The problem with articles
Most of the time, Google Analytics or other tools give you the average engagement time and the scroll rate. These are two useful but very imprecise metrics. You can of course try to detect whether the user reaches the end of the article, but nothing here lets you automate a clear report showing retention per paragraph.
Number the paragraphs
Each paragraph receives a "number/total" label, for example 04/12 for the 4th paragraph of an article that has 12. Each time a paragraph is displayed on screen, an event goes out with its label.
To do this, ask your developer to change the structure of the article to add a data-para attribute to each paragraph (or any element that justifies it, such as a visual or a table that takes up just as much space). This is the cleanest solution when you have access to the site's code. Put a zero in front of single-digit numbers so that the labels sort in the right order.
<p data-para="04/12">Paragraph text…</p>
Send the event to GA4
In Google Tag Manager
- Create the trigger. Go to Triggers > New and choose the Element Visibility type. Set the selection method to CSS Selector with the value
[data-para], the firing to Once per element, the minimum visible percentage to 50 and the minimum on-screen time to 1000 milliseconds. - Create the variable. Go to Variables > New and choose the Auto-Event Variable type, with the variable type Element Attribute and the attribute name
data-para. Name it Paragraph - label. - Create the tag. Create a GA4 event tag named
paragraph_viewed, with the event parameterparagraph(value: the variable from step 2) and the trigger from step 1. - Test. Start GTM preview mode, open the article and scroll down the page: a tag fires for each paragraph displayed, with the right values.
In GA4
In GA4, go to Admin > Data display > Custom definitions and click Create custom dimension. Name it "Paragraph", set the scope to Event and select the paragraph parameter. It splits events by paragraph label and fills in from the moment it is created.
Reading the reports
Open the Explore menu, create a Free form exploration and add the following items:
- Rows: Paragraph
- Values: Event count
- Filters: Event name exactly matching
paragraph_viewed, then Page path exactly matching the address of the article
The filter on the event is essential. Without it, all the other events of the page are grouped on a "(not set)" row. Then sort the Paragraph dimension in ascending order: the table reads from 01 to the end of the article.
Worked example
An article of 12 paragraphs, 1,000 readers.
| Paragraph | Readers who reached it | Drop from the previous one |
|---|---|---|
| 01/12 | 1,000 | - |
| 02/12 | 940 | 60 |
| 03/12 | 890 | 50 |
| 04/12 | 850 | 40 |
| 05/12 | 820 | 30 |
| 06/12 | 640 | 180 |
| 07/12 | 610 | 30 |
| 08/12 | 590 | 20 |
| 09/12 | 560 | 30 |
| 10/12 | 540 | 20 |
| 11/12 | 520 | 20 |
| 12/12 | 480 | 40 |
Typical drops range from 20 to 60 readers per paragraph.
The drop from paragraph 5 to 6 is 180 readers, which suggests that the 5th is the most problematic, or that the start (before 50%) of the 6th does not attract readers.
Reread your article and challenge it: paragraph too long, change of subject, jargon, inserted link, ad block…
Note: a loss of retention can be beneficial in some cases where the user is directed during the article by a CTA to convert.
Tools
- Google Tag Manager: Element Visibility trigger, Auto-Event Variable, GA4 event tag, preview mode.
- Google Analytics 4: custom dimension, free form explorations.
Measuring by paragraph is part of the tracking expertise, with the GTM setup and the data layer.