Developing Enterprise Response Functions

A brief review of approaches

 

The best approach for developing enterprise response functions is to collect activity data and sales results and use statistical analysis to quantify the impact of each activity.  Essentially, we are looking for proportional increases (decreases) in sales whenever a specific promotion is increased (decreased). This requires collecting weekly/ monthly sales data, as well as weekly/ monthly promotional activity data at the most granular level possible.

 

Ideally, we would sales data source at the customer level.  For example, IRI and AC Nielsen collect data on in-store sales for several consumer categories.  In some cases, this data would need to be supplemented with direct data feeds from some large retailers.  If a company does not have access to demand data, then we can leverage inventory/ ship-to data with some reasonable assumptions on how this translates to demand.

 

On the activity side, the granularity is often a function of the marketing channel/audience.  For example, emails would be linked to individual customers and/or ZIPs, digital advertising (display and on-line search) to consumers would be collected at the ZIP level, and TV / Radio would be collected by Designated Marketing Area (DMA).  B2B activities would be collected at the HQ (parent) or store (child) levels as available.  We would account for the different levels at which promotions are run and simultaneously account for their impacts in the statistical models.  In those cases where a sequence of promotions happens before the sales is made, we can use our understanding of these sequences along with pathways analytics to model these sequences.

 

Note that this data is only required for a representative sample of customers and/or geographies.  A company does not need to invest in acquiring and integrating this data for every customer.

 

When sales and activity data are not available, then we still have a few options:

 

  • develop a robust set of impact estimates through in-market A/B test (pilots)
  • leverage existing forecasts/ plans that include different promotional spends (e.g., + 10% and -10%) to derive an aggregate response curve
  • conduct an internal trade-off session where brand team members must choose between different tactics under artificially low budgets

 

The latter two methods quantify the company’s thinking on promotional impact and can be used in proof of concept models.