Attribution model: How to determine the most effective advertising channel. Standard attribution models

How do your advertising channels interact with each other? What is the best way to distribute funds between them? Should you disable an advertising campaign if it does not bring conversions? All of these painful questions can be answered by studying user behavior and their path to purchase. In this article, I'll show you how to do this using assisted conversions and attribution model comparison in Google Analytics.

What are assisted conversions?

Effective channels attract users who perform targeted actions on the site (transactions, registrations, ordering a call back, etc. - it all depends on the method of monetizing the project). At the same time, sometimes one interaction with the site is enough for a visitor to convert, but not always. More often, the “seven touches” rule works—that’s why a separate tool is used for each stage of the sales funnel. For example, display advertising helps users learn about your product, while search advertising attracts already interested users.

Associated conversions— target actions in which the analyzed channel was an auxiliary source (that is, the final interaction occurred after the transition from another channel). Imagine that you are selling children's toys.

1. The user saw media advertising and went to your site. Among the assortment of the online store, he liked a toy minion, but the user did not complete the transaction because at that time he was not interested in purchasing (this happens often with banner advertising - read).

2. A week later, this visitor was invited to his birthday and he remembered the toys on your site. I searched for “minion toy” and saw yours. search advertising and saved the site in your browser bookmarks to quickly find it after receiving your salary.

3. Finally, during the third direct visit the user ordered a toy. By default, Google Analytics assigns all conversions a value based on the last indirect source of the visit, in our case, search advertising. At the same time, in general reports we will not see that display advertising was one of the factors due to which the user bought a toy on your website.

If one of your channels or sources does not show conversions in regular Google Analytics reports, don’t rush to abandon it, this could be a key step in the user’s path to purchase.

How to view assisted conversions with Google Analytics?

To find out whether a channel or source contributed to a user's conversion journey or not, use the Multi-Channel Funnels report. To do this, go to the “Reports” tab and in the left panel select “Conversions” - “Multi-Channel Funnels”. 1. In the “Overview” sub-item you can see a general summary and visualization of the relationship between different conversion sources.
2. In the sub-item “Associated conversions” you can see direct information about the channels of associated conversions, their quantity and value:
3. The Time to Conversion tab provides useful information to find out how many days it takes your users to make a purchase decision. This information can be used to properly set up remarketing.
Please note that the line “12-30 days before conversion” displays the sum of target actions for the analyzed days. By clicking on the plus next to the line, you will see more accurate information.
4. The last sub-item is “Main conversion paths”. This displays information about how many interactions users make with the site before making a purchase and what channels they use. In our example, direct visits and search advertising lead. This is not all the possibilities for analyzing associated conversions that Google Analytics provides. Next, we'll look at the attribution model comparison tool.

What is attribution and what models exist?

Attribution is the distribution of conversion value between all user interactions with the site before making a transaction.

As I already wrote, by default, Google Analytics reports assign a value to the last indirect user interaction with the site. This information will be useful if the user most often makes a purchasing decision after the first interaction. For example, display advertising for a pizza delivery service can bring conversions already on the first visit to the site.

Let's look at each model in detail, using illustrations from the Google presentation.

100% of the conversion credit is assigned to the first interaction. This model is well suited for measuring the effectiveness of display advertising, since its goal is to familiarize the user with your offer.

In the chain of interactions, 100% of the conversion credit is assigned to the last channel, even if it was a direct link to the site.

3. Google Ads Last Click Model

The last click on a Google Ads ad receives 100% of the conversion value.

Each interaction is assigned the same conversion value. This model can be used when every point of user interaction with the site is equally important.

The closer the interaction is to the moment the target action is completed on the site, the greater its value.

The first and last channels in the chain of interactions will be assigned 40% of the value, the remaining 20% ​​will be evenly distributed among the remaining channels. This model will be useful if you are interested in both the first interaction, when users first learned about your offer, and the last interaction, when the target action was completed on your website.

With this model, you independently distribute the value of conversions between interactions. You can create such a model directly in the Googe Advertising interface.

This model is available in the Google Marketing Platform. It distributes value across all sessions in the chain based on the correlation between the presence of the source in the chain and the conversion of the chain.
The data-driven model can only be used in accounts with a large amount of data (minimum 20 thousand clicks and 800 conversions in 30 days).

1. Select “Reports” in the top panel, then in the left menu follow the path: “Conversions” - “Attribution” - “Model Comparison Tool”.

2. Choose goals that interest you. For example, you may not take into account related actions, such as adding an item to your cart, but only transactions.

3. In the lookback window, select how many days before the conversion to consider for analysis (from 1 to 90 days).

4. Next, you need to select the attribution model with which the report will be built.

4.1. You can choose one of the default attribution models.

4.2. You can also create your own attribution model or import a ready-made one from the Google Analitycs Gallery.

4.3. Another important feature is the choice of several attribution models (maximum of three). For example, let's take attribution models for the last and first interactions.

5.1. By default, you can analyze by sources, channels and their groups.

5.2. You also have the ability to select any parameter from a list of traffic sources, custom parameters, and Google Ads data. 6. And lastly, you can segment the report. For example, compare conversions that occurred as a result of advertising on the first or last interaction.
By applying the segments selected above, you will get the following type of report:
Now you've learned how to use the attribution model comparison tool.

The times when you could work with only one traffic source (for example, SEO) and still have good sales are long gone. Today, only an integrated approach provides truly effective sales growth. However, when working with multiple sources, we are faced with an important question - what role does each channel play in the chain of user interaction with the site (multi-channel sequence) and how to understand the importance of each channel? After all, it depends on understanding:

• how much investment to invest in each source of visitors,
• what is the return from each channel,
• how channels interact with each other.

Pay special attention to interaction. For example, users from social networks may not make a purchase immediately after the transition, but at the same time, it is social networks that inform visitors about your company, and after subsequent interactions, for example, through contextual advertising, visitors make a purchase.

The rules by which the value of a completed conversion is distributed between channels are called attribution. Knowing which channels the visitor used, we can assign a greater or lesser value to each of the channels (or one of them) and, based on this assessment, make a decision about the effectiveness of the channel.

There can be many attribution models, the most common are:

You can select an attribution model in the report Attribution → Comparison Tool :

We wrote more about the tool in the article below; first, let’s look at what the main attribution models are.

1. Last click attribution

In this case, the entire conversion value is assigned to the last source of user contact with the site. It is clear that this is not entirely correct, since on almost all sites, even those offering very cheap products, the user usually makes 2-3 transitions before conversion.

For a site with an expensive or complex product, there may be significantly more such transitions, as the user thinks about, compares, and gets acquainted with information about the product.

2. Attribution based on last indirect click

This is the default attribution model in Google Analytics. All conversion credit is assigned to the last channel if it is not a direct visit (for example, from bookmarks or a URL entered into the browser bar). In the case of a direct site visit, the conversion value is assigned to the previous channel. The logic is quite simple - if a user came to you from bookmarks, it means that in the beginning he must have learned about your site from somewhere.

3. First-click attribution

What is link building in SEO? As the name suggests, it’s the other way around – all the conversion value is assigned to the first channel that allowed the visitor to find out about your offer.

4. First and last click

The value is divided equally between the first and last channel the user clicked in the chain that led to the conversion.

5. Linear attribution model

The conversion value is divided equally between all sources the user clicked on.

6. Attribution model taking into account the recency of interaction

The “closer” a channel is to the moment of conversion, the greater its value. The significance of each interaction decreases as the time until conversion increases.

Google Analytics reports to evaluate the contribution of each traffic source

Understanding the importance of correctly assessing each traffic source and knowing the main types of attribution, we can turn to special Google Analytics reports:

Already looking at the general information in the tab "Review" , we can formulate a general understanding of how traffic sources interact with each other. Each source is indicated by a colored circle; we clearly see what percentage of traffic “intersects” - this means that the visitor used several sources before making a purchase.

Please note that in the upper left corner of the screenshot there is data on associated conversions.

Associated conversions are a visit from some source that was at the beginning or in the middle of the chain of visits, but not at the end, i.e. the number of interactions that did not lead to a conversion, but participated in the chain.

As you can see in the screenshot, out of 744 conversions, 423 (more than half) had preparatory visits. The sources that provided these visits did not lead to a direct sale, but with a high degree of probability we can assume that without these associated conversions there would not have been the conversion itself that generated income.

Important! The Multi-Channel Funnel Report uses a last click attribution model, unlike all other reports which default to last indirect click.

To evaluate associated conversions for each source in more detail, there is a special report called - "Associated Conversions" :

For example, in the screenshot we clearly see that clicking on links gave us 48 conversions over the specified period, in addition, another 58 times this source was an intermediate step for users who ultimately converted.

With e-commerce set up, this report will help you estimate the revenue from each traffic source much more accurately. As you can imagine, this is very important when we decide which sources are worth investing in. You can, of course, focus on the number of conversions without e-commerce, but, of course, this is a less accurate indicator when creating an advertising budget.

To evaluate in more detail how traffic sources interact, you should go to the report "Basic Conversion Funnels" :

This shows all the source combinations that led to the conversion.

For example:

Additional reports that will help you better understand the chain of visits until the moment of conversion - "Time to Conversion" And "Sequence Length". In them you will see statistics on the number of days from the moment of visit to the moment of conversion and the number of visits from any sources until the moment of conversion.

Google Analytics also gives us the opportunity to compare different attribution models Conversion→Attribution → Model Comparison Tool :

This tool allows you to better understand the differences between different attribution options and visually see the value of each channel at different stages.

For example, let's compare the attribution model for last click, first click, and linear attribution:

Note: free search, if we measure conversion only by the last interaction, loses to the direct traffic channel. The site owner, seeing such a report, will immediately shout: the SEO specialist is not working well!

But, comparing with other attribution models, we will see that search traffic is the most powerful for the first interaction, i.e. It is from this channel that real customers will learn about your site. The importance of search traffic is also confirmed by the linear attribution model, where its share is also the highest.

Keep in mind that comparing attributions allows you to look at the success of each channel from different points of view, but to compare channels with each other and evaluate the success of each of them, you must choose a single attribution model.

For example:

- for a short-term campaign aimed at an immediate purchase - by last click;

- for an SMM campaign that increases overall awareness - by first click, etc.

You can also create your own unique attribution model in Google Analytics, but you need to spend quite a lot of time creating it and first evaluating standard models.

If you're serious about budgeting and assessing how each source contributes to your business's success, you can't do without multi-channel funnels and attribution measurement. You need to understand the importance of the channel, not just its contribution to direct sales.

Pay attention to these Google Analytics reports, work with them and various attribution options - this will help you use all traffic channels more efficiently and intelligently. Based on these reports, you can intelligently plan your advertising budget across different channels.

When analyzing website promotion and the profit received from advertising campaigns, it is very important to trace the entire user journey - from the moment they visit the website until they make a purchase. This will give us the opportunity to understand how to further distribute the budget between advertising channels, how these channels interact with each other, which of them is the most effective, and much more.

In practice, such a path may consist of a chain of different traffic sources. For example, a visitor first came to our website through contextual advertising (Paid Search), viewed several pages of the site and left. Later I switched again, but from Organic Search. A few days later I went to the site through a direct source (Direct), entering the address in the browser bar, and made an order.

Example of a user's purchase journey

Thus, before making a transaction (conversion), the user interacted with the site through three different traffic sources:

1. Contextual advertising;
2. Organic search;
3. Direct entry;

To which of them will Google Analytics attribute the achieved goal in its reports? To answer this question it is necessary to understand such concepts as attribution And attribution model. Attribution in web analytics is the rule of distributing the value of a conversion among all interaction stages in the conversion path and assigning a certain number of points (in %) to calculate its effectiveness.

An attribution model is a set of rules by which you decide to determine the value of a conversion. There are 7 different attribution models in Google Analytics:

1. Last interaction;
2. By last indirect click;
3. Last click in AdWords;
4. First interaction;
5. Linear;
6. Temporary recession;
7. Based on position.

Last interaction (last click)

100% of the conversion value is assigned to the last channel in the chain of interactions. In our example this is direct channel.

Attribution Model - Last Interaction

"The Last Crossing".

The advantage of this model is that you can say with 100% certainty which visit resulted in a conversion. However, this also has a disadvantage - it does not take into account the user’s previous interactions with the site. Thus, according to our example in Analytics reports, we will not be able to understand that the user made his first touch through advertising (namely, we spent money on it and through it the user became acquainted with our offer for the first time), and we will also not be able to see that then he carried out a similar search and came across us again, but only through organic matter. The last source took all the value!

This model is recommended to be applied to those projects whose audience is ready to buy immediately and without additional time to think. As a rule, these are goods or services with a quick response - food delivery, calling a taxi, car towing, equipment repair, etc.

By last indirect click

This model is the default for all Google Analytics reports except Multi-Channel Funnels reports. The difference from the first model is that attribution ignores direct visits, and 100% of the value is assigned to the last channel in the chain of interactions. In our example, this is organic search.

Attribution Model - Last Indirect Click

Yandex.Metrica has a similar attribution model called "The Last Significant Transition", in which all sources are conditionally divided into significant and secondary (insignificant). Insignificant ones include direct visits, internal transitions and transitions from saved pages.

Since it is basic in Analytics, it should be used when comparing with other models. The model comparison tool is available in the section "Conversions - Attribution". This will be discussed in more detail in the following chapters.

The disadvantage of this model is that the value of direct interactions is deliberately underestimated.

Last click inAdWords

100% of the conversion credit is assigned to the last AdWords ad in the interaction chain. In our example, this does not mean that 100% will go to contextual advertising (Paid Search channel), since in parallel with Google AdWords you can run campaigns in other advertising systems.

This model is used if you have an advertising campaign in AdWords, and users from your ads come to the site to make transactions. And Google, when introducing such a model into the list of standard Analytics attribution models, did not think about other advertising services except its own.

Web Analytics Guru and Google Evangelist Avinash Kaushik in one of his articles he called this model useless. Therefore, we will adhere to his advice and move on to analyzing the next one.

First interaction

100% of the conversion value is assigned to the first channel in the chain of interactions. In our example, this is contextual advertising.

Attribution Model - First Interaction

Yandex.Metrica has a similar attribution model called "First transition".

Linear attribution model

All channels in the conversion funnel are assigned the same value. In our example, 33%.

Attribution model - Linear

This model is used when the user is exposed to various channels throughout the entire conversion cycle and all points of interaction with a potential client are important when calculating effectiveness. For example, when analyzing blog posts.

Temporal decline (taking into account the duration of interactions)

This model is based on the concept of exponential decay, and the value of the goal increases closer to the last channel. The term comes to Google Analytics from nuclear physics and gives a comprehensive understanding of the essence of the time decay model: the closer to the conversion the touchpoint is, the more valuable it is considered. The remaining points lose value as the time interval increases.

Under this model, the default half-life is seven days. This means that an interaction that occurred seven days before a conversion is half as valuable as one recorded on the same day, and two weeks before it is four times less valuable. Exponential decay occurs throughout the entire period retrospective analysis(by default it is 30 days).

In our example, the channel closest to conversion is direct approach. He gets the most value, then organic search and the smallest %, taking into account the duration of interactions, has contextual advertising.

Attribution Model - Time Decline

The model is applicable to analyze purchases resulting from promotions in order to assign more value to interactions on promotional days. And those completed a week earlier will be rated much lower.

However, some marketers use it in their work more often than the classic one. "By last indirect click", since it is applicable in almost all topics. One can argue for a long time about the value of some transitions compared to others. But everything here is quite logical - the farther this or that channel is from the moment of conversion, the less value it should receive. After all, if previous transitions to the site were no less effective, then why didn’t they lead to conversion?

One of the advantages of the Time Decay model is the ability to specify the length of the half-life and compare it with other baseline models.

Ability to set half-life

Based on position

Based on position, 40% of the value is assigned to the first and last interactions, and the remaining 20% ​​is equally distributed among the others. Attribution model "Position Based" is a hybrid of models "First Interaction" And "Last interaction."

Attribution Model - Position Based

This model is the closest to real life and is recommended to be used when you need to track all touchpoints: from acquaintance and the first expression of interest in your brand, and to the last interaction that led to conversion.

All listed models are standard Google Analytics models. However, users have the ability to create their own attribution models. You can do this using the settings "Attribution Models", which is at the presentation layer in user tools and objects.

Presentation-level attribution models

In the initial stages of working with Google Analytics, I recommend thoroughly understanding the 7 main attribution models and multi-channel funnel reports (we will look at them in a separate chapter), and only then move on to creating your own.

• Vk.com -

We've released a new book, Social Media Content Marketing: How to Get Inside Your Followers' Heads and Make Them Fall in Love with Your Brand.

Conversion attribution is a multi-component concept consisting of the terms “attribution” and “conversion”.

Conversion in business is when a user completes a target action for the company. For example, if your goal is for a user to make a purchase, then your most important conversion would be completing a transaction.

Conversion is divided into macro and micro:

1. Macro conversion is the ultimate action we strive for. For example, the same purchase of goods.
2. Microconversion is the user steps that lead him to macroconversion: registering on the site, adding products, etc.

Attribution is the rule of distributed conversion value. Simply put, assigning “points” to a conversion to calculate its effectiveness.

Ultimately, conversion attribution is a method of determining the effectiveness of a marketing channel, calculating the contribution of a specific entity (channel, keyword, landing page) to the conversion.

Conversion attribution models

What is the attribution model Google Analytics or Yandex Metrics - this is a set of rules by which you decide to determine the value of a conversion. There are 7 standard attribution models in Google Analytics and 2 in Yandex.Metrica.

Let's imagine that a visitor came to your site throughGoogle advertisement, then, after some time, follows a link from a social network, reaching the site on the same day through mailing and a direct link.

1. “Last Interaction” (“Last Touch”) Model: All “points” are given to the last channel in this conversion chain. In our case, this will be a direct link.

1. Last indirect click model: In many ways similar to the previous model, but with the condition that direct visits are ignored, and all the value is assigned to the last channel through which the user came before the macro conversion. In this example, it will be a mailing list.

1. Model " Last click in AdWords: All conversion value is invested in the last click on an AdWords ad. For us, it's a single click.

1. First Interaction Model: The first channel in the chain through which the user made a transition is considered valuable. Here it will again be AdWords.

1. Linear model: All channels in the conversion chain are assigned the same value. There are 4 elements in our chain, and each is assigned 25% of all “points”.

1. Model “Taking into account the duration of interaction”: The closer the user touchpoint is to the time of the macro conversion, the more valuable it will be. The term “exponential decay” is used here, and the default half-life is 7 days. That is, if an action occurred more than 7 days ago, then it is 2 times less valuable, and more than 2 weeks - 4 times less valuable. Let's say that the transition through AdWords was made 8 days ago from the date of purchase. Then this channel will receive 2 times less value than all the others. The most valuable will be the conversion from mail and direct conversion.

1. Position-Based Attribution model: This is a fusion of the first and last touch models. The first and last link in the chain receive 40% of the total conversion value, the remaining 20% ​​is evenly distributed among all participants. In our case, 40% will receive AdWords and direct visits, 10% each will receive a social network and mail.

Yandex metrics attribution models include the first and last transition, the last significant transition.

1. In the case of the first option, the entire value of the conversion goes to the channel through which the first touch occurred.
2. In the second - to the last click that led to the conversion.
3. If everything is clear with the first two, then the last one needs to be dealt with more carefully. The last significant transition is similar to attribution by the last transition, with the only difference that transitions from bookmarks are discarded and only significant sources remain: search, context, social networks.

You can customize your conversion model depending on the conditions of your advertising campaign.

How to choose an attribution model?

The conversion attribution calculation model is selected depending on the characteristics of your business and advertising campaign:

1. The “last interaction” model is right for you if your business is based on purchases and transactions that do not involve a decision-making stage.
2. “Last indirect click” is suitable as a base for comparison with other models and if you do not want to take into account direct visits to the site.
3. AdWords Last Click is used to determine the highest performing AdWords ad.
4. If you want to know what piques your visitors' interest and achieves the first touch, then use the First Touch model.
5. If you are constantly in contact with a potential client through advertising all the way from the first click to conversion, then use the Linear attribution model.
6. For short-term advertising campaigns, it is recommended to use the “Accounting for the recency of interaction” model.
7. If you value the customer's first experience with the product and the final conversion equally, then you should use Line Item Attribution.

• Use the “Last Hop” model for technical analysis of the site and detection of pages without a counter code.
• Set up a “First Transition” model if the user takes a long time to decide on a target action and return to the site from any other traffic sources.
• “Last significant transition” for sites with fast conversion occurring within one visit.

Conversion attribution is a powerful tool used in online marketing to analyze and adjust an advertising campaign. Using it, you can always find effective advertising channels and reduce the costs of business development.