# What you meant, what players did, what you decided

Analytics tell you what happened. Player Reports tell you what it felt like. Studio Intelligence is where those meet a question you actually care about, and where the answer stops being a browser tab.

## An investigation

You start with a question. Against it you gather evidence: analytics results, player reports, playtest findings, feedback, notes you write yourself. Every piece keeps its provenance, so months later you can see where a claim came from rather than having to trust it. When you have enough, you record the decision and the follow-up you are trying next, and that stays attached to the question rather than being lost in a chat log.

## Release comparison

Inside an investigation you can put one release beside another and see the same metrics for both. Read it as what changed, not as what caused it: a comparison shows a difference between two windows, and it has no way to know what else you shipped in between.

## Analysis

Asking Ravensight to analyze an investigation is a bounded AI action, metered exactly like an analyst question. It reads the evidence you gathered, not the whole account. The decision stays yours: nothing here marks a question resolved on its own.

## Where it connects

A player report can be sent into an investigation from the Player Reports page. Findings the investigation produces are kept in the same studio context graph as the analyst's, so the next game starts from them, and the analyst can search them with `search_studio_intelligence`. A playtest finding needs no such step: once a run finishes, its worst findings already sit in the analyst's memory and in the studio graph as evidence, ready for an investigation to cite.

## Market research and publisher drafts

Market research on the Studio Intelligence page works three ways: you collect sources you name, you ask a question and get a quick look in seconds, or you start deep research that keeps digging in the background. Starting a quick look or deep research needs a member role or above.

### Collect the sources you name

A research run collects up to five comparable sources you name: a game's store page, or any public HTTPS page you are permitted to collect. Each comes back as a dated excerpt with its limitation stated, a relevance estimate and a category. You choose which passages to include, up to five, and only those are used as context, labelled as external research rather than facts about your game.

### Quick look

Ask a market question of up to 1,000 characters and the answer arrives in seconds. Ravensight answers first from the shared market fact base (below), which is free, and two AI providers research the same question in parallel with a small number of web searches each. The answer shows each claim with its status and its sources. If one provider times out, the other still answers. A quick look is one shared AI unit, the same as a research run, and may use the included monthly allowance. When neither provider returns anything usable it is refunded, and the fact base hits are still shown.

### Deep research

Deep research is a background job for a question that deserves more than a quick look. Pick one to four focus areas (competitors, player needs, opportunities, pricing) and, if you like, the quick look it follows from. The run moves through stages you can watch: queued, researching, extracting, reconciling, merging, then complete. A project runs one deep research job at a time, and collecting sources can run while it does.

It costs **25 AI units ($5.00)** from the prepaid balance only, never the included allowance. An empty balance is refused before anything runs, and so is a run when no provider can run at all, so nothing is charged. A provider kept out by a platform limit is shown on the run as skipped. When one provider fails, the run finishes as **partial** on the other provider's results, keeps the charge, and says so on the report. When neither provider returns anything usable it is refunded in full. A run whose charge is uncertain is held for billing review rather than retried or refunded blindly, and an owner can settle or refund it from the run itself. A studio also has a daily research limit.

When a deep run follows a quick look, it lists what changed: claims it added, claims it confirmed, claims it now disputes, and quick look claims it dropped. Each provider's raw report can be read in full from the run when it was kept. A report too large to keep, or one from a run where report storage is not configured, is not kept, and the run says so.

### What a claim's status means

- **Corroborated.** Both AI providers found the claim, each citing public sources. (A fact in the shared fact base is also corroborated when a person added it.) This is agreement between two AI providers on cited public sources, not independent verification.
- **Single source.** Only one provider found the claim.
- **Disputed.** The two providers contradict each other. Read both sources before relying on either.

Player reviews quoted in a report are samples, not a measurement of every player. Nothing from a quick look or deep research enters your studio's context by itself: a person still accepts what is used.

### The shared market fact base

Reports stay private to the studio that asked. The public facts drawn from them go into a market fact base that Ravensight keeps for every studio. A fact is only ever about a public subject: a public game on Steam, a genre, a tag or a platform. Each one carries a public source link, a short quote that supports it and the day it was collected. Facts are written by a separate extraction step that sees only the cited source quotes and titles, never your question, your notes or your studio's names, and a fact the quote does not support is dropped. Anything naming your studio, your projects, your games or a phrase your question puts in quotes is filtered out and stays in your private report, and a fact never says who asked. Because the fact base needs no AI to answer, it keeps answering during an AI provider outage. Reading it needs a read-only role or above.

Facts change in a few plain ways. Finding the same fact again strengthens it in place, adding its sources, rather than writing a copy. Facts that disagree are both marked disputed. A research finding can dispute a fact both providers found only when both providers agree on it from two different sites, and a fact Ravensight added by hand is never disputed by research. A newer figure for the same measure replaces the older fact only when both providers found it; the older fact is kept, marked superseded. A fact can also be retired, after which it no longer appears in search and cannot be cited.

A market fact can be cited as evidence in an investigation, on a [Brainstorm](https://ravensight.io/docs/brainstorm/) card and in a [Strategy matrix](https://ravensight.io/docs/strategy/) cell. When a fact a cell cites is disputed, superseded or retired, the cell is flagged for you to look at. It is never re-scored on its own.

### Publisher drafts

From the included passages you can generate a draft in one of six formats: a positioning brief, a store description, a launch announcement, a trailer outline, a publisher pitch, or the questions and evidence gaps a buyer would raise. Every external claim cites its source by number, a draft that cites nothing is refused and refunded, and the export carries the source list with it. A research run and a draft are one shared AI unit each, and a project keeps up to 50 research runs.

## Background work, on a ceiling you set

The work Ravensight does without being asked each time, weekly digests, pattern synthesis and feedback enrichment, runs under a monthly budget in cents that the owner sets and that is stated against the current rate before it is accepted. It is a ceiling on spend, not a subscription: nothing renews, and switching it off stops the work. Each of the three can be enabled on its own.

Reading an investigation needs a read-only role or above. Creating one, gathering evidence and recording a decision need a member role or above. Setting the background budget is owner only.

Source: https://ravensight.io/docs/studio-intelligence/
Last updated: 2026-11-15
