Using Steam reviews to estimate sales

May 1st, 2018

Back in 2014, before SteamSpy, I wrote this article about using Steam’s review count to estimate sales. Then SteamSpy came along and we all used that instead.

However, now that SteamSpy is unable to show the number of owners due to Valve’s privacy policy changes, the good old review count method has become valid again!

I decided it was worth getting an updated “Boxleiter” number (how much you should multiply a game’s review count by in order to estimate units sold). So as per usual I asked a bunch of indies to supply accuate data. This time I got over 50 data points. Read on to discover my conclusions!

The Headline

The average from my dataset is 82x and the median is 77x. e.g. if a game has 1000 reviews you can estimate it has sold about 80000 units. But please keep reading as there are some important things to note.

I think someone mentioned the Boxleiter number was 50x, but in my blog post from 2014 I came up with 30-100x with 70x being the middle, but that included AAA as well as indie games. This new number is for premium indie games only (no F2P), although I did collect data from small titles and big (complex) ones, and the range in success is huge as well.

The Range

There’s also a pretty wide range from 30x to 150x. So if you use the median of 77x, you might have overestimated or underestimated by about 2x. I think it’s still useful though as you can say 1000 reviews could mean 30,000 to 150,000 sales but it’s more likely to be in the middle somewhere.

Notes

- This ratio was calculated from looking at Steam sales only (not “retail” sales from Steam keys such as via Humble Bundle) and “All Reviews” from the top of a game’s store page just under the description (this filters out reviews made by people who got the game through a Steam key).
- These results could contain errors although I’ve done my best to check them with the devs when I got strange readings.
- I’ve excluded any games that launched before Steam reviews went live in Nov 2013 even though some of those results are still probably pretty accurate.
- I also excluded two outliers with values of 18 an 226 but they didn’t really alter the average/median anyway.
- You still can’t estimate revenue using this number because you don’t know how many sales are from discounts and how big those discounts were. If a game is newer you can assume it’s been discounted less and get a better figure, say for the 1st month or 1st year.

Caveat

As with all my posts, I’m just collecting this data and sharing it and my conclusions for fun. Please don’t base any critical business decisions on it.

Does Review Score Matter?

I noted down the current Steam review score for all the titles and you can see the results above. To be honest I can’t see a decent correlation there, maybe a slight one indicated by the trendline. So you could assume that a well-reviewed game will have slightly LESS sales per review, but not many.

Does Launch Date Matter?

YES! I bet you are glad you read this far because I bucketed the results into launch year (see chart above) and there does seem to be a decrease in the ratio as time goes on.

If the trendline is somewhat accurate it could mean that newer games do get more reviews and that over time reviews become less frequent, OR it’s a change in players’ reviewing habits on Steam (especially as reviews were a relatively new thing in 2014), OR it’s a change in the types of games being made because I believe genre probably affects the amount of reviews you get the most (more on this below).

Also it’s possible that games get the most reviews per sale in the first month when true fans buy them but I don’t have decent data on that. Though I was pretty sure I saw this happen for Shadowhand and I’ve checked the ratio for a couple of games released in the last 3 days and it is low (average of 38x). So make of that what you will.

The Headline V2

OK so now we need to tweak the headline a bit. Basically if I ignore games launched in 2014 the average is more like 71x and the median is 64x. It makes sense to use more recent data anyway. The range is still about 30x to 150x.

So what about 2017? Well the average drops to 63x and the median is 65x. The range is about 30x to 100x but I only have 8 data points, so be careful with that.

Does Genre Matter?

I’m pretty sure it does but it’s really hard to attach a label to many indie games and also I’ll end up with too few data points for many of the labels.

But based on what I can see, I think that Roguelikes/sims/strategy have a higher ratio, and adventure/casual/puzzle with story elements have a lower ratio.

This could be because players who feel personally “affected” by a story-based game are more likely to leave a review than someone who was playing a more mechanics-driven game? OR it could be that the player demographics of those games are different and people in those demographics are more or less likely to leave reviews. But it’s just speculation at the moment (and may have to remain that way!)

Anyway, potentially you could use genre to chose which end of the range to edge towards. e.g. for story-based, go lower than the average/median, and for mechanics-driven go higher.

What else could be a factor?

Could the quantity of sales affect things? The idea being that later sales are to less hardcore fans who are less likely to leave a review. Well I plotted that (see below), and no it doesn’t seem to be a factor, which is kinda interesting. Though probably I need more data for highly successful games.

What about Median Play time? Well unfortunately I don’t have that data, but perhaps if players play your game for longer they might be more likely to leave a review. But then again, this goes against my theory that story-based games (which are typically shorter) get more reviews per sale than longer mechanics-driven games.

OK so that’s it for now. If you feel like sharing your data (don’t share actual sales numbers in public), just post your sales per review ratio in the comments (see the notes section above for how to accurately get that data).

Let’s play Shadowhand

April 27th, 2018

In recent weeks Jake has made a series of Let’s Play videos on YouTube featuring our strategic RPG card game, Shadowhand.

The first offers offers tips and tricks for Chapters 1-3:

The second offers detailed strategy for Chapter 4, Smuggler’s bay:

And as you’ll see, there are further videos, currently taking us up to Chapter 7.

If you are already playing the game and need hints, there is a guide, The Outlaw’s Almanack that you will find useful, too.

And of course if you haven’t got your copy yet:

How many wishlists should you have when launching on Steam?

April 24th, 2018


Y axis is exactly what the title says it is. X axis is just anonymised game number. This applies to all graphs in this blog.

Well of course the answer is: As many as possible!

I recently spoke to a bunch of devs and asked them to share their wishlist-related data because I wanted to know the relationship between wishlists at launch and sales at launch. I figured that this would help me to calculate the approximate amount of wishlists required at launch for a “decent” launch.

7-day Wishlist conversions at launch

On the Steam sales reporting system you can view a page of details about your game’s wishlists. If you set the date range to a week (or 8 days to be safe) that encompasses your launch day, you will see a small table at the bottom of the page. That table shows you how many wishlist emails were sent out at launch and how well those wishlists converted into sales during the first 7 days. I asked devs to send me their data and they kindly obliged.

Note that if you launched without setting up a “coming soon” page, this table will be missing. Also if your game is more than a couple of years old, the data isn’t available.

I’ve plotted the results from 14 games that had regular or early access Steam launches and the results are at the top of the page. You can see that there’s quite a large range. The average is 10.5% and the median is about 6.9%. Yes, I could do with more data points, so please let me know yours if you feel inclined!

Note that the highest conversion may be an outlier because the dev already had a huge player base for on online game and he encouraged them to wishlist the Steam game. So this may have resulted in a large fanbase buying during the first week.

Note that our game, Shadowhand, converted at 20% which is also pretty high. We also have a decent fan base from previous games and so did the other game on the chart with 19.5%.

If I exlude the possible outlier at the top then the average is 8.9% and the median is 6.2%.

Some games seemed to have a low conversion rate. Perhaps there weren’t many existing fans, or the price was too high, or the initial reviews weren’t positive. All those things and more could affect conversion rate.

Early Access 7-day Wishlist conversions at full launch

What about games coming out of Early Access into a full launch?

Well I managed to get 7 data points for those and I discovered that the 7-day conversion rate is much lower than for EA or regular launches. The average and median are both 2.4%.

This suggests that most “fans” (I’m using this term to mean customers who are prepared to pay close to full price for a game that they really like the look of) have already bought the game and the others are just hanging on for a deep discount. Also I suspect that the length of time a game is in EA may affect the conversion rate because if it’s in EA a long time then probably more fans will get round to buying it, thus lowering the conversion rate at full launch.

Wishlist sales vs total sales

So how many ADDITIONAL sales will a game make during launch week on top of wishlist sales? Well I asked devs that too and charted the results above as a ratio of wishlists sales to total sales. This chart includes regular and EA launches, as well as EA to full launches because there was virtually no difference between them.

The average ratio was 21.8% and the median was 20.5%.

Many things could affect this ratio. For example, a large fanbase who have already wishlisted the game would push it up resulting in comparatively less organic sales. Or the game could get some great launch press/streamers or a Steam main capsule feature which would lower the ratio due to all the extra organic sales.

Sales per Wishlist

Now things start to get very interesting because we can say that if you had 10000 wishlists before launch then approx 10% of them will convert to sales at launch, which is 1000 sales. Then you can expect 4000 more “organic” sales for a total of 5000 sales in your first week (based on the wishlist sales to total sales ratio of 20%).

Which means that each wishlist is worth approximate 0.5 sales!

In fact I charted it (see chart above) and the average is 0.58 sales per wishlist and the median is 0.36. The full range is from 0.14 to 1.8.

This is super-useful info, especially if you plan to spend money on ads to get wishlists.

Of course, there’s a big range in here. Worst case your 10000 wishlists might only generate 2% conversions for 200 wishlist sales which account for 33% of total sales. Therefore total sales = 600 and each wishlist equates to approx. 0.06 sales. A poor result but also pretty unlikely!

Best case would be something like 10000 wishlists converting at 31%, so 3000 wishlist sales which are only 10% or total sales. Therefore total sales = 30000! So each wishlist would be worth 3 sales in an unlikely scenario where everything was fantastic.

How many wishlists should I aim for?

This year at GDC I was in a meeting with some devs and a Valve rep and they said 50,000 wishlists would be a good figure aim for if you want featuring to kick in. Wow, previously I’d heard about devs trying to get 10K wishlists, and we struggled to achieve that with Shadowhand as we only got about 4000!

Ultimately you want your game to make a profit, or at least to break even. So if you know your total cost (and you should) then you can work out how many units you need to sell to break even.

Let’s say you need 10,000 units to break even in the first week, then you’d need about 20,000 wishlists according to my research (using the 0.5 sales per wishlist ratio).

For a $10 game that would mean $100,000 in gross revenue in the first week, or about $65K net revenue after Valve takes their cut and deals with sales tax.

Then you can use my week 1 x5 formula to calculate your year 1 revenue as per my recent blog post on the subject.

Well I hope you found that interesting. Please let me know your thoughts and feel free to email me any data that you want to share in order to improve my results. Thanks!