How to build an effective conversion funnel and calculate an attribution model for it in OWOX BI

The task of an effective attribution model is to help you define which of your paid traffic channels actually lead your users to conversion, and which channels fail to do so.

Knowing this, you can redistribute your advertising budget by rerouting it into effective traffic channels instead of pouring money into the channels that don’t work. As a result, you get more users converting without spending more on your ads.

However, the effective attribution model is not solely about traffic sources. After your users interact with an ad, they go through a sequence of actions that lead them to the eventual conversion.

These actions form a conversion funnel, and your attribution model will be effective only if you define such actions among all the interactions a user makes with your business and build the funnel out of them.

In this article we will tell you:

  1. How to build an effective conversion funnel that considers all the user actions that valuable for your business
  2. How to increase the efficiency of your ad budgets using this conversion funnel and ML Funnel Based Attribution model from OWOX BI

How to build an effective conversion funnel

To build a conversion funnel that considers all the user actions critical for your business, we recommend you to try using the AIDA model.

AIDA is a consumer behavior model that breaks down the purchase-decision process into four stages: Awareness → Interest → Desire → Action.

AIDA is efficient for building an attribution model upon it because it doesn’t limit you in terms of how many user actions you should use to form a conversion funnel. With AIDA, you can take any events you think are impactful for the conversion and simply distribute them within the AIDA hierarchy.

To build an AIDA funnel, you need to:

  1. Define your users’ conversion actions — all the user actions that affect their progression toward the conversion
  2. Analyze if the conversion actions you’ve defined are sufficient to build an effective funnel out of them
  3. Break the conversion actions into groups — the funnel steps — so each step would correspond to one of the AIDA stages: Awareness → Interest → Desire → Action

Having these steps as the basis, you’ll be able to calculate your attribution model and learn which of your paid traffic channels bring new customers and which don’t.

Funnel example

Here’s how a funnel can look for a SaaS business, and how traffic sources are engaged on each stage of the funnel:AIDA_Funnel_example_en.png

Now, let’s look closer at how you build an AIDA funnel.

1. Define the conversion actions based on which you’re going to make up AIDA conversion funnel steps

Conversion actions are all the users’ actions that move them towards the conversion itself.

Define such actions for your website.

And don’t limit yourself in the number of actions to pick. The more meaningful user actions you use for the analysis, the more conclusive the analysis will be.

Yet, make sure you’re able to gather all these events in the Google BigQuery storage — either using the Google Analytics User Behavior Data → Google BigQuery pipeline in OWOX BI or the BigQuery Export feature for Google Analytics 360.

Once you’re done with your conversion actions, you’ll be able to analyze them in Google BigQuery to see if they actually fit to make up an AIDA funnel.

2. Analyze the defined actions to understand if they can form an AIDA funnel

Сonversion actions can be analyzed in different slices. Like:

  • Time from action to conversion
  • Conversion probability after performing a conversion action

To perform the analysis, make up a table with your conversion actions as in the example below:

Conversion action Number of conversion actions for the last 6 months % of conversions happened after a conversion action Average time to conversion after a conversion action
Action 1  257 432  95%  3 days
Action 2  145 765  29%  8 days
Action 3  56 391  7% 14 days
Action ...  .....  .....  .....


Once you assembled the table, build a chart based on the data from it.

Here’s how your chart can look like whether you define your conversion actions right or wrong:AIDA_charts_en.png

The chart shows how the time to conversion changes depending on a conversion action’s AIDA stage. The less the time to conversion, the more probability that an action will lead to conversion.

Right is when the actions you’ve picked for the funnel are scattered evenly across the chart.

In this case, each AIDA stage can become a separate funnel step in your attribution model.

Wrong is when actions ended up being piled together in the chart.

This means you can’t effectively break them into four sequential funnel steps, and the attribution model calculation results will be unreliable. In this case, try to regroup your conversion actions.

3. Group the conversion actions into buckets that will make up the funnel steps

To make up an AIDA funnel, pick the actions you can then group into four AIDA stages.

Awareness: actions that prove a user’s initial knowledge about your product or service. It can be actions like clicking on an online ad.

Interest: actions that prove a user is interested in your product or service. Examples: clicks on links within a blog post or email.

Desire: actions that prove a user is willing to perform a conversion action. Examples: adding a product to cart, searching for a specific product on a website.

Action: the conversion action itself.

Here’s an example how your AIDA funnel may look based on the actions analyzed:

Awareness Interest Desire Action
Actions 1,3,7,9 Actions 2,10,13,19 Actions 4,8,11,12 Actions 6,14,15


Set up and calculate an ML Funnel Based Attribution model based on the AIDA steps you’ve defined

Once you’ve grouped the conversion actions into funnel steps, you can use them as a basis for your attribution model. Having the attribution model calculated, you’ll figure out which traffic sources bring users into the conversion funnel and which of the funnel steps are most important for the conversion.

Knowing this, you can allocate your ad budget into these channels and get more conversions as the result.

All this can be achieved with OWOX BI's ML Funnel Based Attribution.

Why ML Funnel Based Attribution?

Unlike most of the popular attribution models out there (First and Last Non-Direct Click, for example), OWOX BI’s ML Funnel Based Attribution employs machine learning algorithms to carefully consider how every single funnel step affects the conversion.

In the case of the AIDA funnel, ML Funnel Based Attribution will consider all the actions you’ve built the funnel upon and will miss none of the traffic channels that led users to these actions. The Last Non-Direct Click model, on the contrary, will probably miss most of such channels.

Create an attribution model in OWOX BI

  1. Create a model as described in this guide.
  2. In the model settings, create funnel steps. There must be four steps, each representing a separate AIDA stage: Awareness → Interest → Desire → Action.
  3. In each step’s settings, go to Edit and add all the conversion actions that you’ve picked for this step as step conditions. Set conditions using the OR precondition:Attribution_OR_en.png
  4. Now, calculate the attribution model. With the calculation results, you’ll learn what traffic sources led users to the conversion actions in each step.
  5. Next, compare the ML Funnel-based Attribution calculation results with the Last Non-Direct Click attribution results from Google Analytics. Compare them by source/medium/campaign dimensions. You can do it using any other data analysis system convenient. Having this done, you’ll be able to define which traffic sources were impactful to the conversion but missed by the Last Non-Direct Click attribution.

Now you can redistribute your ad budget in favor of the undervalued traffic sources and — as a result — increase the number of conversion actions and conversions themselves.

Was this article helpful?
2 out of 2 found this helpful
Have more questions? Submit a request


Please sign in to leave a comment.