Twelve requests
A call note, a ticket and a Slack thread from the same account count three times.
Discover
Research that knows which decisions rest on it.
Bring interviews, call recordings, surveys, support tickets, CRM notes and Slack threads into one place. Sage turns them into claims, ties each claim to the minute, page or row it came from, and tells you when new research changes a decision you already made.
We put the continuous in continuous discovery.
01 — The problem
The most complete persona had the least evidence behind it.
Solveo's Head of Support persona was 95% complete: four jobs, twelve insights, three opportunities. It had been written from what the sales team heard in demos. Checked against the research, none of its nine claims was backed.
Two of those claims carried the whole go-to-market. Meanwhile the interviews and tickets describing a different person sat in another folder. Research gets done. It just doesn't reach the beliefs and decisions it should change.
02 — A better question
Research is worth what it changes.
Storing and finding research is a solved problem. The question that decides outcomes is the next one: what does this finding change, how sure can we be, and which decisions depend on it?
What a library gives you
What a decision needs
A tag that says a quote is about onboarding.
A claim: admins at companies over 100 seats struggle to set permissions during onboarding.
A count of mentions.
A count of independent customers, by segment.
Search that finds what you remember to look for.
Evidence that disagrees with you, surfaced without asking.
A report filed in March.
A finding that updates when April's interviews say something else.
An insights count on a dashboard.
The decisions each finding supports, and the ones resting on old research.
03 — Open the twelve
What a roadmap item looks like with its research showing.
Say a feature is on the roadmap because twelve customers asked for it. Open the twelve and the picture changes in three moves. This is the conversation a PM walks into when someone in the planning meeting asks “how do you know?”
The team that can answer that question in the room keeps its roadmap. The team that has to go and look loses a week, and usually the argument.
Fewer independent voices
One customer counts once, however many interviews and tickets they appear in.
All from one segment
The claim was written for enterprise. Every piece of evidence came from mid-market.
New interviews disagree
They arrived after the roadmap was set. Sage re-rated the claim and flagged the item that rests on it.
04 — The new way
Research that keeps working after the study ends.
Continuous discovery needs continuous research. Every new interview, ticket, thread or experiment result is read against what the team already believes, and every belief knows which decisions lean on it.
0101
Bring it in
Uploads, recordings and the calls, threads and notes your tools import. The original is always kept.
0202
Read and dated
Recordings are transcribed. Each source is classed, dated and its speakers identified.
0303
Claims and evidence
Specific statements, each anchored to the minute, page, message or row it came from.
0404
Linked and rated
Claims in your opportunities, solutions and roadmap get their state from the evidence.
0505
Rechecked on arrival
New research re-rates the claims it touches. The decisions that depend on them get flagged.
05 — How a claim gets its state
Four labels, the same everywhere.
On artifacts, in Research, in documents and in Sage's answers, every claim carries one of four labels. They describe the evidence, not whether the claim is true.
Supporting evidence from enough independent sources
Opposing evidence from enough independent sources. One solid counterexample can be enough.
Evidence found, but not enough to decide either way
The check ran and found nothing in either direction
06 — The Shorter Loop way
Six things your research can do here.
Raw first
Every citation opens at its source
Open at source lands on the page of the PDF, the timestamp in the recording, the message in the thread or the row in the spreadsheet. The original upload is never changed.
Who said it
Context travels with the quote
Speakers become participants, matched by email or CRM contact and filed under their account. Industry, size, plan and role ride along with every piece of evidence.
Splits
Disagreement between groups becomes a finding
When enterprise evidence backs a claim and mid-market evidence consistently contradicts it, the claim shows “Evidence splits by company size”. Split it into two scoped claims, edit the wording, confirm.
Strength
Reasons you can read
Each claim lists its support mix (“2 interview, 1 survey”), what each piece counts as, its age, sample size and study design. “All from one kind of source” is flagged. No score out of 100.
Gaps
It tells you what you don't know
Customer groups you have fewer than three pieces of evidence from. Claims with no evidence from the segment they name. Research questions still open, oldest first.
Lineage
Decisions remember their research
What rests on this lists every decision, requirement and artifact that depends on a claim. Lineage walks both ways, with each link marked Recorded or Inferred by AI.
07 — Personas
Personas built from the research.
Over a few weeks, 207 feedback items and 21 exit interviews came in. Sage kept finding the same person: a support team lead who runs the queue, switches auto-send off after a wrong answer, and never attends a demo. She came up in 14 of the 21 exit interviews.
Sage drafted her persona from the transcripts and tickets, with every job, pain and gain linked to the customers who said it. Five of the six were backed. One went the other way: the team expected team leads to want the AI on more ticket types. They wanted it on fewer, handled reliably.
08 — What comes in
Bring what you have. Connect what keeps arriving.
Imports are dated to when the conversation happened, not when it synced. Each source is classed as customer evidence, secondary research or the team's own, and you can change the class with one click.
- DocumentsPDF, Word, slides, spreadsheets, CSV, text and HTML
- RecordingsAudio and video, transcribed with timestamps and speakers
- StorageGoogle Drive, OneDrive, Dropbox, Box, Zoom
- ConversationsSlack and Teams threads. Channels shared with customers count as customer evidence
- CRMHubSpot and Salesforce notes and call logs, tied to the contact and account
- Research toolsFireflies.ai transcripts, Dovetail research docs
Coming soon: Gong, Fathom, Grain, Zoom calls, Delighted, Maze, SurveyMonkey, Typeform, UserTesting.
09 — Health
Open Research and see what needs you this week.
There is no insights counter. The landing page lists only what needs a person, and says “Research looks healthy” when nothing does.
- Ideas based on untested assumptions
- Important decisions with weak evidence
- Claims where the evidence conflicts
- Findings that differ between customer groups
- Customer groups you know too little about
- Research questions still unanswered
- Decisions based on outdated research
- Findings that changed recently
- Research records missing key details
The research challenge
Ask one question of your research. Ask it twice.
Pick your top roadmap item and ask: is the case for it still sound? Ask Shorter Loop. Then ask whatever you use today: your research repository, or Claude or ChatGPT with the same files uploaded.
Every good tool will find relevant quotes. Read both answers against the four questions and see which one you would take into a planning meeting.
- Did it count customers or mentions?One account in a call, a ticket and a Slack thread is one voice.
- Did it bring up evidence against the item?Retrieval tends to return what sounds like the question, which is usually the supporting side.
- Did it tell you what changed since you decided?That needs the decision, its date and the evidence that arrived later.
- Did it name who you haven't heard from?A gap can't be retrieved. It has to be worked out from what is there.
Customer story · Solveo · after the plan
The plan was set in February. The research kept reading.
Solveo, a B2B customer-support platform, committed to auto-send by complexity on the strength of churn interviews, CRM losses and support data. The plan was done. The research wasn't.
[What the team did: for example, split the claim, scoped the rollout to the segment that backed it, and opened a research question for the one that didn't.] Nobody had to remember that the February research existed, or go looking for what had changed since.
Auto-send by complexity cuts reopens
Evidence from [N] independent customers, all gathered before [month].
[N] interviews and [N] tickets arrive
From [Slack Connect channels / Fireflies.ai / HubSpot], dated to the conversation, with speakers matched to their accounts.
[Segment A] backs it. [Segment B] consistently doesn't.
[One line on what segment B customers reported, from the workspace.]
[N] roadmap items rest on the claim
Listed on Health under findings that changed recently, and on each item under what rests on this.
Get Started
Which decision are you making on research nobody has reread?
Bring the research you already have. Sage shows you what it supports, what it contradicts and what it never covered.