Vision plus three horizons — Now, Next, Later — with success metrics and the dependencies that gate each.
Layout
Product Roadmap
6 blocks · 2-row grid
"A roadmap is a sequence of bets. Without metrics, it's just a list."
What it is
A horizon-based plan. Now is what the team is shipping this quarter, with named owners. Next is what's queued, with preconditions. Later is themes (not features) you intend to invest in, with the caveat that they may move. Vision states where it's all going; Success Metrics tell you whether each horizon worked.
Origin
Roadmaps as a structured artefact came out of product-management practice in the 2000s when teams started sharing forward plans publicly with customers. The Now/Next/Later format crystallised the discipline: replace dated Gantt charts (which were always wrong) with a confidence gradient that's honest about uncertainty further out.
When to reach for it
You're onboarding a new product hire and they need to understand priorities in 20 minutes.
You're publishing externally to customers — they want to see what's coming without being lied to.
You're defending product spend in a board meeting and need a forward-looking artefact.
The blocks
Worked example uses Linear (2023 public roadmap).
What good looks like
The customer impact 18–24 months out, in plain language. What the product will not be, too.
Example — Linear (2023 public roadmap)
Become the default project tracker for ambitious software teams — the one that doesn't make you fight the tool. We will not become a wiki, doc tool, or all-in-one platform.
What good looks like
2–3 metrics per horizon with targets and instrumentation. Leading vs lagging labelled.
Example — Linear (2023 public roadmap)
Activation rate (workspace creates ≥1 cycle in week 1) ≥65%. Weekly retention M3 ≥85%. NPS ≥60. ARR growth 8% MoM.
What good looks like
Ranked list of 3–5 bets for this quarter. Owner, metric, and status (on track / at risk).
Example — Linear (2023 public roadmap)
Customer requests: bulk edit shortcuts (Tom). Cycles: forecasted completion line (Maria). Integrations: GitHub deeper PR ↔ issue linking (Akira). All ship this quarter.
What good looks like
3–5 candidates for next quarter with preconditions and open decisions. Conditional, not committed.
Example — Linear (2023 public roadmap)
Mobile rebuild for native gestures (precondition: design lead in seat). Slack-first triage workflow (precondition: API throttle for high-volume users). AI summary for retros (decision: own model vs. partner).
What good looks like
Themes (not features) for 2+ quarters out. Why each, what would invalidate it, "may change" framing.
Example — Linear (2023 public roadmap)
Theme: cross-product navigation for orgs with 10+ projects. Theme: native goal-tracking for OKR teams. Theme: writing tools (docs) — explicitly flagged as may-not-happen.
What good looks like
Outside teams, vendors, regulations, hires that gate the plan. Mitigation per dependency.
Example — Linear (2023 public roadmap)
Mobile depends on design hire (offer out). API depends on infra migration finishing Q2. AI features depend on EU compliance review before EEA rollout.
How to use it
Vision first. Without a vision, you're sequencing features, not bets.
Now is dated and owned. Next is conditional. Later is thematic. Stop pretending Later is committed.
Each horizon gets a metric. "Ship X" without a metric is theatre.
Update quarterly. A roadmap that's a year old isn't wrong — it's a different document.
Common mistakes
Dating Later items. If a feature has a date 9 months out, it's a fantasy.
Roadmaps with no metric attached — you can't tell whether the quarter worked.
Filling Now with 15 items. That's a backlog, not a roadmap.
Spin up the canvas in one click. Copilot will score every cell against the same rubric this guide describes.
Keep reading