Pilot program closed
CoreFeedback ingests Steam reviews and surfaces recurring experience patterns, friction clusters, archetype splits, and pre/post change shifts — signal you can stand behind in the room. Evidence informs what to fix next; creativity still decides the game you’re building.
A 30-Day Pilot Built Around One Core Question
Most teams already have feedback. What they lack is:
That was the loop we ran with the pilot cohort.
Example Signal
Recurring friction around reload speed and combat pacing, concentrated in competitive play.
Before patch
After patch
Confidence: High (recurrence threshold met)
This isn't sentiment-only. It's pattern movement.
No black-box summarisation.
The output is structured, reproducible, and auditable.
Clusters, archetype splits, time tracking, review-level signals, evidence quality, and emotional journey — so you see what’s recurring, who it affects, and how it moves.
What appears repeatedly across the dataset — not just what's loud.
Each cluster includes:
So you see what's really repeated, who it affects, and how confident you can be.
The same complaint can come from different player types. CoreFeedback splits signals by who said it — story-focused vs min-maxers — so you know who you're fixing for.
Targeted decisions instead of one-size-fits-all fixes.
See how signals change before vs after a patch, campaign, or season — not just a single snapshot.
Define a window:
Measure:
So you can validate whether a change or campaign actually moved the needle.
Every review and snippet gets deterministic signals: emotion/taxonomy, polarity, quality, engagement.
So every insight and weight rests on the same per-review signals — filter and compare consistently.
Experience and engagement bands, reliability scores, and skew warnings so decisions aren't distorted by noise.
So you weight feedback by who said it and how reliable it is — not one vote per review.
Within a single review, emotional tone can move from frustrated to satisfied — you see the pattern, volatility, and resolution.
So you spot recovery or escalation patterns, not just an average score.
The pilot included the full core decision loop. Clustering and the decision layer were simplified; not every signal from the full product was surfaced — enough to run the loop and verify impact.
The loop
Ingest
Snippetise
Classify
Signal Materialisation
Cluster
Compare
Decide
Verify
Ingest → Snippetise → Classify → Signal Materialisation → Cluster → Compare → Decide → Verify → Learn
This is not a one-off report. It's a repeatable decision system.
Who This Is For
Devs & Live Ops
Verify tuning changes and systemic adjustments.
Studio Leadership
Align around what is structurally recurring, not anecdotal.
Marketing
Identify mismatched expectations vs actual player experience.
Publishers & Investors
Assess product risk and trend trajectory.
No lock-in. Exportable outputs.
Deterministic Feedback Analysis. Not Black-Box LLM Summaries.
The pilot cohort has ended. Join the waitlist for wider access — or book a call to talk through the platform.
