evaluating the various types of Attribution Modeling done for Marketing Analytics

Digital marketing today is scattered - People access from multiple devices, clear cookies, or use multiple browsers, which makes it difficult to track their entire journey.

To understand this, consider below example:
On Monday, a visitor sees your post on Instagram, clicks it, visits your site, and leaves.
On Wednesday, s/he clicks on your ad in IMDb.com, visits your site, and leaves.
On Friday, s/he clicks on a Google Search result, visits your site, and leaves.
On Saturday, s/he types your website's URL in a browser, visit your site, and makes a purchase.

Now, which channel (Insta, IMDb, Google Search, Direct) will you attribute your purchase to?

Attribution Modeling is a framework for analyzing which touchpoint(s) or channel(s) or interaction(s) receive the credit for a conversion.

***** Attribution Models types *****

1.

Last Interaction Attribution or Last-Click or Last-Touch

Though many of the interactions prior to the last-click are important, this model simply ignores them and gives 100% credit to the last interaction of your user before s/he converts on your site/app.

Default attribution model in most platforms, including Google Analytics.

In the case of our example, this model attributes the purchase to direct traffic.

Most accurate.

Simple & Straightforward - Easiest to evaluate.

Fit for those who have a short buying cycle.

It gives a good idea of the strongest channel.

Used if the sales funnel is wide at the top, but narrow at the bottom.

2.
First Interaction Attribution or First-Click

Exactly opposite of Last-click, it gives 100% credit to the first interaction.

In the case of our example, this model attributes the purchase to Insta.

Most accurate.

Simple & Straightforward - Easiest to evaluate.

Fit for those who have a short buying cycle.

If there is a tendency to convert customers immediately, then their first touchpoint is especially important.

Used if the business goal is bringing in new top-of-the-funnel customers.

3. 
Last Non-Direct Click

Exactly the same as Last-click, except that it eliminates any 'direct' interactions that occur right before the conversion.

In the case of our example, this model attributes the purchase to Google Search.

By eliminating the direct traffic, this model assigns value purely to the marketing strategy that led to the conversion.

4. 
Linear Attribution

Splits credit equally between all the interactions.

In the case of our example, this model attributes the purchase equally (25%) to Insta, IMDb, Google Search, Direct.

5. 
Time Decay Attribution

Exactly the same as Linear attribution, except that it also takes into consideration when the touchpoint occurred.
Interactions that occur closer to the time of purchase have more value attributed to them.
The first interaction gets less credit, while the last interaction will get the most.

Fit for those who have a long sales cycle (such as for expensive B2B purchases).

6. 
Position-based Attribution or U-shaped attribution

Exactly the same as Linear attribution, except that 40% weightage is given to 1st interaction, 40% weightage is given to last interaction, and the remaining 20% is equally split between the remaining ones.

In the case of our example, this model attributes the purchase - 40% to Insta, 10% to IMDb, 10% to Google Search, and 40% to Direct.

Fit for businesses that have multiple touchpoints prior to a conversion.

7.
Custom Attribution Models

Google Analytics lets you create these.

Obviously, gives deeper insights.

Difficult to create.

Requires a lot of data.

Used for businesses having long buying cycle and plenty of data.

Credits:

Kaushik.net/avinash/multi-channel-attribution-modeling-good-bad-ugly-models/

NeilPatel.com/blog/best-analytics-attribution-model/
Support.google.com/analytics/answer/1033861

Support.Google.com/analytics/answer/1033861

Support.google.com/analytics/answer/1665189

Support.google.com/analytics/answer/1662518

Support.google.com/analytics/answer/2909452

Support.google.com/analytics/topic/3180362

AgencyAnalytics.com/blog/marketing-attribution-models