Blockchain Business Intelligence: Does Size Matter?

Details

A month after joining Steemit, I realized that the community for the most part rewards based on effort and value. I have always wanted to do an analysis that will back this belief with data. This analysis is about the length of posts in count of characters and its relationship with the payout value.

Outline

To show the relationship between length of posts and count of photos with the payout value clearly, here are the data-points that will be used in this analysis:

• Average Payout Value Sorted by A Pre-Defined Length of Post Range
• Count of Photos vs. Average Payout Value
• Monthly Trend of Character Count & Count of Photos Used on Posts
• Top 25 Authors Sorted by Average Payout Value & How Long are Their Posts
• User-Base Scatter Plot Exploring The Relationship Between Length of Posts and Average Payout Value

Scope of Analysis

I wanted to do this analysis system-wide, but found limitations in terms of the size of data involved. I started with generating the data for the month of January, then delimited to the first seven days in February, but both were returning more rows than what would fit in Microsoft Excel. I then decided to delimit the data to the #philippines tag which will be more meaningful since I know the user-base based on proximity.
This analysis is built from 20 months (June 2016 to January 2018) worth of data for posts where the tag #philippines was used. I made some exclusions in doing the scatter plot based on these criteria: auto-posts based on tags, control over whale level SP, some early adopters, communal account, and frequent used of bid-bots. All these categories are seen by the author to have potential or actually skewing the data-points being studied in this analysis.
Here is a list of users excluded from the scatter plot: markboss, vertical, trafficmonitor, monitorcap, purepinay, surpassinggoogle, filipino-trail, hiroyamagishi, jonmylestan, sasha.shade, englishtchrivy, rcarter.witness, steemph.cebu, robin-ho, janz, elizafhaye, janicehung, rockjon.

Tools

I used arcange's Steem SQL Database in generating the data. I used the Comments Table for the posts, comments, payout, and body. I used the column parent_author to delimit the data to posts and exclude comments. I ran this query in a Microsoft Excel Spreadsheet:

SELECT author, parent_author, body, json_metadata, created, total_payout_value, LEN(body)
FROM Comments
WHERE json_metadata like '%"philippines"%'
GROUP BY author, parent_author, body, json_metadata, created, total_payout_value, LEN(body)

Notice that I included LEN(body) to count the number of characters used in each posts, this captures all characters though and are not limited to text. I still think that character used to format the posts are of relevance to this analysis.
To get the count of photos used, I calculated this in excel using the body column using these formulas:

JPG

=(LEN(body)-LEN(SUBSTITUTE(body,.jpg,"")))/LEN(.jpg)-(LEN(body)-LEN(SUBSTITUTE(body,.jpg],"")))/LEN(.jpg])

PNG

=(LEN(body)-LEN(SUBSTITUTE(body,.png,"")))/LEN(.png)-(LEN(body)-LEN(SUBSTITUTE(body,.png],"")))/LEN(.png])

GIF

=(LEN(body)-LEN(SUBSTITUTE(body,.gif,"")))/LEN(.gif)-(LEN(body)-LEN(SUBSTITUTE(body,.gif],"")))/LEN(.gif])

The charts were built using Microsoft's Power BI.

Results

In the chart above, you will see a direct correlation between the length of post and the average payout value. Taking the assumption of 5. characters as the average length of words, this means that a significant drop of the expected reward when the count of words in a post is lower than 200.

When "googled" the average length of a feature story, the first result will say that it should at least be 500 words, but can get as long as 2000 words. This is validated here as shown in the drop in average payout value as the count of characters drop from an average of 980 words in the first range ($14.6), to 686 words ($6.9), to 294 words ($4.7). We see an even steeper drop in the payout value in the next range 147 words ($1.8).

Count of Photos & Payout Value

This next chart shows us that there isn't necessarily a need to have too many photos in a posts. It looks like having at least 6 is the way to go. In fact, posts that have more than 20 photos have lower payout value at the average compared to the 6-10 and 11-20 range.

This tells us however that having less than 3 photos or no photos in a post has an adverse effect on the average payout value.

Monthly Trend Of Post Length & Average Count of Photos in Contents

I included this chart to show the function of learning curve. I am going to point to periods where the activity increase significantly and how it relates to the average length of posts and the average count of photos. Referring to the above chart, this first happened between January 2017 and March 2017. Between these periods the count of posts almost doubled from January 2017 to March 2017, but the average count of characters per posts reduced to barely half. Then the length of posts built up between March 2017 till November 2017. A new inflection point happened between November 2017 and January 2018 where almost the same story of the first inflection in March 2017 happened.

Top 25 Author with Highest Average Payout (Average Length of Their Posts & Average Count of Photos)

These authors who when using the tag #philippines are the top 25 in terms of average payout per post. At the average this group is using 4800++ characters per post which means roughly 950++ words. They are also using 8 photos per post at the average.

Please note that following the practice of the Top 25 authors shown here might not necessarily result to increase in your posts' payout value immediately. This chart does not put into consideration the function of reputation and Steem Power holding. This merely show what the community seem to appreciate and support more in terms of length and count of photos.

Author, Post Length & Average Payout Scatter Plot

This scatter plot shows that a great majority of the authors who are getting at least $10 payout value at the average at writing post that are at least 2000 characters long (about 392 words). Anything less than 2000 words as shown in this chart is unlikely to have more than $15 payout value.

At the same time, this shows that having a long post does not automatically mean high payout.

Conclusions

Overall, this analysis tells us that the community value contents that are more detailed, and have enough photos to complement the texts. The monthly trend shows us that there is a function of learning curve at play, and the scatter plot shows a guide in terms of minimum length of posts to get decent payout value per post.

This analysis gives newer Steemians some guide that are backed by data around the level of details needed in their post to do well in the platform. It is noted however, that there are other considerations in terms of rewards like reputation and Steem Power holding, and length of post does not necessarily guarantee a higher payout value.

---

Posted on Utopian.io - Rewarding Open Source Contributors

---