How to select funnel steps correctly when setting up attribution

If you want to optimize advertising costs and increase ROAS, consider and correctly evaluate the contribution of each channel in the chain before selling. The tools of standard Google Analytics aren't enough here because half of the attribution models available in the service give all the value from the transaction to any channel in the chain. And the second half of the models estimates the channel according to a conditional rule, not on real metrics.

Learn more about the pros and cons of all models in the article Google Analytics Attribution Models: Detailed Review and Comparison.

Unlike the standard GA models, OWOX BI Attribution takes into account the mutual influence of channels on conversion and user advancement on the funnel. You can also compare ad performance across your current GA and OWOX BI models to see undervalued or overvalued campaigns.

As a result, you will be able to redistribute the budget to channels that attract users with conversions, but don’t receive any value by Last Non-Direct Click in Google Analytics.

In this article, we explain how to highlight conversion events for funnel steps on the AIDA model and build an attribution model in OWOX BI based on these steps.

The sales funnel is the path the user takes from first encountering the site to buying. According to the AIDA marketing model, this path consists of four stages:

1. Attention — a person paid attention to the site.
2. Interest — became interested in goods.
3. Desire — wanted to buy goods.
4. Action — made a purchase.

In theory, it sounds simple and consistent. In reality, things are a little more complicated. The audience of your site may not behave as you expect. Visitors who come from different advertising channels behave differently. A person can go to the site many times and leave without doing anything; throw goods in the shopping cart; click on different banners, etc. Yes, his actions will be consistent, but each user will have his own sequence.

Also, the selection of communication channels depends on which stage of the funnel the user is on. For example, media advertising works perfectly with a cold audience at the beginning of a funnel that needs to arouse interest. And emails work better with audiences at the final steps of the funnel. These are loyal users familiar with the brand, so the conversion from mailing is higher.

To increase conversion from site visitors to buyers, you need to examine the actions of your audience and correctly select the funnel steps you will use in your attribution model.

With the correct sales funnel, you can:

1. Correctly assess the effectiveness of marketing, find undervalued sources of traffic and optimally redistribute the advertising budget.
2. Understand the needs of your users to make them relevant offers.
3. Divide the audience into segments depending on which stage of the funnel users are at and personalize communication with each segment.
4. Find bottlenecks in the funnel and optimize them.

The task of your advertising at each stage of the funnel is to motivate the user to move to the next stage — closer to buying. For example, conversion actions can be: view the item card, subscribe to the mailing list, add the item to the cart, and more. How can you determine which of these actions are suitable for building an AIDA funnel?

At this point, you need to take all the business-critical actions users take on the site before they convert, such as the first payment.

The selected actions must be tracked on your site and sent to Google BigQuery. This will allow you to perform the next step.

You can analyze conversion actions in different sections, for example:

• Time from activity to conversion.
• Probability of conversion due to the conversion action.

First, you should make a table like this with all the selected events:

You shouldn’t limit yourself to the number of events. The more actions you use for analysis, the better it will be. At the same time, of course, don’t forget about common sense.

Then you have to make a graph based on this table. Here’s an example of such a graph:

The graph shows how the time before conversion changes depending on the conversion action.

Good — when the events you select are evenly distributed on the graph. This will make it possible to build a funnel by dividing these events into clusters, each of which will be a step of the funnel in attribution.

Bad —when events are grouped on the graph in one place. Then you’ll only get one or two steps of the funnel. In such a case, the quality of the model calculations will be low.

Example of potential AIDA funnel on analyzed events:

If you don’t have a project in OWOX BI, you can start using it for free right away.

Before you run attribution calculations, you need to collect all the data in Google BigQuery cloud storage. With OWOX BI, you can send raw data about users’ actions on the site (page views, events, transactions, etc.), cost data, statistics from call tracking and email marketing services from Google Analytics to GBQ.

You can also load order status information and users from your CRM into BigQuery manually. Data in GBQ can be accessed by SQL queries and answered as tables.

For further work and visualization, these tables can be uploaded to CSV or Google Sheets, for example, using our free addon. This solution allows you to build reports without sampling and other restrictions present in the GA interface.

If you have Enhanced E-Commerce for Google Analytics configured on your site, an attribution model will be automatically created when the first table with session data in OWOX BI appears in the BigQuery:

You can create a new model or modify an existing one. To edit, open the model settings and in each step list conversion events via the OR condition:

Then start the calculations. As a result, you will receive information about which traffic sources led to the conversion actions you chose.

The results of the OWOX BI model calculation can be compared to those of the Last Non-Direct Click model in the section called source/medium/campaign:

This will highlight the sources of traffic that led to conversion actions but remained undervalued. For example, in this screenshot, we see that facebook/organic was undervalued, and it is worth investing in more effort.

Next, you will only have to redistribute the budget to increase the number of conversion actions.

On the screen below, you can see how traffic sources are often distributed at different stages of the funnel. If AIDA funnel events are selected well (see chart above), we can redistribute the budget among advertising sources.

If the events aren’t selected successfully, all of them will fall on one or two advertising channels, and then there will be no options between which channels to redistribute the advertising budget.

P. S. If you need help combining data into Google BigQuery and setting up attribution, we are ready to help. Sign up for a demo and we’ll discuss the details.