Your Steam launch, finally explained.
A one-time analyst report on your post-launch Steam game. Upload your Sales CSV and app ID. An analyst brief lands in your inbox, reading your refunds, reviews, and patch timeline the way a senior analyst would, without the bill.
Steamworks tells you what happened. It doesn't tell you why.
You shipped. Refunds are a number. Reviews are a firehose. Your Discord has opinions. Somewhere in there is the reason your Tuesday patch helped, your French players churn, and your 74% positive rating has stopped moving. Reading it all takes a week. Nobody has a week.
A count, a country, a category you didn't choose.
Valve gives you a refund reason from their list. Not the one the player actually typed. The interesting signal is in the free text. That text is paginated HTML, not a column you can filter.
800 reviews, one thumb, zero structure.
Steam classifies reviews as Positive or Negative. That's it. 'The online is broken' and 'the art is stunning but the servers are broken' both count as one negative tick.
The anomaly is in a row you never scroll to.
Your refund rate looks fine. Spain's doesn't. Brazil's doesn't. The partner portal sorts by revenue, so the spike hides in row 14 and you don't see it for eight weeks.
Did 1.2.3 actually help? You're guessing.
You pushed a fix. The refund rate drifted down. Was it the fix or the week? Without attribution across sources, your retro is just a story you're telling yourself.
This is what lands in your inbox.
Generated · 2026-04-22 09:41
Corpus · 1,247 reviews · 384 refunds
The refund curve is healing, but matchmaking is silently reshaping who churns, and your Spain cohort is the canary.
Every number traced back to the row it came from.
The one page you send to your publisher.
Headline findings, ranked by likely impact on your title. Written, not bulleted.
The reasons players actually gave, in their words.
Semantic classification of refund notes into a taxonomy derived from your game. Not Steam's categories. Not ours.
What reviewers praise, bury, and argue about.
Positive and negative themes, cross-tabbed with refund themes so you can see where the stories agree and where they diverge.
The row that was hiding in your CSV.
Per-country refund rate, revenue share, and review sentiment. Statistical flags on anything diverging from your global baseline.
Did 1.2 actually do anything?
Before-and-after windows around each patch you list. Effect size, placebo band, confidence note.
What Steam's UI hid from you.
Refunds whose selected category doesn't match their note text. Surfaces real buckets that are bigger than the dashboard suggests.
What moved first, in your data.
Retrospective lead/lag across review sentiment, refund rate, and wishlist churn. The patterns show what moved first in your data.
Concrete, ranked, non-generic.
Prioritized actions tied to specific findings. Instead of 'improve retention' you get 'investigate Spanish matchmaking queue length before your next patch cycle.'
Four steps. No calls. No onboarding.
Paste.
Steam app ID, your Sales Data CSV export from Steamworks, and a plain-text list of your patches with dates.
We fetch the public stuff.
Reviews, store metadata, language splits. Nothing that requires your Steam password. Your partner-portal session stays yours.
The pipeline runs.
Two-pass classification, anomaly detection, patch attribution, cross-source synthesis, fact-check, render.
Report in your inbox.
Hosted URL plus PDF. Share the URL with your team. It stays live indefinitely. You own it.
The ones we get most.
What does the waitlist actually do?+
You get on the list. When reports go live (targeting Q4 2026), the first 50 waitlist emails get a link to buy at the $199 founding price before we open up publicly. No card today. No obligation. If it's not useful when you see it, you delete the email.
Why $199? What's the standard price?+
$299 is the planned standard price. The first 50 customers get $199 as a founding-price thank-you for buying before we have public testimonials. After 50 customers, or once price testing completes, whichever comes first, it moves to $299.
What do you need from me, exactly?+
Three things: Steam app ID, Sales Data CSV (standard Steamworks export), and a plain-text list of your patches with dates ("1.2 shipped Mar 8, hotfix Mar 11"). Reviews, store metadata, and language splits we fetch from Steam's public endpoints. No partner-portal password, no API key.
My game is small. Will the report be thin?+
If your combined refund and review count is too low to analyze, we flag it at intake before you pay. We'd rather refuse a report than ship a bad one. Under a few hundred combined data points, the product isn't for you yet. Come back after your next update.
What's your refund policy?+
Full refund if there's a demonstrable factual error, a section we promised is missing, or delivery fails. 14-day window. "I already knew these things" isn't a refund, but it's a sign we need to do better on the next one, and we want to hear about it.
Who else is this for? Publishers? Console devs?+
v1 is Steam-only, English-only, and aimed at indie and small-studio developers 1 to 6 months post-launch or post-major-update. Console, mobile, non-Steam, and multi-title portfolio features aren't on the near roadmap. If that changes we'll mention it on the waitlist.
What happens to my data?+
Raw inputs (CSVs, refund notes, reviews) are retained 30 days post-delivery, for defect correction only, then purged. The hosted report and its aggregates stay live alongside the report. We don't use your data for any aggregate or benchmark features unless you explicitly opt in at checkout.
Is this written by an LLM?+
The analysis pipeline is LLM-assisted. That's what makes $199 possible. Every numerical claim is schema-enforced to cite its source row, a second model fact-checks the first, and for the first 50 to 100 reports a human (the founder) reads every one before it ships. We'd rather be honest about this than pretend otherwise.
Get on the list. Lock in $199. Ignore us until it ships.
No card. No spam. One email when it's ready, and maybe one earlier asking what you'd want us to look for in your data. That's it.
Join the waitlist →