Today’s daily dilemma is “What attribution model is better?” Well, each case has its own solution. Discover with us what are attribution models, distinguish them and find out how to choose them for specific campaigns.
What is attribution model
The whole chain of interactions with the brand usually leads to the online request but not the single channel. Imagine a company that has a website, Instagram profile, launched paid ads and remarketing. Here’s an approximate scenario that makes a customer interact with the brand.
- Instagram user has seen a product photo and followed the profile. Thus, after a while, he liked the range so much that he decided to look at the entire catalog.
- A customer googled the brand name and was transferred to the website from paid ads. However, he decided to make the purchase later.
- Two days after he was “caught” by a remarketing with a promotion offer.
- A customer was driven by the remarketing advertisement to the website and has made a request there.
Thus, we have three interactions with the brand before the purchase:
- Instagram;
- paid ads;
- remarketing.
Which one of them is the most valuable from the sales point of view? The customer wouldn’t get acquainted with the brand without Instagram. Without paid ads, he wouldn’t open the website. However, without remarketing the deal wouldn’t be closed. It means that each of these steps and tiny conversions that were made by a customer has its own value. This value needs to be considered to judge the effectiveness of advertisements, count the cost recovery of promotion channels and separate ad campaigns.
Attribution is the rule that transfers the conversion value to different points of contact with the brand. Based on the attribution the analytics system defines the value that will be divided between specific channels.
Channels that helps to show off are important for a business that only enters the market. But for a company that is recognized on the market is more valuable to know sources that lead to the purchase. That’s why attribution models need to be chosen with special care. Otherwise, you may underestimate a channel that influenced the whole process of the deal and get rid of it. All in all, you’ll receive the sales that are down.
Attribution models types
Last Non-Direct Click
This model is used as a default for non-Multi-Channel Funnels Google Analytics reports. Due to this model, 100% of conversion value is attributed to the last transfer to the website that took place before a direct one.
Let’s say a customer came from paid ads. He liked the product, however, he couldn’t fill in a request at the moment. He bookmarked the website, then opened it from the tab and made a purchase. Transfer from paid ads is the most important here. Without this model, it may seem that the direct/(none) was the most valuable as the customer made a direct transfer before the purchase. Nevertheless, the Last non-direct click attribution model shows that the final target conversion was made due to contextual advertising.
The best usage
- You are a newbie and scared to get confused — it’s easier and more understandable than other models as it provides a general presentation of the value of each channel. Well, it is used as a default by Google for a good reason.
First interaction model
This model attributes all value to the first interaction.
The best usage
- If you are only entering the market and your company is unknown. It is a good choice for campaigns that are launched to make the brand more well known, for example, SMM, display advertising, etc.
- For launching brand activities. In this case, the most important thing is the source of the first interaction with your company.
- For monitoring the growth of interest to your brand.
However, such an attribution model is not appropriate to understand the contribution of other channels in the conversion.
Last interaction model
Due to this model, the whole conversion value is attributed to the last channel of interaction.
The best usage
- If your business model underestimates the fast purchase, without considering it. For example, merchandise, food delivery, etc.
- If a campaign is launched to attract customers right at the moment of the purchase. Products selling on Instagram like backpacks that are protected against hustlers, phone covers or makeup.
The disadvantage of this model is that anyway a business is having other online marketing activities, meanwhile. And it is always better to know what was “before”.
Last Google Ads Click
The maximum weight is attributed to the latest Google Ads advertisement that the customer clicked. Other channels are not considered.
The best usage
- If you launch advertisements in Google Ads only. And you need to figure out which creative (campaign or ad group) is the most effective.
Linear model
The conversion value is equally attributed to all interactions in the chain.
The best usage
- If you have a long sales cycle. So the customer needs to be accompanied by advertising activities during each step before the purchase itself.
- When it is important to see all interaction channels.
The disadvantage. This model attributes the same weight to all events, however, you can’t define the ones that were crucial.
Time decay
The most valuable is the last touchpoint and the value of others decreases as time passes.
Considering the time of approaching the conversion, then 7 days after the value of the interaction becomes two times more. Let’s say, a customer opened the website, 7 days after he was driven there by a remarketing and only 7 days more he made a purchase. The first website visit is two times less valuable than the transfer via remarketing. The same situation is with the final conversion. It is two times more important than remarketing.
This model is based on the concept of exponential decay. It means that time slowly changes the value of interaction. Concerning to the advertisement, the closer to the conversion the more valuable the channel that brought it.
The best usage
- For short or one-time ad campaigns. For example, devoted to a holiday sale. Thus, you will define sources that performed better exactly during the promotion days.
Position Based
The greatest value is attributed to the first and to the last channels that performed in the chain of interactions. Due to this model, each of these ad sources receives 40% of the value and the rest 20% is equally divided among channels that are between them.
The best usage
- If it is important to understand the channel that interested the customer and the one that led him to the target conversion.
The Position Based model is rather popular and suits almost all business niches. As any company wants to know not only what influenced the purchase but also where the customer got acquainted with the brand for the first time.
Data-driven
This model is available only if enough data is collected. Read more on this in Google Ads Help. The advantage of this model is that the system independently analyzes all chains of interaction, as well as the probability of the conversion, depending on the customer’s actions. The data-driven attribution model takes information from all products integrated with Google Analytics.
The system defines patterns based on this data and attributes weight to each step. Channels that more often precede the target conversion receive greater weight. The same data is used for an automated rate strategy.
The best usage
- To have a clear understanding of the advertisement’s efficacy and its increasing.
- If a company needs to track only search ads in Google, as well as conversions on the website and conversions that are tracked via Google Analytics.
- If you have enough data. This model simply won’t be available for a website with a low traffic volume.
According to Google, this attribution model allows receiving more conversions having the same budget. But there is one nuance. Data-Driven model is available in Google Ads, and you need to meet the requirements that we referred to above, to get it. There is an analog of this model in Google Analytics 360, however, to receive the access you would have to pay about 150 000$.
Custom model
If none of these models is convenient it is possible to create your own. You can do it in the Google Ads interface. In this case, you independently divide the conversion value between all interactions. To create a custom model it is important to understand the most valuable for your business type of users’ behavior. The following question is as well important: whether there are tiny conversions in the interactions chain that also bring profit to a company, etc.
Use the Model Comparison Tool in Google Analytics to understand which model is the most suitable for your business. It will show by the chosen metric how many conversions consider the models you are interested in. It is also possible to set up to 4 segments to compare.
This screenshot is taken from Google Analytics Demo account
Ringostat Event-Driven attribution model
Ringostat created its own attribution model — Event-Driven. It helps to evaluate the contribution of the ad campaign to the final conversion depending on the users’ actions on each funnel stage. This model is used in the end-to-end analytics of our service. To work with it you need to understand the steps that customer is getting through before the purchase, define the probability of its passage, and set priority for each funnel stage. For example:
Let’s say a potential customer comes through the following funnel stages: visit, registration, another visit, payment. The first website visit does not guarantee a purchase. It means that, for example, its value for the conversion is 2%. The registration value is 40%. Another visit increases the probability of the closed deal, however, it still doesn’t mean that a customer will buy something. That’s why, the value is 20% here, and so on for the following stages. Thus, Ringostat user can customize any model, convenient for his business.
The advantage of Ringostat end-to-end analytics is that it also uses time decay that we described above. A user can set it according to his deal cycle. In this case, Ringostat will also consider the time that has passed after the deal.
Using Ringostat end-to-end analytics you can also upload costs manually. Thus, you won’t be limited by systems that we are integrated with. And you can evaluate the cost recovery of investments made, for example, in SEO, banner advertisements, etc.
We described the work of this model in the article “Ringostat released end-to-end analytics: find out more about effective and profitable ads”. Connect this functionality and you will be able to draw conclusions on the effectiveness of advertisements based on the most illustrative data — sales.
The best usage
- If your business has a rather difficult sales funnel. For example, Ringostat provides a more simple model for a pizza delivery service.
- To understand the value of each user’s session before he makes the target conversion.
- You are ready to spend your time on easy setup, that has to be done only once. However, instead, you receive a clear evaluation of advertising channels’ efficacy considering how a user moves by the funnel during each visit.
Resume
Perfect models suitable for any business simply don’t exist. Everything depends on the time that requires the decision and complexity of the product. If you are a newbie it is possible to use last non-direct click or position based models. When a business belongs to the niche that implies a lot of touches with the customer before the purchase it is better to compare existing models and set custom ones. Alternatively, you can use Funnel Based or Data-Based models.
The most important here not to forget that a user often needs more time to make a purchase. He chooses the product and compares it to the competitors’ offers. It is important to think that if everyone makes purchases by the direct transfer to the website, it means that the advertisement doesn’t work. Pay attention to how many times a customer interacted with the website before the purchase, as well as how many devices he used.
