Insight into Media Efficiency

Attribution Modeling and Reporting

Abstract

With advertising budgets in the hundreds of millions, large enterprise retailers must go beyond accurately tracking the results of their spending and implement the systems and processes that empower decision makers to better allocate funds to the marketing channels delivering the best return.

Even with a bountiful range of available tools designed to track, analyze, and recommend, impactful course-corrections are still problematic for many marketers operating without an accurate assessment of the efficacy of their spending and no clear understanding of the media combinations that yield results larger than the sum of their parts.

This use case details the steps 89 Degrees took as part of an Attribution Modeling and Reporting project to help a multi-billion dollar retailer who had recently completed a strategic overhaul of its marketing mix, but still lacked insight into:

  • Media effectiveness over time or by designated market
    area (DMA)
  • Media performance as a whole or by tactic; e.g. TV vs
    Radio vs Direct Mail vs Paid Search
  • Media impact on existing vs new customers

With insights from attribution models, clients can carefully adjust ongoing spending and realize measurable performance gains.

Related Content

Visual analytics on map
Person watching tv

Project Goals

89 Degrees’ approach to attribution modeling relies on the most granular level of data. This use of individual data provides greater insight into the effect of each media type and gives better information about the interaction of media. This approach enables our client to:

  • Accurately report on the products and services driven by and attributable to marketing across all channels
  • Gain insight into the most and least effective marketing mixes influencing customer behavior
  • Inform the decision-making process for allocating budget and resources that increase return on marketing investment

Project Details

Stakeholders, Influencers, and Systems

The primary stakeholders involved were the Vice President of Marketing, Director of Customer Intelligence, and the team responsible for media buying. Non-users, but key project influencers, also included the CEO and CMO.

The project focused on the development and implementation of attribution models with output rendered in SAS Visual Analytics. External systems would be involved as well with third parties providing data for television and radio ad buys, paid search, and display reporting. Our client would provide its own Salesforce email data and CRM database, while 89 Degrees handled the inclusion of direct mail reporting.

Requirements

In building the models, the project team used data from the past two years for all media spend and advertising impressions. This data could be reported by week for each DMA, ZIP code, or individual, using in all cases the most granular view available. Media sub-tactics – including paid search and display, alternate languages, national versus local TV buys – were used and proved to be valuable.

SAS visual analytics data model

Process

To ensure success, the 89 Degrees project team focused on individual level attribution, evaluating 104 weeks of data for each customer or prospect within our client’s target market, resulting in more than 5 billion distinct observations. The team evaluated prospects and customers separately, taking steps to tease out the percent of organic sales not attributable to media spend. 89 Degrees started by acquiring all data from the various vendors, cleaning and manipulating data to generate the needed information. With much of the data provided in non-standardized Excel feeds, 89 Degrees created a repeatable, automated format to speed up processing and prevent errors.

Next the team created Decay Curves to determine how long media impressions influenced the target audience. The Decay Curves provided a specific value of an impression or Television Rating Point (TRP) over time and helped highlight for each channel the point of diminishing returns. 89 Degrees aggregated the media data to the lowest level possible along with transaction history to the customer week level. This enabled the project team to create one data set for existing customers and another for new customers.

The team then used logistic regression techniques to develop attribution models for existing and new customers. The models describe the relationship between media spend and customer sales, providing insight into whether the media had reached saturation levels – meaning additional impressions would deliver little to no lift – or if there was an opportunity to test increased impressions. Finally, the team applied attribution scoring at the transaction level, historically and on an ongoing basis. This allowed each transaction to be attributed differently to the various media tactics.

 

Results

To complete the project, 89 Degrees delivered a dashboard for all required reporting objectives via SAS Visual Analytics, including:

  • Measure media KPIs over time
  • Track the effects of changes in media buying strategy
  • Drill down to individual or groups of DMAs
  • Assess media impact on key objectives

The findings of the project allowed our client to redeploy its media and increase ROI. They were able to understand the point of diminishing returns for different media types (as shown on the charts above), as well as the relative benefit of an incremental dollar going to one media source versus another.

The results also showed how media in combination did or did not create incremental sales at different spend levels. These insights and the ongoing review of media efficiency in the SAS Visual Analytics reports allows our client to continue to improve and refine its media strategy, and to make better decisions in instances in which the media budget is either decreased or increased for a period of time.

Software

SAS Visual Analytics provided the ideal mechanism to track ongoing performance and media attribution which allowed our client to identify opportunities to improve the media mix in challenging markets. They could also more easily identify what wasn’t working compared to other DMA’s, and set up tests – tracking ROI by media channel over time – to see how changes in strategy would play out.

Results icon

Related Blog Posts