Cyberpunk 2077 Steam Reviews
Buggy Start + epic post-launch dev effort + Over 700k reviews = Cool data Playground
Every year at Christmas my wife gifts me a box of chocolate Pop-Tarts and every year I hold out for a day or so before consuming all of them at once. This results in two things, me having a sugar rush (on account of the sugar) and also me not wanting to move (on account of the other 20+ ingredients). Nostalgia win + new years resolution motivation.
What can you do while you are high on a sugar but also incapacitated by food additives? Try out Substack I guess.
So I wrote a script that scrapes all reviews for all games with revenue on Steam. It is taking a few days to run because Steam asks that we only do 100k requests per day which leaves me in a terrible place of not having enough data to do what I want but still wanting to justify avoidance of house chores.
So I decided to analyze Cyberpunk 2077’s Steam reviews. It’s very important and it isn’t just because it’s the biggest file I’ve pulled in that fits nicely in excel. If you see my wife please let her know that this is serious business with global ramifications.
Anyway, please enjoy my Pop-Tart fueled post.
Quick recap of the highlights surrounding Cyberpunk:
100+ awards won at E3 2018
8M preorders according to CD Projekt Red, 30M lifetime game sales plus 8M expansion pack sales
9.4% decline in CD Projekt Red’s stock price at launch
Big hype? Check. Buggy release window? Check. After launching with 80% positive reviews in December, the monthly reviews fell to 65% positive in March. Console was worse.
CD Projekt Red continued fixing bugs and reviews steadily improved, but it’s tough to get enough new/updated reviews to raise the overall score. This is especially the case for a heavily marketed game. Today Cyberpunk has a 94% positive recent (30-day) review score but it’s lifetime sits at 84%.
Reviews aren’t just about the developer+player relationship, there’s a real revenue impact. Here’s how games are grouped based on the number of reviews and relative positive %.
If you are 70%+, congrats, you’re good.
If you are between 70% and 40%, you will still get impressions but in many places your game will have the dreaded orange “~” next to it which can mean up to a 50% decrease in click rate in Steam search results. It also results in lower revenue per visit (RPV) to your Steam product page for most titles because users care about other user’s opinions. Steam partially promotes based on revenue so you feel it on both fronts.
If you are below 40%, Steam will limit your visibility on the platform. Their goal is to pair users with games they like. If a majority of users are saying they don’t like a game, Steam is going to take their word for it even if it is making them money.
So back to Cyberpunk, the CD Projekt Red team was smart and capable. They had a buggy game that needed fixes. Their reputation (and revenue) was on the line so they did something about it. It took them almost two years but they managed to bring the monthly review average up to a 93.5%. Pretty amazing stuff.
Now for the Data:
Updated vs New Reviews:
I don’t know if you’ve ever been on a dev team that is panicking about lower than expected user reviews. I’ve been on a few and inevitably the corporate question of who do we target next, existing or new pops up. It might sound a bit tone deaf, obviously you want your game to be good across the board but sometimes you have to make a choice between a broken FTUE and late-game perf. I don’t have a hard answer for which approach is best but looking at Cyberpunk, new users drove the majority of review improvement. Here is an annualized graph:
Consider the following:
In 2024 Cyberpunk was reviewed over 122k times.
98k of 2024 reviews were first time reviewers and 95% of those reviews were positive.
23k of 2024 reviews were updates to existing reviews. 88% of those reviews were positive.
So by volume and ratio, the winner is new users. That said, for a game four years in market, a major reason people are showing up is probably good word of mouth from the existing audience.
Time of Play vs Initial Reviews
For Cyberpunk, positive reviewers took over 50% longer to leave an initial review than negative reviewers. That’s 41 hours of playtime vs 63 hours of playtime for negative and positive reviewers respectively. It is normal for positive reviewers to come in a little slower than negative reviews but 50% is a bit longer than normal.
Here’s a breakdown of time-to-review by language:
A few points to make here:
Reiterating that positive reviewers take longer to leave a review. They are probably too busy having fun.
This holds for all languages.
If you launch a game that has lower reviews in the first week, it might not be as bad as you think! Remember the hours in the chart are in-game time, not total time from purchase. It takes users weeks or months to leave their first review
The Japanese took a long time to decide if they liked Cyberpunk while Turkey decided quickly.
This is an average not median so the data is going to be skewed to benefit the longer playing initial reviewer.
Time of play by Review Type
Well we know that positive reviewers take longer to review, do they end up playing more and if so, how much more?Here’s a curve with the percentile distribution called out.
Remember that unlike the previous section, this is life-to-date playtime. The median reviewer has over 100 hours of play! Incredible but also a good callout that the audience we are considering has a strong core-bias.
User reviews are important, but don’t take them too seriously
When I’m feeling down and am out of Pop-Tarts I sometimes go look up PUBG negative reviews for users with over 1000 hours in-game because you just know they are raging over losing a match and I empathize with that. I haven’t played much Cyberpunk 2077 but there are 343 players with 1000+ hours that gave the game a negative review. I thought about making a word cloud of them but opted for a tasteful selection of quotes:
“Bait and Switch” -1214 hours played
“Tried it out didn’t like it” -1231 hours played
“Editing this review after 3 years and almost 1700hrs later. And keeping thumbs down.” -1673 hours played, tastefully self-aware
“Sad excuse for a game” -1977 hours played
“not good” -16055 hours played (no that isn’t a typo, 16k+)
Review Scores By Language:
It’s pretty normal for a non-Asia developed game to get lower reviews in East Asia. It’s also unsurprising to see a Warsaw-based studio get good Polish reviews. Dawaj Polsko! Here’s another way to look at that data but over a time series.
For the above graph I filtered just to languages that had 20k+ reviews and bundled the rest as “Other”. Korea is comparatively low but, Busan is wiped out by a weaponized virus in-game so a 73% approval rating is actually pretty good.
Did Developer Responses help?
Look I was hoping for an answer in this dataset too but CDPR only responded to 117 reviews and most of the responses were copy pasted. I’m not the math guy but I am pretty sure that isn’t enough info. I spent the time looking through things though so I’m going to share thoughts:
46 of reviewers updated their review after the developer response. 61% of those reviews were positive so it might have helped a bit. Probably low ROI action.
53 of the dev responses were localized copy paste to get the word out about a specific audio software conflict. No idea why this was so highly prioritized.
The most prolific day was October 15, 2021 where some CM engaged with a bunch of positive comments. I like to think its because that is the day that someone woke up and decided that they were going to read positive reviews and feel good. The result? Twenty five unique dev responses in a day!
Will check back on this once I have all the reviews for all titles pulled in.
Are Free Reviews more positive?
Probably not scalable but users that got the game for free gave the game about 3% more positive reviews than paid users. If you’re desperate to get from 69% to 70%, this might be an idea to sell to an out of touch exec.
I have run out of Pop-Tarts. Hope you have a great New Year!