Custom Attribution Models: Helping Marketing Specialists, Analysts and CMOs Achieve Zen
Vlad Flaks, CEO OWOX
Custom attribution models are simpler than they might seem. To get the most benefit out of them, an analyst, a marketer, and a CMO should work closely together. Even if all of the above roles are filled by the same person, this article will help you streamline your workflow and achieve better results during all stages.

1. Analyst prepares data
The challenge for analysts is to bring together all the data sources needed for the calculations. Here’s the good news: if transaction data is already sent to Google Analytics and advertising campaigns have all the necessary UTM tags, it will only take about half an hour to establish the settings. You won’t even have to seek assistance from developers. Just use OWOX BI and Google Tag Manager. More details about custom tags in Google Tag Manager can be found in this blogpost. More about combining data can be found here. The very process of bringing data together is described in full detail in our Help Center.

After combining the data, an analyst should set up the attribution model. Marketers can also create new attribution models or change existing ones, but it would be better if the first model was created by an analyst. This will help make sure that the data from the connected sources is accurate. For example, if the probability of progression to the next step of the conversion process is more than 90% or less than 5% (or the ROAS is about 1000%), there’s definitely something wrong with the data.

The result of an analyst’s work is all the data brought together to enable calculations in Google BigQuery, and at least one calculated attribution model.
2. Marketing specialist gets reports and applies the results of calculations
The challenge for marketers is to evaluate advertising campaigns and to draw conclusions about how to reallocate the advertising budget in the most efficient way. The good news for marketers is that in order to see the real value of advertising campaigns, they need only start typing their questions in the OWOX BI Smart Data.

...and see what campaigns were undervalued, or overvalued, in the Last Non-Direct Click attribution model:

This is a valuable report. It demonstrates how campaigns could be re-evaluated if you considered their contribution to the progression through each stage of the funnel, not only to the last one. A marketer can easily manage the results by adding and editing stages within the funnel, combining data from multiple devices, or ignoring free traffic sources in the attribution model.
For example, if the user sessions on multiple devices are combined using UserID in the attribution model, the average length of a conversion path will increase. This happens because a customer can use multiple devices on their journey to a purchase. The credit designated for advertising campaigns will be allocated correspondingly to their channels.

How do you apply the results of the attribution? This might be a genuinely challenging question for many marketing specialists. True, there’s no such option as "consider the contribution of campaigns in Bing Ads" in Google AdWords. Much like advertising agencies, if you work with a few at the same time, they’re pretty unlikely to share notes.
If a marketer does nothing with that, the marketing channels will compete against each other trying to drive conversions. The success of one campaign will always mean that the other has failed. As a result, you’ll only be able to predict one thing about your advertising budget: there won’t be enough money.
You won’t be able to answer such questions as "Why does CPA increase for one channel when I exclude the other one?" or "How do I increase sales without seeing ROAS drop?".
To ensure that your advertising campaigns work as a team towards a common goal, marketing experts should set individual goals for each of them. The goals should be set with their strengths and mutual impact in mind. Display campaigns shouldn’t be evaluated only in terms of the transactions, and Email channels shouldn’t be evaluated just in terms of the number of attracted customers.
The easiest and quickest way to do so, is to adjust the target CPA for each campaign, while taking into account the special correction factor. This factor is calculated as a ratio between the revenue attributed to an advertising campaign in the Funnel Based attribution model, and the revenue attributed in the Last Non-Direct Click attribution model. An example of the correction factor is given in the last column in the table below.

Campaigns highlighted in green have a correction factor of greater than 1. These campaigns are undervalued in the Last Non-Direct Click attribution model. This happens when a campaign more often than others, contributes to the progression on the upper stages of the customer journey, but is followed by another campaign, which gets all the credit for the conversion.
The campaigns highlighted in red have a correction factor of less than 1, these campaigns are overvalued. For example, the Last Non-Direct Click attribution model could assign all credit for the conversion to the Email channel. Meanwhile, campaigns which attracted visitors to a website where they submitted their email address garnered no credit.
With OWOX BI Smart Data, a marketer can determine the efficiency of each advertising campaign and reallocate the advertising budget accordingly.
For example, reallocate the budget towards undervalued campaigns, and reduce the budget for overvalued ones. More details about the shortcomings and advantages of different attribution models can be found in our blogpost.
As a result, by applying the attribution model, which evaluates each touchpoint in the conversion process, marketing specialists can achieve an up to 25% increase in revenue in the project, at the same level of advertising budget:

It’s important to understand that the cost per conversion will increase for some campaigns. However, for the project as a whole, as a result of this well-coordinated teamwork, the cost per conversion will decrease.
Marketing experts can always check the accuracy and the efficiency of attribution models. There are several ways to do this and they have been previously covered in another blogpost.
3. Chief marketing officer controls the efficiency of advertising expenses
What’s the challenge for CMOs, if the data’s already brought together by analysts, and advertising campaigns are managed by marketing specialists? A chief marketing officer plays a principal role in inspiring their team along the way from being Last-Click apprentices to becoming real attribution masters. In the next stages, a CMO motivates their team, by tapping into their strengths and resources while controlling the accuracy of the calculations.
How can you make sure that the ROAS you’ve got provides an accurate evaluation of the performance of your advertising campaigns?
Run a test for two user segments (eg. by country or by region) and compare the results. In segment A, use your old attribution model. In segment B, apply the Funnel Based attribution model. Success would appear as displayed in the chart below:

As a result, your business will be rewarded with transparent analytics, and a more effective way to achieve your sales targets.
FAQ
-
What are the common challenges faced in attribution modeling?
Some common challenges businesses face in attribution modeling include identifying all the touchpoints, measuring the effectiveness of each channel, and the lack of a standardized approach to attribution modeling. -
What are the various types of attribution modeling?
The different types of attribution modeling include first touch, last touch, linear, time decay, position-based, and algorithmic attribution. -
What is attribution modeling?
Attribution modeling is the process of determining the value of each touchpoint a customer interacts with before making a purchase. This process helps businesses to allocate their marketing budgets appropriately.