I developed these attribution solution models the afternoon following a long meeting where I was trying to explain to my coworkers at the advertising agency why data integrity is the most important task in advertising. This wasn’t the first time I was trying to speak science to a group of creative experts. As you may well imagine, my opportunities in an advertising agency to speak to coworkers about data in general are infrequent. One of the important elements in advertising that I have learned from my coworkers is that advertisements need to be believable. In digital advertising we preach – relevancy in advertising. The brand message must also be passionate. These are all key components for a data scientist, and I think maybe even more emphasized when working at an advertising agency.
Attribution Solution Models
Believable, Relevant, and Passionate
The attribution problem has two models. The first model is when the client business is a franchise and the second model is for non-franchise businesses. At the top of both models is the consumer purchase. This works for all cases including website properties that are not ecommerce enabled. The end result the client is seeking for the return-on-advertising-spend (ROAS) is a sale to the consumer. All of the relevant information about this sale is then uploaded to the customer relationship management platform such as Salesforce, Zoho, Oracle. Netsuite, Hubspot, etc. Some of this information is used by the marketing team and the advertising team to develop advertising strategy in the feedback loop. From the feedback loop the advertising methods for the offline and online channels is rolled out. The advertising directs the consumer’s to the website.
Once the consumers are at the client website the two models separate. For the franchise the consumer is typically introduced to the products, services, and features and then directed to a particular franchisee. For example many banks have a corporate website that helps to guide the consumer to the closest branch office. Each branch office may have their own website or a subdomain of the corporate site. The national basketball association (NBA) is an example of this method where each of the individual teams has a subdomain of the primary, nba.com domain. The non-franchise attribution solution models skip this step. From there moving forward, the two models become identical once again.
The attribution problem model now identifies the first problem. “Lack of data reliability.” This is usually the case when working with Google and Adobe Analytics as the primary data management platform (DMP). Not because Google Analytics is a bad platform, it is an excellent platform. The cause is usually that the platform rarely gets configured to meet their business strategy. There are over 100 customization settings in Google Analytics that must be configured to correctly track and to correctly report website information. The customization also affect advertising performance, user experience and behaviors, and conversion metrics. Usually, the business installs the tracking code on the web property and that is the extent of the platform development. In short, the reliability of the data coming from the website ranges between 25% and 30%.
When a chief marketing officer, digital marketing professionals, and other executives saying something similar to: “we’ve tried optimization techniques like A/B testing, SEO and SEM and other methods and none of those ever increase the bottom line revenue.” It’s a “Reliability Issue” red-flag.
Moving up in the franchise model to the franchisee website typically leads to a dead end for data collection and data analysis. At best – it’s more of the same reliability issue. Some franchisees have autonomy for their advertising strategy, website tracking, performance optimizations, etc. and the only information they feedback to the franchise are completed sales. The attribution solution models represent this lack of transparency with a large question mark.
The In-store, Brick and Mortar Attribution Model
On the far left side of each attribution solution models are the brick and mortar variable. This is usually called the in-store visit. The consumer has seen the advertisement(s) and visited the website and then made a purchase at a store location. Alternatively, they saw the advertising and did not visit the website but did go to a store and purchase. The later posing the nearly impossible to solve attribution problem. As such, I’ve left that consumer path out of the attribution solution models. While there are some sketchy (at best) methods for tracking these consumers it is usually in violation of privacy policies. Many of these methods do not follow good advertising practice. Besides the possible legal implications, I leave it off of my attribution problem models because my attribution solution models (ASM) provides a fairly precise method for accurately attributing to that consumer segment.
An Attribution Solution Model
The attribution solutions models is a process I’ve developed over the last 12 years while working with many different business verticals. Over this time I’ve primarily used Google Analytics and Adobe Analytics for data management platforms. The solution begins with the development of the attribution solutioning model guidelines. The Attribution Solution Model Guidelines (ASMG) is a comprehensive technical document. The content of this document is intended for a Tag Management Developer and a Google Analytics Developer. The guidelines provides specific information they will require to accomplish the attribution solution shown in the article images. There are four sections to the ASMG .
- The website analysis or Site Tracking Assessment
- A list of specific request for the client and the website developer.
- The tag management developer guidelines
- The Google Analytics developer guidelines
If your company or your agency would like more information on the attribution models discussed above please contact me.
A primer for attribution modeling