See it in action
Five launches. Fourteen days each. Read properly.
The public Steam reviews from the first two weeks of five recent launches, read and classified. The cards are the proof. Below them, what each studio could have done with it.
Flip a card for the numbers.
Each review is split into points, the individual things a player says, classified by theme, emotion and playtime. Act, Investigate and Monitor are the platform's own flags. No studio data was used.
What this changes
The numbers are the easy part. The decision is the point.
Five moments from the cards above, and the call each one puts in front of a studio.
Gears of War: E-Day
What should we prioritise after launch?
What players said
Crashes, freezes and stuttering. Thousands of players, most of them with a proper session behind them, writing the same thing in the first fortnight.
What we found
All four issues flagged Act at launch were already at Act in the August beta. Nothing new appeared. The warning was there eight weeks early.
Audience context
These aren't passing through: 53% of the reviewers had played a proper session and 29% were past 30 hours. Most came for the franchise — 65% of the praise compares it to past Gears. The friction sits with the committed players this game is built for.
What you'd decide
Proposed: make the August fix list patch one — crashes, freezes, stuttering — and hold new content until it has. The issues didn't change between beta and launch, so the fix order shouldn't either.
Proposed experiment
Ship the beta fix list as patch one, with the flagged crashes named in the patch notes, and watch the technical friction before the next content drop.
How you'd check it
In the 14 days after the patch, new reviews: points about crashes, freezes and stuttering, the Technical Issues friction share, and whether the positive rate climbs back toward the 71% of the head start.
Star Wars: Galactic Racer
Is the problem the product, or the price of getting in?
What players said
Buttery smooth and not worth $59.99, often in the same review. 83% positive and a chart-topper, with 1,611 points about value.
What we found
Friction sits with players under three and a half hours. Past that it falls away. The complaint is about the door, not the game.
Audience context
The value complaints come from the edge of the audience: a quarter of the reviewers never passed the drive-by stage, and the friction sits with players under three and a half hours. The 23% past fourteen hours aren't the ones saying it's not worth $59.99. What the player group came for points at the order of work: fix matchmaking, investigate the immersion, protect the art style — and keep watching multiplayer, top of both the friction and the praise.
What you'd decide
Proposed: treat value as a door problem. Test the front of the funnel — a price window, a storefront pitch that leads with the value — before touching the game, because that's the side marketing can move.
Proposed experiment
Run a limited price test — a discount window, or a storefront pitch that leads with the value — and compare the value-friction points in new reviews against the period before.
How you'd check it
In the two weeks after: value-friction points in new reviews, the share of friction from players under 3½ hours, and whether the 83% positive holds.
EA SPORTS FC 27
Which complaints are ours to fix?
What players said
Last year's game with new bugs. One in three would recommend it; the other two are annoyed, and say so.
What we found
40% of the friction is about the industry and the franchise, not this game. The early-access week told the same story at a ninth of the volume.
Audience context
The players writing this are the franchise's core — 52% had played a proper session, 16% were past 53 hours — and the early-access week told the same story at a ninth of the volume. Same audience, louder now.
What you'd decide
Proposed: spend the patch on this game's 60% — the bugs, the balance, the monetisation — and answer the industry-level 40% with communication, not code.
Proposed experiment
Ship one patch that names the regressions it fixes, then split the follow-up friction between 'this game' and 'the industry'.
How you'd check it
In the 14 days after: the this-game versus industry friction split (it was 60/40), points about the named regressions, and whether the recommend rate moves off one in three.
Permafrost
Who are we actually answering?
What players said
Seven patches in thirteen days. Players noticed, and they're still waiting. 71% positive, almost all of them Early Access buyers.
What we found
Only 41% of the reviews are in English. German is 14%, Russian 11%. Most of the conversation isn't in the language the team reads first.
Audience context
The players waiting are the audience the patches are for: 99% of the reviewers are Early Access buyers, 52% had played a proper session, 27% were past 16 hours. And 59% of them are writing in a language the team doesn't read first.
What you'd decide
Proposed: answer in their languages. Translate the next patch notes — German and Russian first — and make the cadence explicit: what's fixed, what's next, and when.
Proposed experiment
Ship the next two patch notes in German and Russian alongside English, and watch the 'still waiting' complaints in those languages.
How you'd check it
In the 14 days after: points about patches and waiting in the German and Russian reviews, and whether the 71% positive moves.
Far Far West
What do we do when nothing is on fire?
What players said
Smooth as butter and funny with friends. 9,845 reviews in fourteen days, 96% positive, nearly all of them Early Access buyers. The 4% who weren't positive write about the same two things as everyone else: bugs and frame rates.
What we found
One topic at Act out of twenty-three: bugs and clipping, 448 points. Three hotfixes in six days fixed the save corruption and bought trust along the way. The grind complaints come from players past fifteen hours, and nobody is asking for a different game.
Audience context
The praise comes from Early Access buyers playing with friends — a third of the reviewers were still under two and a half hours, so the fun is reaching new players. The friction sits deeper in, with the 14% past fifteen hours. Same audience, different parts of the game.
What you'd decide
Proposed: clear the clipping list, explain the unlock system in one note, then ship the melee weapons — in that order. The order is the decision.
Proposed experiment
Ship the clipping hotfix, publish one note explaining how the progression is meant to feel, and hold the melee weapons as the next visible step.
How you'd check it
In the 14 days after: points about clipping (it was 448), points about the grind from players past fifteen hours, and whether the 96% positive holds.
Next step
Now see it on your game
Run it yourself in about two minutes, no sign-up — or book twenty minutes and we'll read your game live on the call.
Run it live
Point it at any game on Steam
Pick any game and watch the product read it, live. It pulls the latest reviews from Steam, reads every single one, and hands back the same read a studio gets — with the real clock running.
- Pick a game
Any Steam title, or one of the one-click examples.
- It reads every review
The pipeline pulls the latest batch and classifies each one.
- You get the read
The surfaces it produces, and the decision they point to — live, on the real clock.
Each square is one review already saved to the corpus.
-
Fetch reviews
—
— -
Classify
—
— -
Surfaces
—
—
Pricing
See what each plan unlocks
Launch, Live and Portfolio, priced per seat. Pick the job you're doing; every plan reads reviews and discussions in audience context.