Now/Next/Later roadmap

Now/Next/Later replaces the date-driven roadmap’s core lie (that estimates made months before discovery are accurate) with an explicit confidence gradient. Janna Bastow, co-founder of ProdPad, invented the framework around 2012. A customer renamed it from “Current/Near Term/Future” to Now/Next/Later, and the name stuck because it does exactly what it says: sorts work by certainty, not calendar.

The three columns

Now is the only column that carries a genuine delivery commitment. Items here are in active development, scoped, and staffed. If someone asks “what ships this cycle?” the Now column should answer that question directly. Write Now items as outcomes, not features: “Reduce cart abandonment 15%” rather than “Rebuild checkout flow.” This gives engineering room to find the right solution, and it keeps the column honest if the chosen implementation changes mid-build.

Next is validated direction. These are problems with enough evidence to be next in line, but where the solution is not yet locked. Treat Next as a priority signal, not a delivery promise. Sales can know about Next items in general terms; they should not pre-sell them with a date.

Later is a curated set of strategic bets, not a junk drawer. This is the column that teams get wrong most often. Items land in Later because they are real opportunities without enough signal to act on now. They need periodic review: if a Later item has accumulated no new signal across multiple review cycles, it should be culled or explicitly re-parked with a reason. A Later column that never shrinks destroys trust with the people who submitted those items and stops functioning as strategy.

The confidence gradient is the point. The roadmap makes uncertainty visible rather than hiding it behind invented dates. That honesty is its main asset with both your team and your stakeholders.

Worked example

A PM at a B2B analytics tool runs three initiatives. Written as a Now/Next/Later:

Column	Item (outcome-framed)	Why it lives here
Now	Cut time-to-first-insight below 5 minutes for new signups	Discovery done; solution scoped; engineering started
Next	Enable analysts to share dashboards with external stakeholders	Problem validated in three customer interviews; solution still open
Later	Predictive anomaly detection	Technically interesting; no signal yet that customers will pay for it

Notice the Later item is specifically “no signal yet that customers will pay for it,” not “we might do this someday.” That specificity makes it a strategic bet, not a parking lot entry.

Moving items between columns: the gate

Items do not drift between columns on vibes. The gate from Later to Next is a discovery spike: a brief problem brief with evidence (customer quotes, usage data, competitive signal) that justifies elevating the item. Without a gate, Next fills with wishful thinking. From Next to Now requires a scoped solution and a staffed team. Connecting this to RICE or weighted scoring gives you a defensible mechanism for the Later-to-Next promotion specifically.

The stakeholder conversation

Sales wants dates. Executives want a Gantt. This is where the framework gets tested.

The most effective move is to give sales a “Now/Next only” version for customer conversations. Later items are visible internally as strategic direction; they are not a menu sales can order from. When a VP asks for a Q3 date on a Next or Later item, the honest answer is: “I can tell you it is validated enough to be next in line. I cannot give you a Q3 date without first doing discovery, because any date I invent today will be wrong by the time we scope it and I would rather not start a conversation by breaking trust.” Most executives respect this more than a fictional Gantt, once they hear it stated plainly.

Review the roadmap often enough that the Later column reflects current signal, not wishful thinking from a previous cycle.

When it is wrong

Do not force this framework on work with hard external constraints. FDA approval milestones, SOC 2 certification windows, hardware supply chain commitments, and contractual enterprise SLAs all require a milestone roadmap because the date is a real constraint, not an estimate. Using Now/Next/Later for those contexts loses credibility with the stakeholders who actually understand those constraints. The framework earns its place in continuous discovery contexts; it does not belong everywhere.

Interview application

strong

"I use Now/Next/Later when the team is doing continuous discovery and the right solution to a problem is not yet known, which covers most meaningful product work. The practical discipline is treating Now as a real commitment while treating Next and Later as hypotheses about priority, not delivery promises. In practice, I write the Now column as outcomes, so engineering has room to find the right solution. The hardest part is organizational: sales will try to sell Later items as features with dates. My approach is to give sales a Now/Next-only view for customer conversations, and to keep Later visible internally as strategic direction rather than a menu they can order from. Items graduate from Later to Next through a lightweight gate: a discovery spike that produces a problem brief with evidence. If a Later item sits across multiple review cycles without new signal, I cull it or explicitly re-park it with a reason. I do not force it on every context: hard regulatory milestones and hardware supply chain commitments need a milestone roadmap, because the date is real, not estimated."

weak

"We use Now/Next/Later to organize our roadmap. Now is what we're working on, Next is what's coming up, and Later is longer-term stuff. It's better than Gantt charts because we don't have to commit to specific dates, which gives us flexibility." This is a definition recitation. It tells the interviewer nothing about how items move between columns, how you handle a VP demanding a Q3 date on a Later item, or whether you write outcomes or features in the Now column. It is also the exact answer everyone gives after reading one blog post about the framework.

The 2026 case

The argument for Now/Next/Later has only gotten stronger, for a specific reason: AI products add a second layer of uncertainty on top of the ordinary “we don’t know the right solution yet” problem. Model capability, inference cost, and competitive benchmarks shift on roughly six-month cycles. A Q3 date for an AI feature you are scoping in Q1 assumes the capability landscape is frozen. It is not. A model update between scoping and delivery can make your chosen implementation unnecessary or make a previously impossible approach trivial.

Writing the Now column as outcomes rather than features matters even more here. An outcome-framed Now item (“answer 80% of support queries without human escalation”) survives a capability shift and lets the team re-route. A feature-framed one (“add RAG retrieval”) can become irrelevant the moment the base model improves enough to make retrieval optional.

The viable/lovable lens applies too. Viability (is there a paying market large enough to cover costs?) is the harder problem when the underlying capability changes every six months and user expectations reset with it. Later items in an AI product are often bets on a capability curve, not just on a problem. Making that explicit in the Later column, “blocked on model cost reaching X” or “revisit when benchmark Y is hit,” turns the junk drawer into a genuine strategic instrument.