We at AnswerMiner are proud when such a review about our products landing in the company mailbox. In the following blog post you can read about Bence Buday’s experiences with AnswerMiner and also how he made advantages from the information of his dataset. Let the story begin with his words.
Facebook analysis
by Bence Buday
The pain
In the past two years, Facebook did a significant change in the algorithm that had a huge effect on the post impression of pages. From that time (mostly for marketers), making a well-performing post is a struggle.
Of course, this caused a big change in organic reach (not that big on paid reach). Why? It doesn’t matter. All matters that how can we live with this? No, not rewriting or hacking this algorithm, neither running only paid posts in the future.
Then what can we do?
Page Insight
Those who are working with Facebook pages are aware of Page Insight. Insight is a handy surface within Facebook to have a short retrospection of the performance of your page or your page posts.
Unfortunately, this was not enough for me, since I could not get what I wanted. What I wanted to get from my Facebook Page was knowledge. Not just information about my data, but causes and conclusion. I was looking for answers to my questions:
- What is the reason behind the performance of different types of posts?
- Does the post length count?
- How could I improve my posting quality?
- What kind of Facebook post performs well?
Those who know Facebook Page Insight will recognize this look immediately. Is it enough for you to make decisions about how and what to post?
Dead-end
Within Insight, I have no chance to do an assessment, and something more important what I must mention is that visualization options are limited here. You are not even able to share it with others, and showing them the Insight itself.
From my point of view, it is a must to get a broader look at the life of the page and content. It is not possible here. But, Facebook allows exporting data. After a quick cleanup, I will be able to analyze it so that I can find my answers.
The Solution
Since humankind works with computers, we can use great applications for problem-solving. Thus I was convinced that there must be an application for data analysis and even Facebook data analysis. Something available online, and there is no need for any programming knowledge. That’s how I have found AnswerMiner. This company has offered what I wanted: ”Understand your data in 5 minutes, no coding required”. Say no more! Let me try the application!
The first impressions with the user interface of AnswerMiner
I was pretty much surprised how an application can be so clear and user-friendly. Even after I knew that Facebook connector exists (since it was mentioned in the pricing plan) in AnswerMiner, I had no idea how it will work and how easy it will be. Trust me. It is easy. So I started to generate my first dataset directly connected from Facebook via API.
Dataset within a few clicks
Right after clicking on the Facebook connector in AnswerMiner, a popup window will show and ask you for a few clicks. Linking, choosing, selecting (as you can see in the pictures below).
I recommend selecting “post” instead of “page” since this information is responsible for the performance of the Facebook Page itself.
Without Page Post, the Page won’t perform. At this point, we are not done, because AnswerMiner asks us for what data we want to get from Facebook.
Choose wisely, as some of them are not important. A clear dataset could support your analysis. If you are not sure about what you need, pick as much as you can.
Great Thing! Unlike in Facebook Insight, you don’t have to define the time range. AnswerMiner allows you to examine the whole data archive of your posts!
What happens after the dataset connection?
Insight analytics gives us a clear and nice view of how our page performs and how it did in the past (where past is now minus 180 days, maximum!). However, you only can visualize 28 days inside Facebook. So if I got something better here (as in Insight), I am a winner.
One-click here and another there, and I have found myself in the middle of my Facebook export report. This application had already prepared some charts and the dataset for me in a beautiful way. 4620 cells are waiting for analysis! Go ahead! To be honest, signing up for AnswerMiner was not an “accident”.
I have read all I could about the possibilities and features within the interface of the app. Thus I knew what I wanted to try and where I would find it. My very first visual impression is shown in the picture below this paragraph.
.png)
So I made 74 posts in 2019, a bit more as in 2018. Will more post clicks result more impression? Or inversely? And why is the average negative feedback on my posts are bigger in December?
I immediately understood that my journey in data visualization has just started. Fortunately, Facebook Post is a simple thing. We make it for impressions and attention. All are important (for business purposes) to reach as many relevant people as possible.
How to boost the impression?
The Facebook algorithm has one goal: providing the best user experience. A dull, aimless, harmful, silly post won’t appear on anybody’s timeline. They do not want to lose their users due to a lame post. That is understandable. So what great content is? What gets many impressions without paying for it, right? And what else can I do for a positive impact on my post?
Easy, first, I have to analyze its attributes, which are:
- type of the post (picture, text, video, link, gif, mix, etc.)
- creation time hour, day, month (when the post was published)
- length of the post
Let me explain it! A post can perform badly if I publish it at 2 AM (when I am exhausted, and the followers are asleep), without giving any pictures or video, but a simple link, that would lead the reader outside of Facebook. Nobody will ever give any action to a post like this so that the algorithm will keep it back from people’s timeline.
Also, based on how Facebook works, if the post gets more comments, shares, actions on it, it will be shown for more people. So I clicked on suggested charts in AnswerMiner and checked, is there any connection between length and shares?
AnswerMiner is a professional application, but…
It is a playground if you like visualizations. I spent hours trying to find my favorite one, since there are many, many, and even more. However, the tool will give you charts automatically, so if you want a quick result and don’t want to play around as I did, you will get the job done fast. I prefer columns, but for a better view of the possibilities of different chart usage, I will add some other ones. Please find them at the end of this blog post!
Length and its effect
Well, length and shares. It turned out that those posts, where the length of text was less than 300 (characters) got half as many shares where this number was more than 300. Can you imagine just to get half of your income due to missing characters? All right, move on! Likes, I mean likes summary, which is actions in total in this case. I was surprised again.
Post Shares:
Post Likes:
Those posts, where the length was over 400 characters had performed ca. +30% better than other posts with less than 400. Please have a look on my canvas below and see what the result was at the comments.
Facebook Comments:
(Mean is not equal to the median, please check AnswerMiner blog post about the differences!)
When I saw the result for the impression, I was amazed! It is remunerative if I add some extra alphabets to my 399 characters long text.
Facebook post Impression:
After a few minutes, I was sure that in the future, I would have to write posts over 400 characters. But what if I am not correct? I can destroy my full future social media activity, so I checked the same charts for only 2018 and then in 2019.
2018 length showed me that my statement seems to be correct, since my longest posts had generated the highest median of shares. Nevertheless it also drew my attention to see that if I want to save energy and time a 100-199 character long post also could perform well. Well, what to do now?
Facebook post shares 2018
As 2018 was a long time ago. My brain whispered to me to focus on last year. 2019 showed me that 400-500 seems to be a reasonable interval for the number of letters in a single Facebook post. Writing more characters can be harmful to my social media activity.
Facebook post shares 2019
From now, in this article, I will only analyze my 2019 posts.
Could the day name of posting affect my post?
Why not? The chart below helps me to understand that those posts which were published on Thursday or Friday got a great amount of impression. My understanding is, at this point, that for my Facebook Page, it is efficient if the posts are made at the end of the working week.
Impressions by week day - 2019
Due to my previous experiences with 2018 and 2019, I decided to check this statement from another angle. So I was curious, on which days did the more people engage with my post on Facebook? Even the median of engaged users and the median of shares performed much much better on the last two days of the working week.
Have you read What is Scatter Plot and How to use it? (Guide)? If not yet, you’re not late, but for a better understanding of the chart below, I suggest to read it. I am still focusing on the day name of the week to analyze the performance of posts and for this I tried scatter plot within AnswerMiner. How did I set it up?
Chart settings in AnswerMiner:
- X axis:
post_impression - Y axis: length
- Coloring: day of week name
- X axis min. value: 2k
- Y axis min. value: 300
Why? My goal was to see on which days got my posts the most impressions beside the length of the post. And the result is?! As it is visible on the chart below, there are 3 posts over 300 characters length and 13k impression. Thursday!
Is there any relation between friends of impression?
For the first read, this question can be strange, I know, but after getting to know what “friends” means in this content, you will want to see the answer. There are many “friends” of impression, such as:
- engaged fan
- engaged user
- impression fan
- impression organic
According to AnswerMiner: yes, there is!
This relation-matrix shows how strong the relationship between the data is. You could ask: ”Ok, your data has a great relationship according to AnswerMiner Relation Matrix, but how do you use it?”
Great question. It is easy. If I see that the percentage is high in the table (over 75%), that means it is enough to affect one of them, and the other will change too. Like the more effort I take on post_engaged_users, the more effort will be generated on post_impression. On the other hand, changing the number of post_clicks won’t harm that much post_impression_nonviral.
Summary
There is much more in this small dataset, however I wanted to share my experiences in a blog post, not in a novel. One last chart for the perfect ending.
On this one you can see that I was not that active in the first 4 months of the year and that is right for the weekdays from January to April. Any other things to see? Yes, I should forget to post in November.
After the end of the summer season, before Christmas not that many people are interested in summer pictures.
Not forgetting my pledge, here you can find some beauty chart based on the same dataset Canvas Link
Imagine working with AnswerMiner on a huge Facebook page where there are more than 100.000 cells (not just 4620)!
In the upcoming few months, I will publish 300-499 characters long Facebook page posts (on different days of the week) with pictures or videos in the attachment. I hope that you will read my experiences that time too.
About the author:
Bio: Bence Buday (29) is a proud dad and a happy husband. Currently (next to the full-time engineering job) he is in manager position at Panorama Apartment House (Croatia, Sutivan) where Bence is responsible for the marketing related tasks (and for some others too). He started to widen his marketing knowledge a little bit more than a year ago with learning web page building, social media marketing, copywriting and self development. In his future plan data analyzing placed on a favorable position.
