White Paper: The impact of titles and descriptions on the engagement of the French Media's Facebook posts

Following the results of two studies carried out in October 2019 by Nonli's Data Science team on the impact of titles and descriptions on the engagement excluding clicks on Facebook (reactions, comments and shares) of type link posts published during the 2nd quarter of 2019 on the pages of 133 or the largest French media in terms of web traffic (PQN, PQR, Magazines, TV, radio), we have decided to condense the results.

The fact that this document specifically focuses on the type link posts on Facebook is justified by 3 main elements:

- the #1 social network in France is Facebook with 27 million daily active users;

- type link publications generate direct traffic for the media - and therefore, revenue (through advertising or subscriptions);

- they accounted for over 84% of media publications on Facebook in 2018.

 

The impact of title size on engagement

Number of characters

The titles of posts over the period studied reached up to 254 characters (including letters, numbers, spaces and special characters), with a majority of them consisting of about 70 characters. 

Let us remind that beyond a certain number of characters, the title is truncated by Facebook. This number differs mainly according to the size of the characters used and the screen size of the device used, but we can assume that beyond 100 characters, the title will be truncated.

As can be seen in the graph below, the engagement tends to increase with the number of characters in the titles and it seems preferable to favor long titles, unlike what we observed in the case of Facebook publication messages for the same media over the same period of time (Impact of messages on engagement 2019).

The difference is particularly noticeable between the range of less than 46 characters and that of more than 95 characters, with a difference of 76%.

As this result may seem counter-intuitive because truncated titles would engage more than titles displayed in their entirety, we wanted to compare titles whose length allowed us to be sure that they were either truncated or not (less than 40 characters versus more than 120 characters).

In this case, the results allowed us to observe, as can be seen in the second graph below, that the difference between these two ranges is 97%.

Fig2-7Fig3-6

 

Number of terms

Using the same methodology as for characters, we identified that the number of terms in the titles ranged between 0 and 50.

The results are in line with the previous ones: the higher the number of terms, the higher the engagement.

We thus obtain a difference of 92% between titles containing less than 6 terms and those with more than 16 terms.

To confirm these results, we considered two ranges: titles with less than 10 terms and titles with more than 20 terms.

As shown in the graph below, publications with titles containing more than 20 terms are significantly more engaging than those with less than 10 terms, with a difference of about 86% between the two.

 

Fig5-3Fig6-1

 

The impact of title semantics on engagement

Terms used in the titles

By analysing the terms used in the titles from the publications that generated the most engagement, it became clear that they could be categorized into 2 groups: 

1. Hot topics

Here we find all the topics related to the hot news of the period such as the Vincent Lambert case, the Balkany case, the Notre-Dame de Paris fire, the “yellow vests”, Meghan Markle eand the birth of the first child of the princely couple, etc.

2. Generic words

These words include all "generic" terms, whether they are qualifiers ("horror") or common nouns ("child"), refer to persons (such as "Macron") or categories of information ("weather").

We can also see that many engaging terms are linked to negative news, but also that most of the terms identified are linked to political, judicial, financial, catastrophic events, public figures, or football.

Fig7

 

Titles in the form of questions and containing quotations

As many titles are formulated in question form or contain quotations, we wanted to know what the impact on engagement might be.

As can be seen in the graphs below:

- Across all media studied, the engagement of publications is higher when the title is not formulated as a question, about 55% more than when it is not. 

It is important to note, however, that this is not true for all media, the example of So Foot shows an opposite result with a gap of 33%.

- In the case of quotations in titles, the result is similar across all media, with a higher engagement of about 56% for titles without quotations.

A previously stated, these figures are not a reality for everyone, with Sport24 as a counter-example in this context, which presents a 28% higher engagement when titles are quotes.

Fig8Fig9

 

Titles containing numbers

In order to identify whether the presence of numbers in the title had an impact on the engagement of publications, we divided the publications into 5 categories, with titles:

- containing an age or temporality in years,

- containing an amount of money in euros,

- containing a percentage,

- containing a number at the beginning of the title,

- containing no figures.

We can see in the graph below that when the title contains a number, the publication generally engages better, except in the case where the title begins with a number.

The graph also highlights the fact that the temporality in years is linked to a higher engagement than any other category. It should be noted that publications with titles that mention an age are often linked to miscellaneous news reports.

Fig10

 

Titles containing emojis

The first observation regarding emojis is that they are used 6 times less in titles than in messages (see Nonli's study on messages).

The second is that 68% of the titles containing emojis come from Le Figaro’s page, whose strong presence affects the data obtained.

All media combined, it can be seen that publications whose titles contain emojis obtain 21% more engagement than those without emojis.

In order to identify the magnitude of the impact of Le Figaro, we wanted to analyze their particular case. It shows that the gap is slightly higher in their case, with a 27% difference.

Fig11

 

The impact of the presence of a description on the engagement

Here, we wanted to determine whether the presence of a description had an impact on engagement.

As illustrated by the graph below, this parameter can have an impact that varies according to the page that publishes the post, so we looked at 4 scenarios.

1. All media combined

Publications that contain a description have a median standardized engagement that is 13% higher than those without a description.

2. BFM TV

The case of BFM TV is representative of the opposite scenario to that of all media, with a much greater engagement for posts without a description, with a gap of 206%.

3. BFM Eco

No significant difference is to be noted in the case of the BFM Eco page. The results are very close and the error bars too large to allow us to identify and confirm any gap.

4. Ouest France

With Ouest France, we see results that correspond to those obtained for all media, but to a greater extent with a 46% higher engagement with a description.

Fig1_Am-4

 

Conclusion

In view of all the publications studied, several conclusions can be drawn:

1. Type link publications with long titles containing many terms generate more engagement.

2. A major part of the most engaging terms are related to hot news or news that are negative in nature.

3. The engagement is generally higher when a title contains no quotation and is not formulated as a question.

4. Titles containing numbers generally engage more than when they do not, except for titles that begin with a number.

5. Overall, emojis in the titles tend to have a positive impact on engagement - it is important to note that the volume of publications from Le Figaro in this format influences this result.

6. The presence of descriptions in type link posts seems to have a positive impact on the engagement of the latter when considering all media, however the results are very disparate when analyzed individually.

Our recommendations:

- Prefer long titles whose terms have meaning and interest for your audience.

- If possible, avoid quotations, questions and contain numbers in the title.

- It is important to continuously test and adapt the way you publish depending on the reactions of your audience.