Steam data tips & insights advent
This December don’t just pick up presents, but knowledge from Santa and the elfs! Get a handy tip, hack or Steam insight every day via our advent calendar and learn to work with Steam and data just that bit more cleverly.
As we share knowledge ranging from Steam insights, refresher maths lessons and dos/don’t s, this promises to be a valuable resource for all Steam pros!
Get new insights and tips every few days until Christmas!
Day 1: Balance your why's
Why? The most asked, and least answered, question in data analysis. Why? Because the why is the final step in truly understanding any problem, anomaly, etc., but it’s also the hardest to answer. Why? Well, for starters it requires you to look at all the variables, every nook and cranny, before you can really answer the why. In the quest for why there always comes a point where data is too labour-intensive to justify, or plainly missing.
That’s why it’s generally accepted when doing analysis to not go all the way, accepting you’re not likely to find a 100% irrefutable answer, and instead find a balance between acceptable probability and effort. Why? Because usually having part of the picture, the most relevant part, is enough. And we have other stuff to do too, don’t we?
Day 2: Gross to net price factors
Looking at 400k transactions in August 2022 we found that the average share of a Steam price that goes to taxes is 15.58%. Depending on your specific audience this can differ, but it’s a handy number to calculate with when you’re looking to set your price point.
Dissecting the listed price further you can say that at a 30% revenue share 25.33% of the gross price goes to Steam, since the 30% is calculated based on net revenue. That leaves 59.09% of the listed price to be paid out. So roughly;
Net = sales price x 0.84
Net net (payout) = sales price x 0.59
These calculations exclude F2P games. We included both full price and discounted sales since the difference was minimal.
Day 3: Locale of spreadsheets
When you’ve ever imported .csvs into spreadsheets, you know this can get messy. Splitting the columns can be a bit tricky, but the decimals is where it gets annoying. And risky.
After importing you notice that some numbers are inexplicably, suspicously high or low, that’s how you know this issue found it’s way into your spreadsheet.
You see, like in many other things the world is divided in how we use separators for decimals, and thousands. For example
2,022.12. Now you and I can see it’s the same number, but during the import process spreadsheet programs like Excel and GSuite are struggling to decode what you need. In cases where the separator doesn’t make sense to the spreadsheet it will plainly take the whole number as text. As an exception it may be able to properly, or improperly decode some numbers, for example ones that don’t happen to have decimals like the rest, or if they are lower than 1000.
When this goes wrong the solution is simple. Change the ‘locale’, the way the spreadsheet will show you the thousand separators and decimals, to match your import before importing. In Google Sheets this is done by going to File, Settings.
Day 4: Don't average averages
Percentages, ratio’s or rates; they do an amazing job of summarizing data. In case you want to summarize even further however, it’s tempting to go too far and make this common mistake: averaging averages.
A quick example. You have a list of daily storepage conversion rates, and you want to know what your conversion rate for a particular week was. Would you take the average of the daily conversion rates of the relevant dates? That result would be wrong.
Think of it this way; every time you make an average you lose a lot of information. That’s great, because averages are the number one way to summarize, and that’s why we love them, but keep in mind a lot of context is lost too. Namely the numbers that contributed to the rate, and their relationship amongst each other.
The best thing to do, is to go back to the numbers the averages were based on, sum all the values on one side, in the example visitors, and the other, in the example conversions, and calculate the overall average from there. This way the weights of each data point is appropriately included in the end result.
Day 5: New tags in 2022
Game trends keep shifting, as they do. To keep up, the available tags on Steam keep evolving, 2022 not being an exception. Welcome 14 new tags! Sometimes based on a big game like Musou or emergence of a new genre like Hobby sims, these tags usually fill an obvious gap.
Most of them, 11, describe a genres. Sports and board game genres were
Tile-Matching. although the latter is a bit ambiguous, which makes for an interesting category page. Other genre tags were
Job Simulator and
For themes & moods we gained
Cozy, the latter two of which have been adopted quite strongly, with over 300 games associated with each.
New tags are not necessarily very popular, but if they fit your niche they could help your discoverability.
Day 6: Calculating change
There’s a few usual suspects when calculating percentages from absolute numbers. The father Christmas of them all, as far as we’re concerned, helps you understand the relative growth between two data points, and should be included in everybody’s mathematical toolbox.
(new-old)/old = change
This looks at the absolute change as part of the old value, and indicates the relative increase, or decrease in case the outcome is negative. The result will be a ratio where 1 means 100%. This connects perfectly with how spreadsheets store percentages, but in other cases you may want to multiply the outcome with 100% to make the ratio into a proper percentage. Do remember to put in the brackets to make sure the subtraction happens before the division.
Day 7: Don't accept simple estimates
Whether you’re building a pitch deck or making financial projections; there’s an undeniable need to understand the sales performance of Steam games you don’t own. It’s easy to find formulas and websites that help you with what’s being referred to as the Boxleiter method, involving the number of reviews of a game to deduce sales. This leads to an easy calculation, but unfortunately a result we know to be wildly unreliable. Maybe it was different in 2015, but reviews do not correlate to financial success today.
That’s why it makes us cringe. We know from analysing 400 million game sales over the past 5 years that Boxleiter results are too inaccurate to be used the way they are. Even for ballpark estimates.
The problem is not in specific outliers that don’t fit the mould. There simply is no one mould. There’s hundreds of small moulds for individual segments within Steam. Even with the Boxleiter method slowly evolving, taking more and more factors into account, like release year, they’re still not suitable to produce acceptable results in this complex context.
The main cause of this is the huge variation in success, and many of the factors leading up to this not easially accessable or even quantifyable, like virality, production budget or marketing effort. ‘Fitting’ game performance into sales, revenue and wishlist predictions is something even our predictive machine learning algorithm is sometimes struggling with, eventhough we’re feeding it actuals from thousands of games.
With a decent sample, about 2.5% of all games on Steam, we are your best bet at getting estimates right.
Day 8: Visibility quality vs. quantity
In the process of getting players into your game the first two steps are getting them to see a reference to your game, and then getting them to click through. The stats for these, accessible via Steamworks, easily accessible in Steam Data Suite, teach us a lot. Impressions being the quantity of your visibility, and the conversion rate from impressions to storepage visits as the quality.
The quantity can be improved by running promotions, changing tags and securing featuring. The quality can be improved by making assets like the capsule images more exciting. Even though these two have different ways to optimize, they are always affecting each other.
Quick example. Imagine adding some interesting tag, that doesn’t quite fit your game. You’ll get additional impressions from the new tag’s tag pages, but those impressions will perform worse at getting clicks since it doesn’t fit what the players are looking for on that page.
Visibility quality, conversion rates from impression to visit, vary strongly per page, or what valve calls the source. Interestingly enough, this means that more successful games have lower conversion rates. The reason is simple. They get more love from the recommendation algorithm and therefore are displayed in more places, more often. But these are usually more generic places as opposed to audience specific. Next, references to these games are displayed more often to the same people. After every exposure the likeliness to click declines. So the 30th time a player sees GTA V pop up is not going to do nearly as much as the first few times.
Analysing traffic patterns for about 700 games, some average conversion rates;
Direct search traffic: 2.52%
Tag page: 1.00%
Upcoming releases: 2.38%
Curator or developer page: 2.60%
We’re looking forward to expanding more on this subject in a separate blogpost, so that we can break the averages over hundreds of games down to more relevant benchmarks.
Day 9: Win+V
Your clipboard, that thing that fills when you do ‘copy’, may sometimes be a hot coming and going of texts and numbers. The limitation of having just one clipboard can mean you have to switch back and forth between apps, pages or sheets more than you care for.
Good thing there’s an upgrade! We all know about Ctrl+C to copy and Ctrl+V to paste, but have you ever tried Win+V? It’s a Windows specific feature that, after turning it on, allows you to not just paste the most recent entry on your clipboard but the ones before it too.
Extremely useful when you’re pasting a small set of texts multiple times, like in a spreadsheet, or to save time not having to go back and forward nearly as much to copy sets of info from one screen to the other.
The clipboard also holds images and even files, so forget about ‘losing’ something important by accidentally overwriting your clipboard! Some people are just all copy, no paste.
On Mac there’s apps that do the same thing, collectively called Clipboard Managers.
Day 10: Cross multiplication
Sometimes you have two numbers that have a relationship (not the sitting in front of the fireplace kind). For one of the two you have an alternative value, and you want to know what it’s partner should be. In comes: cross multiplication.
We like to write a small table on a paper or whiteboard, so it can’t go wrong; each row has the comparable numbers, each column the different value, and one number, of course, is missing. Having the table set up correctly you can multiply the diagonal known values, then divide by the third number.
|Kgs of food/day||44||?|
3 x 44 / 8 = 16.5
The same trick works no matter if the sample value is higher or lower than the ‘original’. For example you can use this to get the sales price without tax. In case tax is 21% and your game is 14 bucks including that 21%, that means
100 x 14 / 121 = 11.57
Day 11: Balance data granularity and capacity
Data is great. These days it’s easy to collect, find and analyse it to your hearts content. Whether it’s data about the market, your performance or player behaviour inside the game, there’s infinite things to track and analyse. But does it always make sense from a business perspective?
There’s no doubt that having the right data based insights helps you make the right decisions and subsequently move your business forward. But it’s also easy to get lost. Because you like analysing, because there’s just so much possible or because you don’t know which insights you really need. ‘Overanalysing’ in this sense is absolutely a thing.
We found that one should try to maintain a healthy balance between spending resources on data collection and analysis and spending resources acting upon those insights. Simply put: it doesn’t matter if you know everything if that leaves no capacity to change anything.
You also shouldn’t underestimate how much having more data makes analysing more complex, leaving even less time to actually iterate (and learn from that). A very basic example of this are decimals. Looking at your data, does it really teach you anything, having 6 decimals for that KPI, or does it only make the overview less readable?
It helps if you know which particular questions you want to have answered.
Day 12: Most successful releases in 2022
|2||Mount & Blade II: Bannerlord||$28.4m|
|4||Dying Light 2 Stay Human||$18.3m|
|5||God of War||$18.1m|
|7||MONSTER HUNTER RISE||$12.8m|
|8||LEGO® Star Wars™: The Skywalker Saga||$12.1m|
Day 13: Hey Google, calculate
Are you still looking for that calculator app on your phone or desktop for some quick maths? What could possibly be quicker than that you ask? Answer: Google! Of course. Input your sum directly into the search bar on your phone, desktop or voice assistant and get an answer immediately.
After doing the initial calculation a graphic calculator shows up in the search results ready for any further mathemagical challenges. That’s easy because, like their physical gray-plastic counterpart, this calculator allows you to use your previous output as input for the next sum.
Make sure to bracket any part of the calculation that needs to happen first. It’s not always needed, but it never hurts. If you’re doing a bit more complex calculations it’s good to know the symbology;
+ or - = add or subtract
* or / = multiply or divide
^ = power
If all of this sounds super obvious, did you also know Google has particular aids for common maths problems? For example type “area of a circle”. And of course you can convert any unit of measurement or even timezones directly too. “10am Amsterdam to Tokyo”. or “20 usd to eur”. Lovely time savers.
Day 14: Don't trust UTM tracking on Steam
Hand in hand with the rise of the visibility issue on Steam grew the need to advertise outside the platform. We need to reach our audience somehow right? To do this properly it’s not enough to just understand which promotions result in clicks, but which actually get people to buy and play.
That practice, figuring out what campaign brought in a ‘buyer’ is as old as the internet and a standard practice in e-commerce and mobile games. But it’s new to Steam. The reason: it’s technically challenging.
Ever since the Steamworks UTM conversion tracking feature was released by Valve about two years ago we’ve talked about it with many companies big and small that all have the same experience: it doesn’t work. Accuracy of attribution in almost all cases being less than 5%. The problem, as Valve admitted when launching the feature into beta, is passing on source information from each step to the next. In particular the step from the web, where the ad lives, to the desktop Steam app, where most of the conversions take place. A tall order indeed.
Our platform includes a campaign source attribution mechanic that is not burdened by this issue, because as opposed to ‘through-tracking’, our approach is ‘around-tracking’. This doesn’t rely on passing data from step to step, instead we identify player activity in the parts of the ad-to-play flow that game publishers control: The ad links and the game code. We’ve learned that trick from mobile and it works great!
Day 15: Part/whole
- Conversion Rate = The share of storepage visitors that choose to buy or wishlist the gameIn any case make sure the part is actually a clean part, and not accidentally includes or excludes more than the whole does. When using this as part of a bigger calculation be sure to encase the calculation in brackets to make sure the order of calculation is right.
- Click Through Rate = The share of impressions on a reference to your game that
- Retention = The share of players that are still playing after a certain amount of time.
- The share of sales that happen in a particular country.
Day 16: Winter Sales trends 2009-2021
The 2022 Winter Sale is almost here, which raises the question for many of us: how do we approach this seasonal promotion the best way possible? There’s many factors that are important; How high do you set the discounts? Are you tying into the season with game content or community? What part of the lifecycle is your game in right now?
This makes it crucial to evaluate the performance of your discounts, so you can at the very least learn from your own experiences moving forward. In Steam Data Suite there’s plug and play dashboards for doing just. From analysing 1800 Winter Sale discount promotions for 1300 games and DLCs since 2009 we’ve obtained some interesting insights.
Between 2009 and 2021 the average discount rate has been varying between 57.5% and 51.5%. Since 2015 this average has been dropping gradually from 57.5% in 2015 to 52.4% in 2020. Last year it seems to have increased again slightly to 53.1%. The right discount for you depends on your overall price strategy more than anything.
The effectiveness of Winter Sales can be summarized in the uplift it’s able to achieve in terms of units sold and revenue. Both of these stats have reached an all time low last year after the previous low of 2018. In 2021 Winter Sale discounts realized on average of 17.5x units sold compared to baseline sales and 12.6x revenue uplift. These numbers vary from year to year, but the effectiveness of Winter Sales overall is trending downward from 39.1x units sold uplift and 22x revenue uplift in 2014.
Day 17: Google Analytics footnote & cleaning
There’s a bunch of reasons to take Google Analytics data from your Steam storepage with a grain of salt. There’s trouble from UA being implemented over GA4, hoe it incorrectly links different sources within the steam ecosystem properly, o, and pretty much all of EU isn’t tracked due to the cookie implementation.
While for the time being this is beyond our control, there’s at least one way to make your store page traffic a little bit more reliable. And it has to do with you. Imagine this. You’re having a really good day in terms of landing page or store page traffic. Awesome! Or wait, was that the day the dev was working on it? Or was that the day I was using it to look up documentation or marketing texts?
To make sure your data stays as ‘clean’ as possible, you can exclude certain IP addresses from being tracked. This is done by creating a view and telling it to ignore calls coming in from your office or colleagues.
To do so, go to Admin from your GA dashboard, then click Create View. You can call it something like ‘Team excluded’. Then with your new view selected and still in the Admin area, click Filters in the column of the View. There create a filter for each IP address you want to exclude.
Day 18: Get significant (conversion data)
You’re running a new version of your store page, it seems to get you a better conversion rate and your excited. Rightfully so. In these cases make sure you don’t forget one thing: significance. If your conversions aren’t numerous it can be many factors, including coincidence, that make the stats turn out the way they do.
Maybe the same week there was a big offline event that got many people informed and excited about your game. Maybe there just happens to be a natural bit of variation in the quality of the visitors and their likeliness to convert.
It’s hard to rule out everything, but please keep these things in mind as they are important context to the conversion KPI’s. Even if nothing particular seems to influence these metrics, we recommend waiting until you get at least 200 sales or wishlists with your new store page before concluding your conversion rate is better or worse, to rule out any coincidence. In Steam Data Suite we provide this significance for your experiments, through a bit more of an advanced formula.
Day 19: Text selection shortcuts
Nothing is linear any more these days! Not even typing. In writing marketing texts, emails or internal reports, you’ll likely find yourself moving back and forward within the text. Sure you can use your mouse to move your cursor, but it’s aaaaall the way over there.
Our coders don’t like that, and don’t do that. Instead they rely on universal text-navigation shortcuts that save them a ton of time. Let them do the same for you! We all know the arrows allow to move the cursor, and PageUp/PageDown, now let’s take it up a notch;
Ctrl + left/right = Move a word left/right
Ctrl + Up/Down = Move to start/end of sentence
Ctrl + Backspace/Delete = Remove previous/next word
Ctrl + Home/End = Move to start or end of the current doc or field
Click 2x/3x = Select word/paragraph
Ctrl+ Shift + v = Paste without formatting
Mac users use the Option key instead of Ctrl. Additionally, holding Shift selects all text between where you stand and where you’re going. This combines with any of the shortcuts above for rapid selection and moving of texts. The shortcuts were used 238 times while writing this text.
Day 20: Visibility & traffic in December
For selling games, December is the most important month of the year. With help of Santa’s purchases, the holidays off and perhaps the generous spirit of Christmas, no other month moves the needle quite like it. Big releases aiming to benefit from this amplify the effect.
Looking at 2022 (numbers adjusted after the year was over) visibility and traffic data from over 500 games we can quantify the effect of the December month like this;
Interestingly enough, visibility, references to your game shown, are down 3% in December compared to the monthly average for 2022. The months between summer and November all have seen more visibility going round. January(-46%) and February(-28%) were by far the worst for visibility, so that promises something for the start of 2023.
In strong contrast, store page visits are up by 39% in December compared to the monthly average for 2022.
The summer sale has proven more effective realizing a similar uplift in visits (+45%), but also a significant uplift in visibility (+49%). So why does the effect on visibility differ more strongly than the temperature? Maybe there were more viral hits this December, or Valve has reduced how many games it recommends on the sales pages. If that’s the case and store page visits have not suffered, maybe the recommendations have become smarter.
Either way, we think the winter is nice, but visibility wise we’re really looking forward to better weather.
Day 21: Who are you really competing with?
Everybody is competing over gamers time and attention, but when you’re looking at competing games to learn from, in most cases you want to stay close to home. Games like CS:GO, DOTA or GTA V have a massive audience, and purely looking at ownership overlap, they sure seem relevant. But can you really learn from what they’re doing?
You want to get inspired, learn how you can stand out, build a business case, and for all of these things relevance is key. Ask yourself; if someone would only buy one game in your niche, which games could they also buy? This is about target audience and niche definition more than ownership.
At the end of the day we’re all competing with the behemoths dominating the game space, other media and even other entertainment. But before operating at the highest level, we believe your focus should remain closer to home.
In Steam Data Suite we can help find who your competitors are from looking at tag overlap, success/scale, and a ton of filters.
Day 22: Our year in review
Last year has been a rocky year for us, but we’ve grown a lot and the results are something to be proud of. That’s why we put them here;
- Over 8 million installs attributed
- $1.3 billion in sales revenue reported
- 120 million unit sales reported
- Nearly 3000 games helped optimize
Numbers were adjusted after the year was over.
Day 23: Causation vs. correlation
We humans love patterns. On our gift wrap, in design, in behaviour and our data; we see them everywhere. So much so we sometimes get over excited seeing hints of a pattern and jump to a conclusion. Helpful in some cases, but dangerous when looking at data.
When two KPI’s seem to have changed in tandem, it’s tempting to believe that one of them caused the change in the other, but often times that’s putting it too bluntly.
We like to think of it this way; KPI’s all exist in a happy family. The trick is figuring out when a pair are siblings or parent and child. Involving a bit more context, zoom into more detail and zoom out to more abstract, can already shine a light on the relationships.
By default it’s good to assume that causation does not mean correlation; Numbers that change in tandem can just as easy have a common cause than be each others cause and effect.
Day 24: Most promising for 2023
On the last day of this advent it seems fitting to look at the new year to come. Looking at our wishlist estimates these are about to be some big successes in 2023.
|1||The Day Before||748.0k|
|3||S.T.A.L.K.E.R. 2: Heart of Chornobyl||571.0k|
|4||STAR WARS Jedi: Survivor™||407.0k|
|5||Company of Heroes 3||358.0k|
|8||Sons Of The Forest||312.0k|