Every product team runs customer interviews. Very few of them connect those interviews to the roadmap in any structured way. The transcripts get saved to a folder. Quotes get pasted into a deck for the next quarterly review. The deeper patterns surface in conversation but rarely make it into a written artifact that influences what gets built.
The teams whose discovery actually shapes the roadmap have a pipeline: a defined sequence of activities, artifacts, and handoffs that move raw user research into product decisions. This article distills that pipeline, building on the eight insight report types Aha! introduced in their discovery framework and adding the workflow that makes them ship work, not shelf work.
Why discovery usually does not influence the roadmap
Four failure points account for almost all the gap:
- Interviews are stored, not summarized. An hour-long recording is not actionable. Nobody re-listens.
- Summaries are shared, not integrated. A Notion doc full of quotes is not a roadmap input. It sits separately.
- Patterns are noticed, not documented. The researcher saw the pattern. The PM did not. The pattern dies in the researcher's head.
- Decisions are made without referencing discovery. The roadmap meeting happens. The discovery doc is open in another tab. Nobody pulls from it explicitly.
The fix is not better interviews. It is a structured pipeline from raw conversation to roadmap commitment.
The 5-stage discovery pipeline
Stage 1: Conversations (weekly)
Every week, the team runs 2-3 customer interviews. Sources rotate: power users, churned users, prospects, support contacts. Each interview follows a loose script focused on jobs-to-be-done, not features. Recording with consent, transcript generated automatically.
Stage 2: Raw extracts (within 48 hours of each interview)
Within two days of the interview, the researcher extracts the raw signal: pain points named, language used, current workarounds, attempted solutions, expressed desires. Stored in a shared system, tagged by interviewee and date. This is the source data layer.
Stage 3: Synthesized insights (monthly)
Once a month, the researcher synthesizes the previous month's extracts into eight standard report formats. Aha! Discovery offers these as templates; you can build them in any tool:
- Executive summary: top 3 themes for the month, 1 page
- Jobs-to-be-done report: the recurring jobs users are hiring the product for
- Journey map: the current path users take, with friction points marked
- Opportunity solution tree: opportunity → solution candidates → experiments
- Persona analysis: updates to existing personas based on new evidence
- Problem validation: which hypothesized problems were confirmed, which were not
- Product messaging: language users actually use vs language the company uses
- Voice of customer: direct quotes selected for representativeness
Each format serves a different audience. Executive summaries go to leadership. JTBD goes to PMs. Journey maps go to design. Voice of customer goes to marketing.
Stage 4: Roadmap inputs (quarterly)
Once a quarter, the synthesized insights from the last three months get translated into specific roadmap candidates. The translation is explicit: each candidate links back to the insights that motivated it. "Build a guided onboarding" links to the JTBD report and the journey map that showed friction at step 4.
The discipline: no roadmap candidate enters the consideration set without a discovery citation. This single rule prevents 60-70% of unjustified features from making it into the roadmap.
Stage 5: Decision (quarterly)
The roadmap candidates compete against each other based on evidence strength, expected impact, and strategic fit. The winners go on the roadmap with explicit links back to the source insights. The losers go back to the consideration set or to the graveyard with documented reasoning.
The weekly handoff
The pipeline only works with handoffs:
| Cadence | Activity | Output | Owner |
|---|---|---|---|
| Weekly | 2-3 interviews + extracts | Tagged raw signal | Researcher or PM |
| Monthly | Synthesis of last month | Up to 8 insight reports | Researcher |
| Monthly | Reading session with team | Shared understanding | PM + design + eng lead |
| Quarterly | Translate insights to candidates | Roadmap candidate list | PM |
| Quarterly | Roadmap decisions | Committed roadmap | PM + leadership |
The monthly reading session is the most underrated step. Without it, the insight reports become documents nobody reads. With it, the team builds shared context that lasts through the quarterly decision meeting.
What the pipeline produces over a year
A team running this discipline for 12 months ends up with:
- ~120 customer interviews documented
- ~12 monthly synthesis cycles producing 60-80 insight reports total
- 4 quarterly translation cycles producing 20-30 roadmap candidates
- ~12-15 candidates that actually ship
- A defensible record of why each shipped feature exists
The defensibility is the unlock. When a stakeholder asks "why did we build X?", the answer is a chain: interview → insight report → candidate → decision. The answer is no longer "we thought it was a good idea."
Tools that fit the pipeline
The pipeline can run in any stack, but specific tool categories save time:
- Recording + transcription: a tool that handles consent, recording, and AI transcription (Tella, Loom, Otter)
- Insight repository: a tagged store for raw extracts (Notion, Coda, dedicated tools like Dovetail)
- Synthesis templates: structured templates for the 8 report types (the Aha! Discovery Advanced suite, or build your own)
- Roadmap with citations: Roaderly, Productboard, or any roadmap tool that supports linking items to source documents
Common pipeline failures
- Skipping the monthly reading session. Without it, reports become invisible. Hold it religiously.
- Letting interview volume drop. 2-3 per week is the floor. Below that, the synthesis runs dry.
- Synthesizing only when ready. Monthly is the cadence. If the synthesizer is too busy, reduce report scope but keep the cadence.
- Translating without strict citation. The rule "every candidate cites discovery" is what protects against feature creep.
The takeaway
Discovery work that does not connect to the roadmap is theater. The 5-stage pipeline (conversations, raw extracts, synthesized insights, roadmap inputs, decisions) with disciplined cadence and explicit citations is what converts research into shipped features. Most teams have the interview part. Few have the pipeline. The teams that build the pipeline ship features that actually serve users, instead of features that just sound smart.
