Part
01
of one
Part
01
Please research Attribution Modelling with Google Analytics
Hello! Thanks for your question about Attribution Modeling with Google Analytics. The short version is we found and summarized relevant sources that explain the importance of attribution modeling for business. Attribution modeling is a subset of Marketing Mixed Modeling (MMM) which helps companies determine which marketing channels drive their sales. However, as digital marketing has popularized, attribution modeling has also become an essential marketing tool, sometimes prioritized over MMM. Different attribution models have been developed and choosing the best one for a company requires marketers to focus on several variables including the type of business, the sales cycle, and the customer journey. Below you will find a deep dive of my findings.
OVERVIEW
We focused our research on the importance of attribution modeling applications for business. Therefore, we included some background information on MMM and how it relates to attribution modeling. We also included some basic principles of the models built in Google Analytics, as well as some customized models. Here, we have summarized the most relevant variables to consider when choosing a model. Finally, we listed a short selection of business cases. Additionally, each of the sections lists specific sources were more information is available.
WHAT IS ATTRIBUTION MODELING?
Attribution modeling is a process through which marketers can assign values to help them understand the return on investment (ROI) of specific digital marketing channels. Therefore, attribution modeling allows marketers to determine which are the most valuable digital marketing channels for a specific campaign or product. Nowadays, attribution modeling refers mostly to digital marketing (i.e. online marketing campaigns), however, similar tools were previously developed for non-digital marketing campaigns. So, for example, a company could estimate how radio ads, TV ads, and printed ads affected the overall results of a single campaign. This is known as Marketing Mixed Modeling (MMM), and attribution modeling is cited as a subset of this larger concept.
MARKETING MIXED MODELING
MMM refers to the set of statistical tools that help marketers "measure and forecast the impact of various marketing activities on sales and ROI. It is used to measure the overall marketing effectiveness and determine optimal ad spend among various marketing channels." Often, MMM will consider regression analysis for market data, economic data, target audience, industry data, competition data, and how it relates to the product, price, placement and promotion (commonly referred to as the 4Ps of marketing) of an advertising campaign. As e-commerce has increased its reach and overall importance in sales, attribution modeling has been developed to measure the effectiveness of digital ads and spending by understanding the online customer purchase journey. Therefore, it considers online and digital advertising only, making it a subset of MMM.
MMM may occasionally be conceived as independent from attribution modeling because some companies do not prioritize the offline features of MMM since their business happens exclusively online. Online-only businesses will mostly focus on developing attribution models, while offline-only businesses will focus on statistical tools that measure the offline market. Meanwhile, businesses that have online and offline points of sale will have to include both, online and offline measurement tools, within their MMM.
You can find more information on MMM in the following sources:
The article refers to the MMM and how it can be strengthened with attribution models for understanding the customer purchase journey through "true multichannel analytics...which takes both, online and offline touchpoints into account."
Nielsen is one of the leading marketing analytics companies. This web page explains the type of data used to provide an MMM solution for clients. They use statistical models that help their clients understand marketing spend and business performance.
The web page provides a brief summary of MMM and the information they might provide considering a diversity of variables including their "contribution to sales-volume, effectiveness (volume generated by each unit of effort), efficiency (sales volume generated divided by cost) and sometimes return on investment (this depends on the client's willingness to share internal P/L information)."
ATTRIBUTION MODELS
According to Google Analytics Help web page, "An attribution model is the rule or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths." So each model assigns conversion values considering different variables and rules, mostly by focusing on where throughout the customer journey the customer converted. Google Analytics has 7 models but it also allows users to customize a model that best fits their needs. The basic principles of these 7 models are:
1. Last Interaction: 100% of credit is attributed to the final touchpoint. This is one of the oldest models since the last touchpoint was the easiest to track.
2. Last Non-Direct Click: 100% of credit is attributed to the touchpoint that preceded the direct click.
4. First Interaction: 100% of credit goes to the first touchpoint of the conversion path. Note that the first touchpoint has a default 30 day lookback period.
5. Linear: It is a multichannel model that assigns an equal percentage value to each of the channels it considers. For example, for a four-channel marketing campaign, a 25% credit would be assigned to each channel.
6. Time Decay: It is also a multichannel model but instead of assigning an equal credit to the channels, it assigns decreasing values to each touchpoint, starting by crediting the last touchpoint with the highest value and decreasing assigned credit for each previous touchpoint. The decrease is calculated based on time passed (minutes, hours, days) between each touchpoint.
7. Position based: 40% of credit is awarded to the first and the last touchpoints. The remaining 20% is equally divided among the rest of the touchpoints.
Each model has advantages and disadvantages which must be considered in order to choose the one that better adjusts to the individual business. All non-multichannel models (1-4) are earlier models which were initially limited by technology. That is, it was technically complex for an analytics tool to accurately trace the conversion path and weigh multiple channels. However, current digital marketing campaigns are more often than not multichannel, using at least social media, paid advertising, and referrals. Because of this, more sophisticated marketers use customized multichannel models based on more recent models (5-7).
For more in-depth information on each Google Analytics model and their advantages and disadvantages we recommend the following sources:
This is a blog post that analyzes each model and provides information on their advantages and disadvantages. Additionally, it suggests a customized model and key questions to develop a model that better fits a company:
A. What type of user behavior do you value?
B. Is there an optimal conversion window you are solving for?
C. What does the repeat purchase behavior look like historically?
D. Are there any micro-conversions defined with engagement type goals, tied to the economic value?
E. Are offline conversions being sent back into GA using Universal Analytics?
The web page includes a brief guide that also has some useful introductory suggestions like knowing how to configure goals, using campaign tags, and having the Google AdWords and Google Analytics tools linked for better measurements. It also has a brief video overview of the models.
This source displays a flowchart that helps businesses whether they should opt for multichannel attribution models or not. The variables that should be considered for making this decision are:
A. Whether the business is B2B or B2C. B2C businesses without a sales team may benefit from using a single-touch attribution model.
B. Number of marketing channels. The recommendation is to use multichannel models if a company is using 4 or more channels.
C. Marketing and sales cycles. Depending on the type of product, the sale cycle may take just one day or may extend to a few months. The general rule is if the sales cycle is longer, then a multichannel model is a better option, while single-channel models may be useful for shorter sale cycles.
D. Marketing expenditure. Higher expenditures often mean more channels. Additionally, larger budgets are better served by analyzing ROI more closely, with tools available in multichannel models.
OTHER ATTRIBUTION MODELS
Bizible's article on attribution models includes two additional models that are not built in Google Analytics:
1. The W-shaped model which assigns a 30% value to the three touchpoints preceding the conversion touchpoint. The remaining 10% is split among the remaining touchpoints.
2. The Full-path attribution model includes the last four touch points, including the conversion touch point. A 22.5% credit is awarded to each, while a 10% credit is divided between the remaining touchpoints.
BUSINESS CASES
We were able to find useful business cases that use different attribution models. One of the most useful sources we found refers to an Adidas campaign. It is an older source, however, it explains how a large retail company, that has online and offline points of sales, can use digital marketing to increase sales.
The focus of their strategy was to bring customers searching for Adidas on mobile phones to actually visit a brick-and-mortar store. Using their analytics model, they were able to determine that 1 in 5 customers searching for Adidas on their cell phones would end up visiting the store. Furthermore, around 13% of store visitors usually make a purchase with an average value of $71 USD. However, customers redirected from a mobile store locator are more likely to buy and to make a larger purchase (around $80 USD). By developing an attribution model that allowed Adidas to understand which was the customer conversion path that drove higher sales, they were able to redirect their marketing budget to channels that provided better results.
Another case study is the Australian store Rebel. This store used Google's anonymized location tools to find how many users that visited the store had previously clicked on a mobile ad for the store. With this information, Rebel was able to identify which product categories drove more traffic, leading to a better-targeted ad campaign. They then experimented a 44% increase in store visits and a "15% increase in category share on key sports categories."
We found other case studies that do not explain the attribution models as thoroughly. However, these are also sources that can help explain the importance of selecting a model that fits with the company's goals and marketing strategy:
The source includes a video that explains the success of the Baby Supermall store. They had a large marketing budget that was not efficient. With a better attribution model, they were able to identify their sales cycle was really long, as expecting families usually take months before actually purchasing products. They then focused on understanding the customer journey better and target ads to the audience more likely to visit the store in the first place.
2. Marketing Attribution Models: 43 Experts Explain How to Choose The Right One (and Mistakes to Avoid)
This is a long blog post that briefly explains what are the specific variables experts notice in order to choose an attribution model. Each case is very shortly summarized and includes only key information that each expert emphasized in order to decide what type of model was best for the business. Some experts also mentioned what mistakes to avoid while choosing or analyzing the model.
To wrap it up, attribution modeling is a subset of modern MMM. It helps businesses understand how their digital marketing strategy is helping them reach their goals and provides information for optimizing advertising and targeting. Choosing the correct model for the business is one of the key challenges and businesses should focus on the customer journey, the sales cycle and the marketing channels for applying a model that works.
Thanks for using Wonder! Please let us know if we can help with anything else!