How to Adjust your Earnings for Seasonality
Today’s NiltoMil topic is a guest post by Rahul Pathak of Lookstat.com about adjusting your earnings for variations and “seasonality.”? This is another great way to know if you’re “winning.”? We always want to know if we’re winning or “doing well.”
How to Adjust your Earnings for Seasonality
If you’ve been tracking your sales in microstock at all, you’ll know that sales on the weekends are about half what they are on a weekday. Seasonality adjustment is a statistical technique that will let you separate out the variation due to seasonal influences (in this case day of the week) from your sales data. The post shows some examples and I’ve included a spreadsheet for those of you that are curious and would like to try this out with your own earnings data.
Raw Earnings
The data used in this example is representative data from last year and covers the period from May 1 2008 to Oct 31 2008. As you can see from the raw earnings data in the chart below, there are a lot of spikes that correspond to the weekends and in general, the chart jumps around a great deal.
Adjusted Earnings
Since we know there is a weekly cycle underlying the earnings data, we can derive a scaling factor that allows us to adjust raw earnings for the effect of weekend days. Conceptually, the process is straightforward – you derive scaling factors and then divide the raw earnings for a day by that day’s scaling factor. This then gives you the chart below.
As you can see the earnings curve has been corrected. The old one is shown with a dashed line to give you a visual reference. Essentially, we have separated the raw earnings curve into two components – a trend component and a seasonal component. If you’re curious about how your data looks with this sort of adjustment, you can simply paste your daily earnings into the attached spreadsheet. It’s pretty simple and will generate all the charts you see in here for your data. You’re welcome to do whatever you like with it. (Blog mentions are always appreciated!)
Seasonal Factor Detail
As you can see from the chart above, the seasonal component for this data set oscillates around 1 (which is the baseline). In this data set, Saturday is the lowest day and Tuesday is the highest. Your data may show a different pattern.
Why Bother?
Ultimately, the point of tracking and analyzing performance is that we want to get better. We study earnings in order to figure out how we’re doing and to find ways to earn more. Part of the how we’re doing question is centered around the deviation between how you did and how you should have done. Questions like ‘is today slow for you?’ abound on microstock forums. At LookStat, we’re working towards helping contributors answer these sorts of questions.
About Me
I‘m the CEO and Founder of LookStat – a web service for microstock earnings tracking and analysis. We’re committed to creating a set of web services for contributors that ultimately help them make more money by showing them how they are doing and highlight. The service is currently in beta and allows contributors to track earnings by image across multiple sites, automatically. You can check out the service at http://www.lookstat.com or read more about it on our blog at http://blog.lookstat.com
Thanks for reading and happy shooting. Also, a big thank you to Matt for his killer blog and the opportunity to post on it.
And thanks to you, Rahul, for blogging for NiltoMil readers!? If you’d like to guest post on NiltoMil please contact me at matt@niltomil.com for more information!
Share This Post
Related Posts
- January 16, 2012 - Can you still make “passive income” from microstock photography?
- August 5, 2011 - July 2011 Microstock Earnings
- July 16, 2011 - Focus on: Steve Heap – photographer, author
- July 14, 2011 - 5 Ways to make more money in microstock
- July 9, 2011 - Focus on: Photocase
- July 7, 2011 - Alamy pseudonyms
- June 17, 2011 - Mid-month update and progress
- June 16, 2011 - Plan your blog posts (or your microstock shoots)
- May 30, 2011 - How NOT to make money at microstock
15 Comments
Feb 9, 20093:50 pm
You’re right about the moving averages being easier and a reasonable approach to take here. One of the future things with a seasonality adjustment is that you can tweak the data for things like holidays etc more easily that you can with an averaging approach and that’s where I’d like to take it in the future.
I’m going to have to go look up your point re: the 6th degree polynomial 
Feb 9, 20094:41 pm
> I?m going to have to go look up your point re: the 6th degree polynomial
It’s a built-in Excel function that uses a simplistic (i.e. x^6) least squares curve fitting method. Algebra 101 type stuff to be sure, and more accurate methods could be used, but this one is fast and easy. Go here to learn a bit more: http://www.efunda.com/math/leastsquares/leastsquares.cfm
Feb 10, 20094:34 am
Thanks for the explanation and the pointer. I’ll definitely check it out.
Feb 10, 20095:09 am
The term seasonal component (factor) is quite confusing for me. It looks like it represents a weekly cycle. “Seasonal” would suggest change on longer time scale.
It is interesting, but … Why bother? How can I improve performance of my portfolio knowing a weekly cycle or a longer trend in my sales?
I am just plotting my monthly earnings. A year ago when my sales were dominated by a pretty stable at that time iStock, it looked that I could predict the performance of my portfolio. Not any more …
Feb 10, 20099:58 pm
Hi Marek,
Great feedback re: seasonal . It’s really a short cyclical factor in this case. I will correct this in the future. Regarding the point, I think as portfolios grow, knowing what is selling and what is not and how that overlaps with your strengths should help. Having said that, your argument has merit as well.
Rahul
Feb 11, 200912:42 pm
I would like to be able to track performance of individual pictures (or groups of selected images) across different microstock sites …
Feb 11, 200912:45 pm
Marek,
you can track individual image sales (across multiple sites) with LookStat today. We will add the ability to create groups (by shoot, keyword, model etc) and let you track those earnings as well.
Happy to chat/email about this anytime. (Matt – hope it’s ok to post a link here.)
Rahul
http://blog.lookstat.com/2009/01/28/image-sales-history-thumbnails-are-now-clickable/
Feb 11, 20091:13 pm
Absolutely ok I like the discussion – it’s good to hear what people want AND what you can already do. I’m holding on for the agencies…I really care more about being able to see little bits of LOTS of agencies than in depth on IStock. lol
Feb 11, 20096:47 pm
Excellent. More sites coming soon btw.
Feb 12, 20094:42 am
The figure I use to see if I’m ‘winning’ is the average earnings per day per month. ($ per mth)/(days in mth) this standardises the result and gives me a figure I can compare from month to month. Although I guess if more weekends happen to fall in a particular month it might be a bit skewed.
Feb 12, 200910:07 am
Tim – that’s exactly why I don’t use avg earnings per day per month. In February we had 8/28 weekend days = 28.5%. May tho is 10/31 = 32% weekend. When figuring out your earnings for that month, that + holidays ALWAYS comes into play. Same for December.
Feb 12, 20099:35 pm
Matt,
This begs the question, how do you keep track of whether you’re winning? If you’ve covered this elsewhere, my apologies.
Rahul
Feb 12, 200910:13 pm
I have two standards for “winning.” 1) Beating other portfolio size (ie. straight growth) and 2) BDE/BME. If I’m surpassing my previous targets, I’m doing well. If I’m not, it’s disappointing.
I should do a full post on this…there are a lot of “ways to win” this game!
Feb 16, 20095:08 pm
>I should do a full post on this
Yes, would be interesting…
Sorry, comments are closed.
« Quick start for fractals Next Post
Microstock and weight loss »
Blog RSS Feed
Subscribe via E-mail
Follow Me on Twitter
12:30 pm
Isn’t it easier/simpler/more accurate just to use a moving average trend line? A 7 day moving average corrects for daily variances, and a 28 day moving average (or better still, a 6th degree polynomial) corrects for seasonal variances.