There is a lot of related terminology with respect to attribution models. This article will help you to understand exactly what these terms mean and how you can use them during analysis, to make sure you make the most of the options offered in the platform.
Attribution Modes
In Billy Grace, there are two different types of Attribution Modes, Session Date and Event Date. Both are explained below:
Session Date: in this mode, for every touchpoint that receives credit for an event, the credit is allocated to the day the touchpoint occurred.
Event Date: in this mode, for every touchpoint that receives credit, the credit is allocated to the day on which the event occurred.
This is best illustrated by an example of a customer journey and how the credit would be allocated between the different modes:
A visitor visits your website through a Meta ad (campaign A) on March 10
On March 14 the visitor comes back through a Google search ad (campaign B) and makes a purchase
Let's assume, that in this example the attribution model gives 90% (0.9) credit to the Meta ad and 10% (0.1) to the Google ad. Let's see how both Attribution Modes would handle reporting this conversion.
Session Date
If we assume this is the only event that occurred in the period, the Session Date report would like this:
Campaign | Date | Number of Events |
Campaign A (Meta) | March 10 | 0.9 |
Campaign B (Google) | March 14 | 0.1 |
As you can see here, not all credit is allocated on the day the event occurred! Intuitively, you could imagine this as the credit being divided back to the days which caused the conversion.
Benefits of this Attribution Mode are:
The Session date Attribution Mode is best suited for judging marketing performance on a specific day using metrics related to spend (ROAS, CPA etc). This, as it allocates credit to the specific day on which it caused a click (and you also spend the amount to get that click!)
Drawbacks of the Session date Attribution Mode are:
For longer attribution windows, it is unclear how far back the conversion is being divided. A good example is the 'unlimited' attribution window - in this case part of the attribution might be given to a touchpoint as far as 6 months back! This effect enlarges any conversion lag effects in the reports (read more here)
Event Date
For this, the event date report would be:
Campaign | Date | Number of Events |
Campaign A (Meta) | March 14 | 0.9 |
Campaign B (Google) | March 14 | 0.1 |
In the table above, all attribution for the specific event is given to the date the event occurred on. This means that campaign A can get credit on March 14, even though there might not have been any session through or spend for the campaign that day!
Benefits of the Event Date attribution mode are:
The attribution credit is always given on the date the event occurred. This helps in reporting and answering questions such as:
Of the events that occurred this month, how many have been caused by Meta?
What is the total number of events caused by paid channels this month?
As the credit is always given on the date the event occurred on, larger attribution windows are easier to interpret.
Drawbacks of the Event Date attribution mode are:
The attribution credit is not necessarily related to the spend you made on that day for your marketing channels. This makes metrics such as ROAS and CPA hard to interpret. This is best exemplified by an example:
If in the example you spend 100 euros on March 10 for campaign A and 1 euro on March 14, the Session Date CPA will be 111.11 (100 / 0.9), while the Event Date will be 1.11 (1 / 0.9)!
For this reason, metrics related to spend are best analyzed using the Session Date report.
Attribution Windows
Attribution windows are related to the period before an event during which you allow a touchpoint to claim attribution for an event. An example can illustrate this:
Customer has a touchpoint 1 for campaign A on March 3
Customer has touchpoint 2 for campaign B on March 11
Customer has a final touchpoint 2 for campaign C on March 14 before purchasing
With attribution window on unlimited, the credit is divided as 80% (0.8) for touchpoint 1 and 10% (0.1) for touchpoint 2 and 3. Different attribution windows would divide this as:
1-Day attribution window:
Touchpoint | In Window | Credit Allocated |
Campaign A (1) | No | 0 |
Campaign B (2) | No | 0 |
Campaign C (3) | Yes | 0.1 / 0.1 = 1 |
You see here that campaign C is the only touchpoint within the attribution window and thus gets all the credit.
7-day attribution window:
Touchpoint | In Window | Credit Allocated |
Campaign A (1) | No | 0 |
Campaign B (2) | Yes | 0.1 / 0.2 = 0.5 |
Campaign C (3) | Yes | 0.1 / 0.2 = 0.5 |
As you can see, campaign B and C are within the attribution window and thus the credit is split between them.
30-day (and unlimited) attribution window:
Touchpoint | In Window | Credit Allocated |
Campaign A (1) | Yes | 0.9 / 1 = 0.9 |
Campaign B (2) | Yes | 0.1 / 1 = 0.1 |
Campaign C (3) | Yes | 0.1 / 1 = 0.1 |
You see here that all touchpoints are within the attribution window, and thus the credit is split between them.
This image displays attribution windows visually:
Food for thought when selecting attribution windows:
If you select a short attribution window, you need to be aware of the fact that this might cut out essential parts of the customer journey that caused certain events to occur. Whether this is relevant for you, depends on the average journey length before a customer converts.
When selecting a longer attribution window with a date period such as 'last 7 days' or 'last 30 days', you could generally see a decline in results when using the Session Date. This is, because of conversion lag (read more about conversion lag here)and the fact that the touchpoint is being moved 'out of' the period you've selected.
In addition, the future touchpoints for which touchpoints can still get credit did not occur yet! Intuitively, you can see this as the attribution window not being 'closed' yet (it is important to emphasize that this effect only occurs for the Attribution Mode = 'Session Date'. See more about attribution modes here). A closed attribution window cannot expect to receive any more attribution from future dates. So we can define it as:
The end date + the number of days in the of the period you've selected has occurred
For example:
For 1 - 10 March with 1-Day is closed on the 12th of March (10 + 1 = 11th of March has occurred)
For 1 - 10 March with 7-Day is closed on the 18th of March (10 + 7 = 17th of March).
As you can see here, selecting a shorter attribution window allows you to have a 'closed' attribution window more quickly, albeit at the drawback of cutting part of your customer journey off. Whether this is appropriate depends on your specific data and customer journeys.
This also highlights why it can be difficult to use the 'Unlimited' attribution window with the 'Session Date' Attribution Mode as the window will never close (it is unlimited after all).
Conversion lag
Conversion lag highlights the fact that conversions don't always happen immediately after a customer interacts with a marketing touchpoint. Some customers may take minutes, hours, days, or even weeks to convert.
What this means is that for more recent days, your ads might have had touchpoints which will receive credit at a later point, but have not done so yet. This might lead to the metrics of these ads (such as ROAS/ CPA/ Attributed Revenue) still increasing for days (or weeks/months!) after the day has passed, as customer are slowly converting.
This effect will be strongest for recent days, as simply put the most recent touchpoints have the highest likelihood of still converting and their attribution windows are not 'closed' yet. The effect of conversion lag also ties into that of closed/open attribution windows. Read more about that here.