No matter if you work (live?) in the traditional or digital side of the media world, you are probably familiar with the concept of attribution modeling. In its most basic form, attribution modeling is simply a way of assigning credit to all marketing channels/touches along the path to conversion. We know consumers will often be exposed to brands via multiple channels on multiple devices that can vary heavily, depending the specific consumer segment, the time of day, the day of the week, etc. That is precisely why droves of marking geniuses have tirelessly pursued ways of using ‘big data’ to connect some of these dots and evaluate media performance from a more holistic perspective.
Types of Models Available
One of the trickiest things about using attribution modeling is picking which model to use. There are several different ways of mathematically approaching the question of how much credit to give each touch point in a conversion pathway, and each of these methodologies can produce very different results. For that reason, it is important to know which model you are using when evaluating your data. Some of the most popular attribution models available include:
- Algorithmic – uses historical data to intelligently calculate the relative weight each touch point likely contributed to a conversion event
- First Click – the entire conversion contribution is attributed to the touch that introduced the lead
- Last Click – the entire conversion contribution is attributed to the touch that closed the lead
- Even Distribution – evenly distributes the conversion contribution across all the touch points involved in the conversion (i.e. each touch given credit equal to total contribution/number of touches)
- 30/40/30 – the first and last touches are each given 30% of the contribution credit and the remaining 40% is evenly split across all the intermediary touches
- Custom – uses custom logic based on placement, source, time, etc to assign conversion contributions to each touch point
Often, it is helpful to compare how channel credit differs in each model. Ultimately, however, we generally recommend our client’s use an Algorithmic model since it allows for the data itself to reveal the hidden correlations and insights within marketing efforts.
In addition to having multiple models to choose from, there are several technology platforms available that can provide the computing power required for processing the large data sets used in attribution modeling. The most popular platforms include:
- Google Analytics
- Convertro – recently acquired by AOL
- Adometry – recently acquired by Google
The Google Analytics attribution model is probably the most widely used since it is included with the free web analytics implementation package. However, the standard reports available only provide data at an aggregate level and don’t go down to the transaction-level touches. Furthermore, the lookback window only goes back 90 days, which can be limiting if you are marketing a product or service with a longer sales cycle. While these reports can still provide very insightful information, the other platforms are more specialized and therefore better suited for doing more in-depth attribution analyses.
At Internet Marketing Inc., we have chosen to strategically partner with Convertro to provide attribution analysis and reporting to our clients. We were impressed by their cross-device matching technology which captures over 40K attributes to uniquely identify users across their devices. Additionally, the data can be analyzed using an infinite look back window and is available at the individual touch-point level of granularity for all conversions.
Like all things tracked online, proper tracking code must be implemented to ensure that the data being collected is clean and usable for analysis. In the case of Google Analytics, the implementation is already included in the standard tracking code placed on the site for collecting website data. Convertro, on the other hand, has specialized tags that need to be implemented independently for both the website and tracked media sources. These tags include:
- Tracking code on each page of the website to a) identify users to the website using finger-printing technology and b) capture the referring sources of site sessions
- Tracking code on each conversion page on the website (i.e. ‘Purchase Confirmation’ pages for e-commerce sites or ‘Thank You’ pages for lead gen sites) to capture the conversion value (i.e. revenue, count) as well as other relevant data collected at the time of conversion (i.e. quantity, lead type, etc).
- Tracking code on each directly tracked media source (i.e. ppc, display, email) to properly identify the sources of overt online marketing efforts
In the case of the Convertro implementation, site visitors are fingerprinted when they first encounter a tracking tag and their online marketing histories are captured once they convert online. This then provides a rich set of historical marketing data directly linked to conversions and allows for a deeper investigation of how online marketing channels influence the conversion behaviors of customers.
Analyzing the Data
As mentioned earlier, the standard Google Analytics multi-touch attribution reports will tell you which channels are contributing to customer conversions at the aggregate level. However, if you have access to more granular data, you can also leverage touch-level attribution data to draw even greater insights including:
- How long it takes customers to convert online from their first touch
- How many online touches are involved in customer conversions
- How long it takes to recognize revenue from past online touches
- Which channels that are introducing, closing, or independently influencing conversions
- The types of touches each channel contributes to conversions
Furthermore, the detailed attribution data can be integrated with other data sources (i.e. CRM data, Google Analytics, etc) for an even richer media analysis. Examples of data sources that we have successfully integrated included:
- Media spend (i.e. PPC, Display, Social)- to determine the return on ad spend (ROAS) from paid media channels
- Customer type (i.e. new vs returning) – to evaluate how the channels influence conversions of different customer types
- Geographic data (i.e. sales region) – to determine the impact of targeted marketing efforts on driving conversions in specifics regions
- Google Analytics e-Commerece – to further segment transactions at a product level
Integration with Media Optimization Decision-Making
In my next post, I will go into detail on how we have been able to leverage Convertro’s Algorithmic attribution model to make strategy optimization recommendations for our clients. Stay tuned for Part 2!