Prioritize: Confidence-Scored Roadmap + Goals

prioritize: lead guest team

Rank what to build by evidence instead of opinion, shape it into a roadmap that survives contact with reality, and set goals the whole team can recite. Built from 324 insights from 133 Lenny's Podcast guests. Lead team: Itamar Gilad, Christina Wodtke, Janna Bastow.

When to use

• A backlog or idea list is bigger than capacity and opinions are fighting (founder, sales, competitors, research).
• You need a roadmap and stakeholders keep asking for dates.
• You are writing or rescuing OKRs / quarterly goals, or the team cannot name its top goal.
• Planning takes weeks and the plan is dead by week three.
• You cannot choose between two features of similar size.
• AI model capabilities keep shifting and a 6-12 month roadmap feels like fiction.
• You must split capacity between core work, tech debt, and bets.

Step 1 - Diagnose (ask before answering)

If the user attached a backlog, roadmap, OKR doc or planning deck, read it first and infer as much as you can. Then ask at most 4 of these:

1. What is the goal for the next quarter, in one sentence with a number? No clear goal means no prioritization is possible (Matt LeMay, Ravi Mehta). If there is none, run the Goals play first.
2. Stage and size? Under ~50 people, seed stage: short cycles (Molly Graham, Gokul Rajaram). Scale-up or enterprise: allocation splits and cross-team forced ranking.
3. Are there hard external dates or enterprise customers who demand a roadmap? Routes you to dated project plans or the 80/50 rolling roadmap instead of pure Now-Next-Later.
4. Is your product exposed to model-capability changes (AI)? Routes you to short-horizon planning (JIT monthly, seasons) versus quarterly OKRs.

Step 2 - Pick the play

If... (situation)	Use	From	Why
Too many ideas, opinion battles	Confidence-scored ICE + Confidence Meter	Itamar Gilad (ICE originated with Sean Ellis)	Forces evidence into the score; stops gut feeling from posing as confidence
You have one numeric team goal	Impact in the same unit as the goal	Matt LeMay	Abstract scores lose the line to the goal
Stakeholders want dates, uncertainty is high	Now-Next-Later + outcome roadmap	Janna Bastow, Itamar Gilad	Granularity falls as certainty falls; dates only where required
Enterprise customers demand a roadmap review	Rolling 6-month roadmap, 80% / 50% commitment	Varun Parmar (Miro)	Satisfies buyers while keeping room to pivot
AI product, capability moves monthly	Work backwards from capability + JIT monthly plan	Nick Turley, Fiona Fung, Asha Sharma	Long roadmaps go stale in weeks
Writing or fixing OKRs	Objective + 'how do we know?' + triangulated KRs	Christina Wodtke	Outcome-based KRs, 3 max, weekly rhythm
You do not know how to move the metric	Frontier-of-understanding goals + execute/understand tracks	Ravi Mehta, Bangaly Kaba	Learning goals beat fake outcome commitments
Core vs debt vs bets fight for capacity	Allocation split (70/20/10, 60-30-10, 5-25-60, Four BBs)	Ken Norton, Shreyas Doshi, Ryan Salva, Anuj Rathi	Makes the trade-off a strategy call, not a per-ticket fight
Sales or execs inject requests mid-quarter	Strategy as escalation filter + sales feature budget	Shreyas Doshi, Jason Lemkin	Requests are judged against strategy or a capped budget
Teams chase the same top-line metric	Distinct goals one step from company goal	Ami Vora, Matt LeMay	Avoids toddler soccer and cascades that do not add up

Step 3 - Run it

Play A: Confidence-scored backlog (Gilad, Ellis)

1. Fix the goal first. Write the top 1-3 goals with numbers. Anything that cannot be tied to one goes in an 'opportunity backlog' (Nan Yu keeps 20-30 problems not yet ready to build).
2. Collect ideas from everyone (founders, sales, competitors, research) but let the team choose which to test first (Gilad). Anyone proposes; the team scores.
3. Score each idea: Impact on the goal (best case, ideally in the goal's unit), Ease (inverse of effort), then Confidence in those two guesses. Fold Reach into Impact: Gilad and Sean Ellis both prefer ICE to RICE.
4. Cap confidence by evidence type (adapted from the Confidence Meter; Gilad's actual meter is a thermometer, this is a working simplification):
   • Own conviction, pitch deck, 'fits our strategy/AI theme': 0 to 0.1 out of 10, whatever the person feels.
   • Peer review, back-of-envelope estimates: still guesswork, barely above zero.
   • Anecdotal evidence, competitor has it, market data, surveys: low to medium.
   • Real tests with users (prototype, fake door, beta): medium to high. Only building and testing earns high confidence.
5. Adjust after small tests. Showed 12 customers, one liked it? Cut the Impact score, do not just lower Confidence (Gilad).
6. Match investment to confidence. Cheap tests first; skip heavy validation for low-risk changes like reordering settings (Gilad: know when to stop).
7. Protect big bets from the RICE trap. For high-reach, high-impact ideas, ignore C and E for about a week, sit with engineers and designers trying to solve it, then re-score (Vijay Iyengar).
8. Willingness-to-pay check (optional, B2B/monetized). Show 10 candidate features in subsets of 6 and ask most important (would pay) and least important (won't pay); rotate subsets (Madhavan Ramanujam). Roughly 20% of what you build drives 80% of WTP.
9. Done when: every top idea has a score, an evidence label, and a next validation step with an owner. Put those steps on a GIST board and review every other week (Gilad).

Play B: Roadmap that does not lie

1. Draw three columns: Now, Next, Later. Place problems or outcomes, not features. Be less granular the further out (Bastow).
2. Put a date only on items with a real external reason (regulation, Christmas, school year). For those, plan earlier with buffer and aim to finish early enough for a soft launch (Bastow).
3. Write items as outcomes: by October we want to achieve this outcome. Move a concrete build onto the roadmap only after a high-confidence test (Gilad).
4. Add a line stating items may be pivoted or cut (Paige Costello); items on a roadmap otherwise harden into promises.
5. Enterprise variant: rolling six months, updated every three; commit about 80% of months 1-3 and about 50% of months 4-6 (Varun Parmar).
6. Story test: write the roadmap as themes with a why for each. Jiaona Zhang wants themes and a story, not a scored spreadsheet.
7. Strategy drift test: lay the roadmap against your top three strategic bets and count the share of resources committed to off-strategy items (Maggie Crowley found cases with 90%).
8. Done when: the Later column is problems, every date has a named external reason, and you can say which bet each Now item serves.

Play C: Goals / OKRs (Wodtke + LeMay + Graham)

1. Check foundations: strategy, empowered teams, psychological safety. OKRs are a vitamin, not a medicine (Wodtke).
2. Write one inspiring objective for the quarter, motivating but not ridiculous.
3. Ask how do we know? List evidence, then convert to three key results that triangulate: one hard number, one quality or delight measure, one tied to money (Wodtke). Brainstorm measures for 10 minutes before choosing.
4. Test every KR: outcome or task? Rewrite tasks. A binary gate is acceptable only if it is genuinely hard to pass (Wodtke).
5. Keep the team goal one step from the company goal: link them with one Y statement or one operator, e.g. upgraded users times lifetime value (LeMay). If the intern-who-started-Monday cannot understand it, it fails (Molly Graham).
6. One goal, one named owner, and say which goal wins in a fight (Molly Graham). Max three company goals.
7. Set the target so it feels uncomfortable but not doomed; expect roughly 70% attainment (Wodtke). All-green means you sandbagged (Jiaona Zhang).
8. Leave next-quarter KRs blank until this quarter's results are in; write only objectives (Wodtke).
9. Cadence: Monday commit, Friday celebrate; weekly status with confidence per KR, last week, next week, max three P1s. Review top 2-3 initiatives per KR, about 10 minutes (Wodtke).
10. Planning time cap: no more than about 10% of the execution period (Lane Shackleton, Gustav Soderstrom): one week of planning in a ten-week quarter.
11. End of quarter: grade loosely, then retro deeply. Ask why you landed there; if you hit, check you know why (Wodtke, Ravi Mehta).
12. Done when: an engineer stopped in the hallway can recite the team goal (Yuhki Yamashita's test).

Play D: When you cannot move the metric (Ravi Mehta, Bangaly Kaba)

1. Name the risk: understanding (levers unknown), dependency (missing tools), execution (cannot ship hypotheses fast), strategic (hypothesis may be wrong).
2. Set a goal that matches: understanding goal, a tooling goal, a count of experiments (e.g. 20), or a test of the hypothesis.
3. If leadership insists on an outcome, commit to it but stage the quarter: weeks of customer talks and analysis, then hypotheses, then execution. A miss then comes with a why and a higher-confidence next plan.
4. Run execute + understand in parallel each sprint: high-conviction items plus 3-4 understand projects. Early in a new domain Kaba started near 60% execution / 40% understand and shifted toward 80-85% execution.
5. For each theme ask: what do we know from data, and what must we first understand?

Play E: Capacity split

1. Pick a template for your stage, then adjust:
   • 70/20/10 core / adjacent or strategic / bets (Ken Norton from Google, Eeke de Milliano; Ronny Kohavi: expect about 80% of big bets to fail).
   • 60-30-10 incrementals / one or two big initiatives / stability and infrastructure (Shreyas Doshi).
   • 5-25-60 bold bets / operations / payoff from earlier bets (Ryan Salva; not for startups).
   • Four BBs: Brilliant Basics, Bread and Butter, Big Bets, Breaking Bad. Ask for allocation of 100 points, write three alternatives with consequences, and make it a CEO + head-of-product strategy call (Anuj Rathi).
2. Growth teams: reserve 20-25% of annual time for new loops, with no metric goals at first (Elena Verna).
3. Do not quantify opportunity cost in a spreadsheet column. Name the ambiguous big bets you are tempted to skip and give them explicit room (Doshi).
4. Done when: the split is written down, one named owner defends the bets slice, and the allocation is revisited at the next planning cycle.

Where the experts disagree

1. Score ideas or do not. Score (Gilad, Sean Ellis): scores let you explain to submitters why an idea lost. Skip abstract scores (Matt LeMay, Jiaona Zhang, Vijay Iyengar): LeMay says use the goal's unit; Zhang wants a story of themes; Iyengar shows RICE buries bold bets. Use scoring when many people submit ideas; use goal-unit estimates when you have one numeric goal. Default: ICE with Impact in the goal's unit, Confidence capped by evidence, and the ignore-C-and-E exception for big bets.

2. OKRs or no OKRs. Use them (Christina Wodtke, Nickey Skarstad, Ray Cao): a shared cross-functional goal framework brings strategy down to the quarter. Drop them (Jason Fried, Tobi Lutke, Geoff Charles, Keith Coleman, Archie Abrams): planning cost more than the work, and core product runs on taste. Wodtke herself says OKRs will not fix a broken company. Use OKRs when several functions must align and the team has a strategy; use a one-pager or milestone goals for small teams with short feedback loops. Default: a light OKR set (one objective, three KRs, weekly rhythm), capped at 10% planning time.

3. How far ahead to plan. Rolling plans (Bastow, Brian Chesky's two-year rolling roadmap, Paige Costello's rolling 12 months, Varun Parmar's six months): customers and go-to-market need predictability. Near-zero roadmap (Amjad Masad, Kevin Weil, Keith Rabois, Fiona Fung): model capability changes the possible set every month. Use rolling plans when partners or enterprise buyers depend on you; use JIT monthly planning with a themes refresh every six months when capability drives the roadmap. Default: Now-Next-Later with detail inversely proportional to horizon (Andrew Ambrosino).

4. Fixed allocation or no quota. Fixed splits (Norton, Doshi, Salva, Eeke de Milliano) make bets and maintenance survive planning. No quota (David Singleton, Katie Dill at Stripe): each team sizes polish and incident time itself, backed by judgment and quality culture. Use splits when the team is large or short-termism wins every fight; skip them when talent bar and trust are high. Default: a written split with a bets slice that is explicitly protected.

Deliverable

Produce this, filled in, as a pasteable doc:

# Priorities and Goals - <team>, <quarter>

## 1. Goals
- Objective: <inspiring, one sentence>
- KR1 (hard number): ...   KR2 (quality): ...   KR3 (money): ...
- Goal link: team goal -> company goal in one statement: <...>
- Owner (one name): <...>   Wins in a fight: <goal>

## 2. Capacity split
| Bucket | % | Why | Owner |

## 3. Scored ideas
| Idea | Impact (goal units) | Ease | Evidence type | Confidence (0-10) | Next validation step | Owner |

## 4. Roadmap (Now / Next / Later)
Now: <outcomes + date only if externally required>
Next: <problems>
Later: <struggles/problems, not features>
Pivot-or-cut note: <one line>

## 5. Not doing (and why)
- <idea> - fails <persona/strategy/confidence> filter

## 6. Rhythm
Weekly: Monday commit / Friday celebrate; status = KR confidence, last week, next week, max 3 P1s
Biweekly: GIST-board review of ideas and validation steps
Quarter end: loose grade + retro; next-quarter objectives only, KRs after results
Planning time cap: <10% of cycle>

## Next 3 actions
1. <who, what, by when>
2. ...
3. ...

Grade existing work

Score each 1 / 3 / 5, total out of 40. Return the score per criterion, the total, and the top 3 fixes with the guest behind each.

Criterion	1	3	5
Goal clarity (LeMay, Graham)	No stated goal, or 10+ goals	Goals exist but not numeric or not one step from company goal	1-3 numeric goals, one owner each, one wins in a fight
Evidence-backed confidence (Gilad)	Confidence = how strongly people feel	Some tests, scores unlabeled	Every score has an evidence type; tests lift confidence
Outcome orientation (Wodtke, Perri)	KRs are tasks or activity goals	Mix of outcomes and tasks	All KRs are outcomes; triangulated hard/quality/money
Roadmap honesty (Bastow, Gilad)	Dated Gantt of features	Themes but dates everywhere	Now-Next-Later, dates only where external, pivot-or-cut stated
Focus (Daniel Lereya, Ebi Atawodi)	Cannot name the most meaningful thing shipped last quarter	5+ priorities	3-5 big rocks; list of not-doing exists
Portfolio balance (Norton, Doshi)	All quick wins, or all bets	Mix exists, unwritten	Written split with protected bets and debt
Strategy fit (Crowley, Doshi)	Roadmap bears no relation to strategy	Partial	Every item maps to a bet; escalations filtered through strategy
Cadence and cost (Shackleton, Wodtke)	Planning eats a month; no weekly rhythm	Cadence exists, heavy	Under 10% of cycle; weekly commit; loose grade + retro

Red flags

• Confidence by conviction. Every terrible idea had someone who thought it was great (Gilad).
• Competitor has it, so validated. Gilad: they do not know better than you.
• Tasks as key results. Wodtke calls this the biggest OKR mistake.
• Goals layer used as a build plan. Gilad: that is planning work, not goals.
• Toddler soccer. Every team gets the same top-line metric and runs to the same surface (Ami Vora).
• Five safe wins, zero ambiguous bets. Positive-ROI logic keeps teams busy (Shreyas Doshi).
• Peanut-buttering resources across many problems (Ebi Atawodi).
• Rushed OKR rollout delegated to HR; leaders skim the book then declare OKRs broken (Wodtke).
• Cannot choose between two equal features. Usually a missing strategy, not a scoring problem (Ravi Mehta).
• Year-long sequential roadmap in AI products (Keith Rabois) and promising dates to customers (Jason Fried).
• All-green OKRs mean sandbagging, and OKRs hit while users notice nothing different is the worst outcome (Jiaona Zhang).

Receipts

• "Guess what? Behind every terrible idea that was ever someone thought it was great, that gives you 0.01 out of 10." — Itamar Gilad, Lenny's Podcast (00:38:47)
• "The further out you plan, the more you're making it up. We know this." — Janna Bastow, Lenny's Podcast (25:08)
• "So, making sure that you have real outcomes that let you move forward, I think that's the biggest mistake people make in OKRs." — Christina Wodtke, Lenny's Podcast (00:25:09)
• "if you don't have a clear and specific sense of what impact means, you can't really prioritize effectively." — Matt LeMay, Lenny's Podcast (01:02:26)
• "we convince ourselves that positive ROI is great. And so we make ourselves busy" — Shreyas Doshi, Lenny's Podcast (01:07:11)
• "two people owning a goal is no one owning a goal. One person owns the goal, who is it?" — Molly Graham, Lenny's Podcast (00:52:09)

Go deeper

• references/frameworks.md - every framework in this skill, how to run it
• references/quotes.md - verified quotes with timestamps
• Related skills:
  • strategy - when you cannot choose between options because there is no strategy yet.
  • metrics - when goals need a north star and a metric tree to anchor KRs.
  • experiment - when a top-ranked idea needs a trustworthy test to raise its confidence.
  • ai-product-bets - when the question is which AI features to build as models improve.
  • price - when willingness to pay should decide the roadmap order.