Offline Channels Attribution

Measure the impact of offline campaigns

Posted on January 5, 2020

What are the problems we face today and therefore what questions need to be answered to drive Organic search

  • How do we define and measure ”Organic Demand” or “Demand Expressions”
    1. What are Demand Expressions?
    2. How does it align to our key KPI – App Organic MAUs?
  • How do we understand what was the impact of just Content Pull Factor on Demand Expressions?
  • How do we understand which marketing channels are contributing how much to increasing “Demand Expressions”?
  • How do we adjust budgets across channels (media mix modelling) to maximize Organic Demand?
  • How do we optimize within each channel to understand campaign/ creative level impact and therefore improve performance?
  • How will this work in practice?



Step followed to answer the questions

  • Define Demand Expression
    1. Organic Search intent demonstrated by users for our brand
  • Establish Statistical Relationship between Demand Expression and App Organic MAU’s
    1. What percentage of installs can be explained by the demand expressions
  • Establish the model of attributing activity impact on Demand Expression
    1. What percentage of the demand expressions can be explained by inorganic spends
  • Establish the model to measure cost effectiveness of campaign activity
    1. Getting Spends/users for each offline campaign
  • Establish campaign level analytics to help channels optimize on an ongoing basis
    1. Deep diving into each campaign to increase the performance of each campaign



 

Steps Taken:
Step 1: Establish Statistical Relationship between Demand expression & App Organic MAU’s

  • For installs: Derive total installs that have happened from the system, removing likely fraud installs
  • For search impressions:
    1. Got overall number from keyword planner
    2. Got pattern from the google trends for the brand keyword
    3. Found the multiplier and used that multiplier on the trends number to get the daily level search impressions for all countries
    4. Removed outliers from the data
  • Normalized and smoothen the data
  • Run linear regression model


Output:
step1_offlineattribution As Demand Expressions go up, Organic App MAUs goes up
X% of the variation in Installs is explained by demand expression



 

Step 2: Establish the model of attributing Activity impact on Demand Expression
step2_offlineattribution Establish Statistical Relationship between Demand expression and in-organic spend (Baseline)
Steps Taken:

  • For in-organic spend: Got in-organic daily channel level spend from the in-organic team
  • For Demand Expressions: Used the same demand expression numbers as used in Step1, explained earlier
    1. Got overall number from keyword planner
    2. Got pattern from the google trends for the brand keyword
    3. Found the multiplier and used that multiplier on the trends number to get the daily level search impressions for each countries
    4. Removed outliers from the data
  • Ran daily level linear regression model on the in-organic spend versus demand expression

Output: step2_2_offlineattribution As in-organic spend go up, Demand expression goes up
X% of the variation in Demand Expression is explained



 

Step 3: Establish the model to measure cost effectiveness of campaign activity
Measuring cost effectiveness of campaign activity step3_offlineattribution
Definitions:

  • Current Spend: Total marketing budget spend on running this campaign
  • Demand Expressions Start/End: Total Demand Expression of brand keyword at the start or end of running this campaign
  • In-organic spend Start/End: in-organic spend at the start or end of running this campaign
  • In-organic spend Uplift: Change in InOrganic spend while running this campaign
  • Demand Expressions Uplift (in-organic): Uplift in Demand Expression of brand keyword because of change in in-organic spend
  • Demand Expressions (removing baseline): Uplift in Demand Expression of brand keyword after removing InOrganic Spend impact
  • MAU's Uplift: Uplift in App Organic MAU's by running this campaign
  • Spend/MAU: Spend for getting each app organic MAU
  • Target Spend/MAU: Target spend for getting each app organic MAU
  • New proposed spend: Proposed optimised marketing spend on each campaign to get best allocation of budget

Output:
Optimised marketing budget (New Proposed Spend) based on:
  • Current Spending
  • Increase in demand expression based on current spending
  • Change in InOrganic Spending
  • Impact on Demand Expressions because of change in InOrganic Spending
  • Removing the impact of InOrganic spend on Demand Expressions
  • Increase in MAU’s because of running this campaign
  • Target Cost per MAU for the industry

New Proposed optimized Organic budget planned



 

Final intermediate output might look something like this:
output_offlineattribution

  • 7th Oct: Teaser was launched
  • 9th Oct: TV started campaigning and have a sudden impact on demand impressions
  • 12th Oct: First peak because of ATL push
  • 13th Oct: Road side banner started
  • 14th Oct: Content was launched
  • 16th Oct: TV channel campaigning started again
  • 20th Oct: Radio (radiocity) started running its campaigns
  • 24th Oct: Radio (BigFM) started running its campaigns
  • 27th Oct: Print ads started running its campaigns