Indie Author Insights: Word Count & Earnings

There are a lot of factors that go into success as an indie author. Productivity is only one of them. Quality is obviously a major influence, despite how hard it is to measure. Genre, advertising, and passive marketing also have an effect.

Still, writing more has to correlate with earning more. Amazon algorithms, as well as Kindle readers, have a short attention span. They’re always looking for what’s new and hot. There’s a rumored “30-day cliff” after which Amazon stops organically promoting books.

Since I have data on my word count and my earnings, I thought I’d see if the two correlate. I’m using data from January 2018 to April 2021. That should be more than enough to see how these variables are related (at least in this one specific, individual case).

For an initial look, I’ve plotted my word count against my earnings for that time period. I’ve modified the scale so it’s easier to see the relationship. (I’ve removed the actual values because they’re not necessary to the analysis.)

I’m also going to check if my earnings are related to the word count from the prior month. That would make sense since it takes time to edit a book and get it onto the market. Looking at a line chart of that, I don’t see any obvious correlation. With the naked eye, I’d say there seems to be more of a correlation in the first chart.

For those who aren’t into statistics, I’ll give you my conclusion now. My word count and earnings are indeed correlated. Word count explains 9% of the variation in my earnings in a given month. It’s not significantly correlated to my earnings in the following month.

Here are the details…

First, I check the assumptions for parametric tests. The word count variable is continuous. The pairs (earnings and word count) are related, since both are from the same month. The observations are independent (each month’s word count is not influenced by any other month). For normality, I need to check the Kolmogorov-Smirnov statistic. Sig. for word count is greater than 0.05, but the one for earnings is not. My earnings violate the assumption of normality.

Now let’s check out a scatterplot. It looks reasonably linear, and aside from a few outliers, it seems to be homoscedastic as well. From the direction of the relationship, I can already see there’s a positive correlation.

Since my number of data points is greater than 30, I believe it’s fair to disregard the violation of the assumption of normality and go ahead with a Pearson product-moment correlation analysis.

Turns out that word count and earnings have a significant correlation at an alpha level of 0.05. r is 0.304, which is a positive, medium correlation. My word count explains 0.304² = 0.092416 = 9.24% of the variation in my earnings per month.

That’s kind of cool! Of course, this is a correlation and not a causation. I’m not saying my word count causes my earnings to go up or down by 9%. There are any number of third factors that could be influencing both variables.

Now I’ll take a quick look at how my earnings correlate to my word count in the prior month. Surprisingly, Sig. is greater than 0.05, which means that we don’t have enough confidence in the results to reach statistical significance. My word count in one month is not significantly correlated to my earnings in the next.

So there we have it — productivity is in fact correlated to the earnings in the current month. Next time, I’ll see if there are any patterns based on the day of the week a book was published.