12 Reasons to Select OWOX BI ML Funnel-based Attribution

The goal of applying attribution modeling in marketing is to evaluate the effectiveness of advertising channels and campaigns and identify which paid traffic channels lead to conversions. Knowing this, you can:

• redistribute your advertising budget to effective channels,
• stop spending money on inefficient channels,
• and, eventually, get more conversions and more revenue.

Your success in executing the sales or marketing plan and growing the business depends on the quality and validity of your attribution model.

The problem is that most conversions happen as a result of more than one session: before buying, the user visits the site several times.

Therefore, to distribute the value of revenue or profit from a specific order, you need to evaluate each session, not just the last one. Most of the standard attribution models don’t take into account all of the user actions before the conversion event happens and give all its value to one channel in the chain. Or they evaluate channels according to a conditional rule and not according to the business model.

At OWOX, we’ve created our own machine-learning Full-Funnel Attribution model that takes into account all user actions online and offline plus real revenue data from your CRM and shows the mutual influence of channels on conversions and user promotion through the funnel.

This article describes what you can do with OWOX BI.

Note: This post was originally published in June 2020 and was completely updated in early 2024 for accuracy and comprehensiveness.

To begin with, let’s refresh your theoretical knowledge.

Attribution modelin is the distribution of value from conversion between channels that advance the user through the funnel. This conversion could be a purchase, a sign-up, a download, or any other valuable action for a business. It helps answer to what extent each channel influenced the user’s promotion through the funnel and the revenue you received in the end.

An attribution model defines the logic of distributing this value — for example, to advertising channels and campaigns that helped attract users. Ideally, your attribution model should be reliable (taking into account the objective contribution of each effort) and transparent (making it possible to understand and verify the results of the calculation).

Attribution modeling is essential for marketers, advertisers, and at the end of the day - businesses who invest in various digital marketing channels. They need it to understand which parts of their marketing efforts contribute most effectively to conversions and sales.

This insight helps them allocate their budgets more efficiently, refine their marketing strategies, and improve the overall ROI of their marketing activities. Attribution clarifies how different marketing channels and touchpoints influence customer behavior, enabling data-driven decision-making.

Learn more in-depth about what marketing attribution is, how it helps companies, and what errors and difficulties marketers and analysts typically encounter when working with attribution.

1. First Click: All conversion value is attributed to the first source that led the user down the path to conversion.
2. Last Click: All value goes to the last channel the user came into contact with before the conversion.
3. Last Non-Direct Click: All value is assigned to the last channel in the chain. If it’s direct traffic, the value is attributed to the previous source.
4. Linear: The value is distributed equally among all sources in the chain.
5. Time Decay: The value is distributed among channels on an incremental basis.
6. Position Based: The channel that introduced the user to the brand and the one that closed the deal receives 40% of the value each. The remaining 20% is divided equally between all channels in the middle of the funnel.

In GA4 (Google Analytics 4), the attribution models have evolved compared to the standard models in Universal Analytics. Plus, Google deprecated most of the standard attribution models.

While the fundamental principles remain similar, there are some key differences in GA4's approach to attribution.

First and foremost, all GA4 attribution models exclude direct visits from receiving attribution credit, unless the path to conversion consists entirely of direct visit(s) - which is sometimes still the case due to business logic or cookie restrictions.

Here's what remains in GA4:

1. Paid and organic last click - Ignores direct traffic and attributes 100% of the conversion value to the last channel that the customer clicked through (or engaged view through for YouTube) before converting.
2. Google paid channels last click - Attributes 100% of the conversion value to the last Google Ads channel that the customer clicked through before converting. If there is no Google Ads click in the path
3. First Touch is not available in GA4.
4. Linear is not available in GA4.
5. Time Decay is not available in GA4.
6. Position Based or U-shaped is not available in GA4.

GA4 introduces more flexibility and data-driven models compared to Universal Analytics. One significant addition is the Data-Driven Attribution model, which uses machine learning to assign credit to various touch points based on how they impact the probability of conversion. This model is more dynamic and adapts to the unique patterns of each business.

Overall, while the names and implementations might vary slightly, the underlying principles of these attribution models in GA4 are closely aligned with those in Universal Analytics, allowing marketers to make informed decisions based on their specific conversion paths.

Different attribution models are available in different advertising services and analytical systems. Most services use the last non-direct click model; some provide post-view, cross-device, or other models. But these models can’t be compared across services: Facebook measures advertising in its own way, while Google Ads has a different approach. As a result, it’s impossible to get an overall picture of all your advertising tools.

Learn the basic principles, pros, and cons of the best-known attribution models to choose the model that best suits your business.

Here is the list of the typical scenarios, when using standard models is limiting the business needs and goals:

1. Multiple Sessions Before Conversion: If most orders on your website are made over multiple sessions, standard models like 'Last Click' might not accurately reflect the entire customer journey. They may overlook the impact of initial interactions that were crucial in the early stages.
2. Large Number of Advertising Sources: When dealing with a multitude of advertising channels, standard models may not effectively capture the synergistic effects or the individual contribution of each channel to the conversion process.
3. Evaluating Mutual Influence of Channels: Standard models often fail to assess how different channels influence each other. Understanding the interplay between channels is crucial for optimizing multi-channel marketing strategies.
4. Understanding Channel Bundling: Standard models might not provide the nuanced insights needed if you want to analyze how different advertising channels work together. They typically attribute credit to channels in isolation, not in combination.
5. Complex and Inconsistent Sales Funnels: In scenarios where sales funnels are complicated and non-linear, standard attribution models might not accurately trace the customer's path to conversion, leading to incomplete or misleading insights.
6. External Factors Influence: These models often don’t consider external factors like market trends, competitive actions, or seasonality, which can significantly impact customer decisions and behavior.

In such complex scenarios, relying solely on standard attribution models could lead to suboptimal decision-making.

More advanced models use machine learning to analyze the contribution of each touchpoint, or custom algorithmic attribution models tailored to specific business needs and customer journeys might be more appropriate.

To assess the mutual influence of all sources, you need to combine data from different advertising services, website tracking systems such as GA, and take the revenue from your CRM, and apply attribution models: Markov Chains, Shapley value, Conversion Lift or customized algorithms.

But these models also have their limitations: a minimum amount of data required for calculations, the inability to consider post-view conversions or combine data from your CRM, hidden calculation logic, complex and expensive implementation, etc.

Data-driven attribution from Google seems like a good solution, however, it is a black box and not surprisingly appreciates Google Ads over the other channels. As a result, the numbers in marketing reports don't add up and marketers cannot trust it…

Machine Learning Full-Funnel (Funnel-Based) Attribution Model allows you to avoid all of these limitations.

The OWOX BI attribution model assesses the effectiveness of your advertising campaigns, taking into account the contribution of each channel to the customer’s promotion through the conversion funnel.

With this model, you can:

1. correctly allocate your advertising budget, taking into account the real contribution of channels to conversions and their mutual influence
2. reduce Cash Reserve Ratio (CRR) while increasing revenue
3. increase the number of new customers while saving your budget
4. increase the number of target users reached by media campaigns while saving your budget.

Funnel-Based Attribution in OWOX BI is based on Markov chains and machine learning. The Markov chain is a sequence of events in which each subsequent event depends on the previous. Attribution based on Markov chains uses a probabilistic model that calculates the probability of transitions between steps of the funnel, allowing you to evaluate the mutual influence of steps on conversions and find out which steps are the most significant.

To calculate probabilities, all steps of the conversion funnel specified in the settings of the OWOX BI attribution model, plus the site login step, are presented as outcomes in the Markov chain. After that, the probability of a transition between these outcomes is calculated.

The graphic above is a simplified example for ease of understanding. In real cases, there can be even more transactions.

Read more about calculating the OWOX BI Funnel Based Attribution model in our Help Center.

1. Collect raw user behavior data from your website in Google BigQuery. To do this, you can use OWOX BI Streaming or [GA4] BigQuery Export. (Don’t worry, we have customizable no-code templates to collect GA4 events into sessions the way you want).
2. Import cost data to Google BigQuery from advertising services. To do this, you can use OWOX BI Pipelines to automatically collect all of your advertising cost data into BigQuery. Fully-managed. No coding or custom connectors required.
3. If you wish, you can collect revenue data from your CRM/ERP, calling tool, or an internal accounting systems. This will allow you to use offline sales and actual CRM revenue data in your attribution calculations. You can also use custom events as funnel steps: calls, emails, personal meetings, etc.
4. In your OWOX BI project, you can apply a set of transformation templates to build 5 attribution models at once and run all of the calculations.
5. You can apply an all-in-one performance dashboard template for Looker Studio to get your calculation visualized in just a few clicks.

Most of our customers use OWOX BI to correctly evaluate existing marketing campaigns and plan future ones. This is necessary to execute a sales or revenue plan with a planned return on ad spend (ROAS). Here are the key 12 reasons why we believe OWOX BI is the best choice for marketers.

Our attribution model is based on complete data: media advertising impressions and post-view conversions, user activity on your website, cost data from advertising services, transaction data from your CRM, and any other online and offline events. As a result, you see the whole picture of users’ interactions with your business and can take into account the impact of all marketing efforts on business indicators.

“Conversion Revenue” in Google Analytics 4 and revenue from purchased goods in your CRM are often missmatched. Often? Always. The only question is… Is that difference affordable to the business decision-making?

With OWOX BI, you can measure the contribution of online marketing to real sales, not just tracked sales, by taking into account the revenue from completed orders in your CRM and even physical stores.

Unlike standard attribution models, the OWOX BI model takes into account every step of the user before the order — not just the final one. The harder it is to pass a step, the more value a session gets that helps make it happen. By knowing the value of sessions, you can calculate the value of campaigns, taking into account their contribution to promoting users through the funnel.

You can grow faster than your competitors by exploiting your business’s unique advantages. With OWOX BI, you can create a funnel that meets the structure of your business.

In addition to the actions of users on your website, you can add any custom events from your CRM, call tracking system, and other systems as funnel steps. For example, you can upload data about calls, meetings, or emails to Google BigQuery. Combine all business efforts aimed at converting users into one funnel to more accurately evaluate their effectiveness.

It’s not enough to get a comprehensive evaluation of your advertising campaign. Your assessment must be divided into stages of awareness, interest, and conversion using the AIDA model.

For example, with OWOX BI, you might see that campaigns don’t contribute enough into Awareness:

As a result, campaigns designed to influence the next stages of the funnel may look ineffective. A ROAS assessment of existing campaigns may show you how to redistribute your budget between them but won’t reveal the lack of effort at the upper stage of the funnel. OWOX BI gives you an understanding of where to focus your effort.

No algorithm can replace your experience. For example, you may know that your brand campaigns are exhausted and that their high ROAS doesn’t mean you should be investing more in them.

Witn OWOX BI, you can easily specify the channels that won’t participate in your evaluation:

As a result, you get an assessment of only those campaigns you can objectively manage. It’s important that you can do this at any time in a convenient interface without contacting developers or your support team.

OWOX BI calculates the value of each session. This allows you to customize attribution models for different user cohorts, calculate ROI/ROAS for new and returning users, and compare cohort profitability.

For example, mobile operators can evaluate the contribution of advertising campaigns to the sale of additional services to current customers (the “current customers” cohort). Companies from the fashion retail niche can evaluate first-time buyers and next-time buyers separately to find out which channels are better at driving new customers to the business.

Also, by knowing the cost of a session, you can calculate how much you spend and how much you earn on each product group. With this information, you can evaluate the effectiveness of advertising for different regions, landing pages, mobile app versions, and applications.

Learn how to group costs and revenue by any session properties using OWOX BI.

It’s important not only to evaluate campaigns but to make evaluation transparent for businesses. Any black box assessment will be questionable and won’t allow you to find errors.

In OWOX BI, you can see how value is distributed across campaigns for each transaction a user has interacted with:

Errors and incorrect conclusions obtained due to attempts to apply machine learning on insufficient data can be costly for businesses. Our algorithm automatically checks and controls the statistical significance of calculations. In addition, OWOX BI gives your analysts full access to a table of confidence intervals for each segment. Thanks to this, you can be sure of your results and understand how they’re obtained.

The effectiveness of online marketing is increasingly dependent not on the technologies used but on the quality of the data on which models are trained. Therefore, attribution on its own loses out to models built using market data.

For an attribution model to be effective, it also needs to meet the following requirements:

1. Data for calculations must be prepared correctly. You can’t just skip the data collected by your JavaScript tracker through machine learning. For this reason, BigQuery ML is suitable for experiments but not for commercial use.
2. You need to consider information about interactions with the upper part of the funnel: media advertising impressions. To do this, we’ve developed OWOX BI Post-View Pixel, which allows you to measure media ads at the user level. We’ve also integrated OWOX BI with Google Ads Data Hub, which allows BigQuery to request information from Campaign Manager (former DoubleClick Campaign Manager) and some other sources so that businesses can consider and transparently assess the contribution of media advertising to the upper stages of the funnel.
3. The model should consider the expert opinion:
   • Which channels are managed?
   • What is the capacity of these channels?
   • What is the role of a given channel in the funnel?
   • What is the conversion window?

The OWOX BI attribution model meets all these requirements.

Data-driven attribution in Google Analytics 360 demands high data requirements: 400 conversions of each type with a path length of at least two interactions. OWOX BI efficiently organizes events into funnel stages, provides meaningful results with smaller amounts of data, and suits many more companies. At the same time, the statistical significance of calculations is automatically controlled at the level of user cohorts, so you can be sure of the results.

With our templates for Looker Studion and data preparation templates in OWOX BI Trasnsformation, you can get reports based on attribution modeling data without analysts or any knowledge of SQL.

Once the attribution model is calculated, a data mart for reporting is automatically created and you get every you need about revenue, conversions, ROI, and ROAS. All done for you. And automatically updated. Here are some examples of out-of-the-box data attribution reports you can get:

This report allows you to compare the results of calculations using different attribution models. In our case, these are Last Non-Direct Click, which is used in Google Analytics, and ML Funnel Based Attribution from OWOX BI. You can see attributed revenue and ROAS by channel, campaign, user type, region, city, and device. Campaigns that are overrated have a negative difference. Underrated campaigns have a positive difference/value. For example, in the screenshot above, we see that the Yahoo/CPC channel was underestimated, and it’s worth putting more effort into it.

The following report shows how the value of sources and channels is distributed by funnel steps. In the example above, we see that the largest number of purchases (the light green stripe) are made offline.

The report above shows how expenses, funnel steps, and purchases are distributed across customer cohorts. With it, you can determine which channels and campaigns attract transactions from new customers and allocate budgets for them. These campaigns help you increase your customer base.

With OWOX BI, you can create automatically updated audience segments based on any of your data and download these segments to Google Ads. In addition, you can automatically send the results of attribution calculations to Google Ads and Facebook Ads to manage bids considering the real effectiveness of your advertising.

In pioneering the development of funnel-based attribution, we’ve gathered unique experiences that we share with our clients in dozens of articles and successful use cases. Additionally, we’re ready to guarantee a level of data collection and processing quality above 96% in our SLA.

With OWOX, you get not just the beautiful idea of using machine learning in marketing but also step-by-step guides, practical recommendations, and examples of successful use cases.

Attribution use cases:

1. Fabelio found out what online customers are doing in their retail stores
2. ROPO analysis: How useful is it for omnichannel marketing analytics
3. How to Monitor Multiple Business KPIs in a Single Dashboard: Success Story of MatahariMall
4. 9 Steps to Know Everything: Building Marketing Analytics for Comfy
5. How to improve the advertising-to-sales ratio by 10%

Learn how OWOX BI attribution can help your grow your business. Schedule a call to speak to us now.