# Kindle Vella - November 2022 - Survey Results

Hey Everyone, I wanted to update you all on my findings from the survey I took around the November Kindle Vella Bonus.

** Disclaimer**

**The data I'll be referring to was manually submitted anonymously by over 100 Kindle Vella Authors and as such we can make no warranty or guarantee regarding the correctness or authenticity of the data. You can read the full disclaimer regarding the data available here: https://www.book-genie.com/kindle-vella-bonus-breakdown . Readers should note that these are informal calculations gathered by unvalidated third party submissions.**

**None of the authors, contributors, administrators, owners, or anyone else connected with Book Genie, in any way whatsoever, can be responsible or liable for your use, reliance, or distribution of the information.**

**During the survey I asked Authors to report their results for each each series individually. The following list are metrics I focused on for this month:**

“Bonus” - The bonus earned for a single series as reported in the Vella Dashboard

“Top 250 Faved Rank” - If the same series ranked in the Top 250 Faved at month end its position was recorded. Series that didn’t rank were marked with a “Did not rank” tag

“Royalties Earned” - The total royalties accrued for the month. Due to Amazon’s rounding in the dashboard, the results may be rounded to the nearest dollar.

“Total Episodes Unlocked” - The total episodes unlocked for the series in the month of November as reported in the Vella Dashboard

“Paid Episodes Unlocked" - The paid episodes unlocked for the series in the month of November as reported in the Vella Dashboard

“Episodes Released in November” - The published episodes released in November. Completed series were marked as 0.

“Gained Followers” - An estimation of gained followers in November (Bucketed into groups of 50) - Followers was a hard metric for authors to accurately measure as Amazon does not show historical data in the Vella dashboard and so will be excluded from this and future surveys. For the same reason series likes is difficult to accurately report on at scale.

When analyzing the bonus it is important to remember that authors are paid from a finite pool each month.

This means that the sum of all authors bonuses cannot exceed the total bonus pool announced for the given month. For example in the case of November the bonus pool was $1,000,000. Adding up every bonus that the authors received in November would therefore give you $1,000,000.

Since the bonus pool is finite each author’s bonus would therefore be relative to the bonuses of all authors on the platform for the given month. In short your bonus earnings aren't just dictated by your performance, but compared to the performance of all other authors on the platform.

This means there is no single formula that can be consistently used month on month to accurately predict your bonus. Monthly fluctuations can be influenced by both your performance and the performance of every other author on the platform. Changes in the average author’s performance, an increase in the number of series on the platform, and changes to the bonus pool can impact your bonus.

It should also be noted that episodes unlocked with free tokens are compensated within the bonus. However, we cannot accurately separate these earnings from the bonus.

We can look at each variable (e.g. Paid Episode Unlocks) that has been publicly stated by Amazon to impact the bonus and look at its individual relationship (Correlation) to the bonus to predict the weighting of each variable. NOTE: Amazon has the authority to change the weighting of each variable at any point in time. Meaning our findings shouldn't be considered outside the month measured. (November 2022)

*The correlation coefficient which I have used to compare each variable against bonus earnings is defined as a statistical measure of the strength of a linear relationship between two variables. Its values can range from -1 to 1. Indicating at -1 a perfect negative correlation and at 1 a direct positive correlation, with 0 indicating no correlation.*

Simplified for the non-mathsy the closer to 1 the coefficient is the greater the relationship between that variable and the bonus. The closer to 0 the weaker the relationship.

For the statisticians reading I understand that there are other models we could use. However, given the time taken to reach out, compile, and write up this post I've opted to focus on correlation. In addition, based on the size of the data I was working with, I determined correlation coefficients were the most practical starting point.

My first experiment was to measure the correlation coefficient between the Bonus authors recorded against every other variable. For the first round to help paint a clearer picture I left significant outliers (Bonuses that were abnormally distant from the rest) in. As such these results were heavily influenced by a small set of bonuses.

(Reminder: The closer to 1.000 the stronger the relationship to the bonus the closer to 0.000 the weaker the relationship.)

Royalties Earned: 0.994

Paid Episodes Unlocked: 0.987

Total Episodes Unlocked: 0.984

Episodes Published: 0.015

At this point I decided to remove bonuses larger than $5000 to have a clearer sense of the weightings for the majority of submitters.

With bonuses under $5000 the correlation coefficients were as follows

Royalties Earned: 0.890

Total Episodes Unlocked: 0.535

Paid Episodes Unlocked: 0.525

Episodes Published: 0.159

In this test Royalties had the strongest correlation of all metrics. This trend continued for bonuses under $1000 where the correlation coefficients were as follows.

Royalties Earned: 0.631

Total Episodes Unlocked: 0.24

Paid Episodes Unlocked: 0.216

Episodes Published: 0.160

I’ve included the scattergraphs comparing bonuses “Y Axis” under $5000 & $1000 respectively against royalties “X Axis”

It should be noted that at smaller values (in particular bonuses ranging from $200-$500) the deviation from the trend-line increased. Hence why the correlation coefficient is decreasing as we filter to smaller values.

The reasoning for this could be one of many factors. For example at smaller bonuses other variables that are known to impact your bonus e.g. engagement rates such as number of faves, follows, and likes earned tend to have a wider variance.

I did attempt to investigate what additional variables would cause the variance for bonuses under $1000. Faves earned in a month was unfortunately unmeasurable. Since the top 250 faved list only measures the top 250 highest crowned series, the quantity of faves earned by a series could not be measured. It also meant many series who did not rank but earned a bonus could not be used in this comparison. As such the impact of faves remains unknown.

Engagement such as likes and follows is not historically reported in the dashboard. For this reason we weren't able to gather accurate metrics to analyze the impact here. Episodes published were compared though with the data collected.

No correlation could be found between a higher publishing frequency and bonus. In future I'd like to include the option for completed series to mark themselves as complete in the survey. Secondly I’d like to add an option for new series to mark themselves accordingly.

From the data gathered I found that for the month of November, Royalties had the strongest relationship to the bonus earned.

While I'm sure you all have questions, I'd prefer not to make any predictions as this is an exercise in learning what we can. I fully expect these findings could fluctuate month on month, but to the extent that this helps you better understand what we can learn about the Vella data, I hope this gives some insight (at least as far as we can extrapolate from the voluntarily submitted information.)

Thank you to everyone who contribute to the survey!