizations are missing out on significant opportunities.

If you’re interested in sharing this information with someone, we also have it in presentation format. Check out the Optimization Explained, Simply presentation below.

 

Want to learn more about Lityx? Request a demo to chat with our about how our software and services can help your business improve its marketing efforts.