Quarterly Planning Without the Two-Day Offsite

The two-day quarterly planning offsite has a strange property: almost everyone who attends one agrees it was exhausting, that the real decisions happened in the last ninety minutes, and that the resulting plan started drifting within three weeks — and yet the same group books another one next quarter. The format survives because it feels proportionate. Planning three months of work seems like it should take days of everyone in a room. It doesn't. The days are mostly spent doing things that are worse in a room: reading context, generating options, and wordsmithing — all of which are async work wearing a meeting costume.

Here's the replacement, tested shape: two weeks of structured async preparation, one 90-minute synchronous decision session, and three named output artifacts. Total synchronous cost for a ten-person team: 15 person-hours instead of 160. The plan quality goes up, not down, because decisions get made by people who read the material instead of people who are on hour eleven of a conference room. This post is the full operating manual — the timeline, the templates, the session script, and the communication step most teams skip.

Why the offsite format fails on its own terms

It's worth being precise about the failure, because the async replacement is engineered against each piece of it.

Thinking in real time is the worst kind of thinking. Offsites ask people to absorb a quarter's worth of context — metrics, customer feedback, engineering constraints, last quarter's misses — from slides, live, and then immediately reason about it. Nobody reasons well that way. The people who "win" offsite debates are the fast talkers and the well-rested, not necessarily the well-informed. Written proposals read on your own schedule flip the advantage back to preparation.

Energy is a budget, and the format spends it on the wrong things. Hour two of an offsite is sharp. Hour eleven — where the prioritization calls usually land, because the mornings went to "context setting" — is sunk-cost negotiation between tired people who want to go home. The async model spends synchronous energy only on the one thing that genuinely needs it: contested decisions.

One person talks, eleven wait. A room of twelve has one conversational thread. Async has twelve. Two days of serialized discussion moves less actual information than two weeks of parallel written argument, and the written version leaves a record.

The plan exists only in the room. The offsite's output is typically a photo of a whiteboard and a deck that made sense if you were there. Three weeks later, disputes about "what we agreed" get resolved by memory and rank. The async model produces the written plan as the process, not as cleanup afterward.

None of this means never gather. If your team meets in person twice a year, spend that scarce time on the things co-location is actually good for — trust, conflict repair, ambition, celebration — and run planning through the process below. Don't spend your only shared days watching each other read.

The shape of the ritual

The full cycle runs about three weeks, starting roughly four weeks before the quarter begins:

Week	Phase	Who	Effort
1	Inputs assembled and published	Lead + contributors	2–4 hours total
2	Proposals written and reviewed async	Anyone proposing; everyone commenting	1–3 hours per person, self-scheduled
3, early	The 90-minute decision session	Decision group	90 minutes, once
3, late	Artifacts finalized, plan communicated	Lead	2–3 hours

Notice what's absent: no all-day anything, no travel, and no phase where twelve people watch a presentation they could have read.

Week 1: Assemble the inputs

Planning quality is capped by input quality, and inputs are pure async work. The lead publishes a single inputs page — one document, linked from everywhere the process happens — containing four things:

1. The scoreboard from last quarter. Each of last quarter's commitments with its final status: shipped, partial, dropped — and a one-line honest note. This is the step most planning skips, and skipping it is how teams commit to 12 things every quarter while finishing 6, forever. Your realized capacity is the single best predictor of next quarter's capacity. If your Project Status or goals tracking is current, this section writes itself in twenty minutes; if assembling it takes days, that's a finding about your tracking, and worth fixing regardless — see Goals That Stick: OKRs, RAG Status and Honest Reporting.
2. Constraints, stated as facts. Headcount changes, committed dates that pre-empt debate (the compliance deadline, the conference launch), budget realities, planned leave. Constraints published up front prevent the classic offsite waste: forty-five minutes designing an option that dies the moment someone mentions the thing everyone in finance already knew.
3. Strategic context, one page maximum. What the company or department needs from this team this quarter, from whoever owns that message. If this doesn't exist in writing, extracting it is the lead's week-one job — a planning cycle without it produces locally sensible, globally misaligned plans.
4. The demand pile. Customer asks, debt, carryover, ideas — linked, not curated. Curation is what proposals are for.

Publish the inputs page at the start of week one and announce the timeline with it: "Proposals due Friday. Comment window through Wednesday. Decision session Thursday at 10."

Week 2: Proposals and written argument

Anyone who wants something on the quarter's plan writes a one-page proposal. Fixed template, hard length limit — one page forces the thinking that slides let you skip:

Proposal: [name] What: Two or three sentences a new hire could understand. Why now: The cost of doing this next quarter instead. If there isn't one, say so — that's useful information against it. Rough size: T-shirt (S = person-days, M = person-weeks, L = a person-quarter or more). Not estimates — magnitudes. What it displaces: Given the scoreboard, what would you not do to make room? Every proposer answers this. It converts advocates into planners. Done in 13 weeks looks like: One observable sentence. Confidence and risks: What would make this fail or double in size?

Proposals go where the whole team can read and comment — a shared docs area with one page per proposal works; Openbook teams typically use a Docs room with a page tree for the cycle (inputs page at the top, proposals nested under it) so the argument and its record live in one place.

Then the comment window: three to four working days where everyone is expected to read everything and comment where they have signal. Rules that keep this from becoming thread soup:

• Comment deadline is real. Silence after Wednesday is consent to proceed. Deadlines-with-consent is the core move of async decision-making generally; the full pattern — including how to handle the person who surfaces objections late — is in How to Make Decisions Asynchronously.
• Proposers revise in place. A proposal that absorbs its objections before the session is the process working perfectly — many decisions resolve here, which is exactly why the live session only needs 90 minutes.
• Disagreements get sharpened, not settled. When a thread hits real contention, the proposer's job is to write the disagreement as a clean either/or with the tradeoff stated, and flag it for the session. The async phase's output isn't consensus; it's well-defined conflicts.
• The lead spends this week as an editor: merging overlapping proposals, prodding silent stakeholders ("Ops hasn't commented on the migration proposal and it lands on you — please read it by Wednesday"), and keeping threads on the page rather than in DMs, where their reasoning would be lost to everyone else.

By the end of week two you have something no offsite ever starts with: every option written down, argued, sized, and either agreed or crisply contested.

The 90-minute decision session

One session, video or in person, with the decision group — usually the team plus its immediate stakeholders; keep it under twelve. Everyone has read everything; the session begins from that assumption and never re-presents material. The agenda, timed:

0:00–0:10 — Capacity math. The lead proposes the quarter's realistic capacity, derived from the scoreboard, out loud: "We planned 12 last quarter and finished 7. We have one departure and a two-week freeze in March. I'm calling our budget six medium items or equivalent." The room adjusts the number if the reasoning is wrong — but from here on, the budget is fixed, and everything else is spending it. This ten minutes is what makes the next sixty tractable: prioritization without a capacity number is a wish-listing session.

0:10–0:25 — Uncontested items. The lead reads the list of proposals that survived the comment window without open conflict and fit obvious priorities. Sixty seconds of "any objection?" each. Typically half the plan locks here — this speed is the dividend from week two, and it's the part that used to cost an offsite its entire first day.

0:25–1:15 — The contested few. The heart of the session: the three to five sharpened conflicts, ten minutes each, hard timeboxed. Format per conflict: the proposer states the either/or in one minute; the strongest written objector (named in advance, so they prepare) gets two; open discussion; then the named decider decides in the room. Two rules keep this honest: no new information may be introduced ("if it wasn't written by Wednesday, it doesn't exist today" — this rule is what makes people actually write things down next cycle), and every decision lands as one of exactly three outcomes — in, out, or cut down (a smaller version is in). "Let's discuss further offline" is banned; an undecided item is out, and may be proposed again next quarter.

1:15–1:30 — Read the plan back. The lead reads the full list — in items with owners, explicit outs, and anything deliberately left as slack — and asks the closing question, person by person if the group is small enough: "Can you commit to this being our quarter, including the outs?" Reservations voiced now get sixty seconds and a decision; reservations voiced in week five get pointed back to this moment. That's disagree-and-commit, made explicit and dated.

If the session runs long, it's nearly always because week two was skipped or skimmed — the session inherits unsharpened conflicts and turns back into an offsite. The fix is upstream, not a longer meeting.

The three output artifacts

The ritual isn't done when the meeting ends; it's done when three artifacts exist. Each has a distinct audience, which is why one deck can't do the job:

1. The plan of record — for the team. One page: the quarter's commitments, each with a single owner, a size, and its "done looks like" sentence; the explicit not-doing list; and the capacity math that shaped it. The not-doing list is half the document's value — it's what the team points to when quarter-week-six requests arrive: "That's on the outs list; it can trade in, but something trades out." If commitments have real dates and dependencies, this is also where they get sequenced into milestones — Milestone Planning: Turning a Roadmap into Commitments covers that translation.

2. The decision log — for the future. One paragraph per contested call: the options, the tradeoff, who decided, and why. Audience: the person in week nine asking "why didn't we do X?", the new hire in month four, and honestly, the team itself next January. Ten minutes to write while memory is fresh; hours of relitigating saved. The comment threads from week two serve as the extended record behind it.

3. The broadcast — for everyone else. Half a page, written for people who will spend ninety seconds on it: what the team is doing this quarter and why in three or four bullets, the one or two most relevant not-doings, and where the full plan lives. Post it where the org reads — the team's feed, the internal newsletter — and have the lead deliver the same summary live wherever the team's work gets discussed. The rule of thumb from internal comms applies: by the time you're tired of repeating the plan, adjacent teams are just starting to absorb it.

Keeping the plan alive for thirteen weeks

A plan communicated once decays. Two lightweight mechanisms keep it real without adding meetings:

Wire the plan into the weekly cadence. Each commitment gets a status line — on track / at risk / off track — updated weekly by its owner in whatever async status ritual the team already runs. The plan of record shouldn't be a document that gets exhumed in week thirteen; it should be the checklist the team walks every week. In Openbook this is a Project Status room seeded with the quarter's goals, so the plan and its weekly RAG history live at the same address the team already checks.

A mid-quarter checkpoint, async by default. Around week seven, the lead posts a scoreboard-so-far against the plan: anything at risk, anything that should be consciously re-traded. If a real trade is needed ("the migration doubled; I propose we drop the reporting item"), it runs as a mini version of the same process — short written proposal, comment deadline, decision — not a hallway agreement. Teams that skip the checkpoint don't avoid mid-quarter change; they just make the changes silently, and week thirteen becomes an archaeology dig into what the plan even was.

Then, in the last week of the quarter, the cycle closes where it began: the final statuses become next cycle's scoreboard, and the honest notes on every miss become next cycle's better capacity number. The ritual compounds — each quarter's planning is easier and more accurate than the last, because the inputs are already lying around.

Adapting the shape to your situation

The two-weeks-async-plus-90-minutes skeleton scales in both directions, but the knobs differ by context.

Small teams (four to six people). Compress, don't skip. The inputs page and proposals can be a single shared document; the comment window drops to two days; the session drops to 60 minutes. What you must keep even at this size: the written scoreboard, the displacement question, and the three-outcome rule. Small teams are the most tempted to "just talk it through," and they're also the teams where one unwritten plan in one founder's head causes the most damage.

Multiple teams planning against each other. Run the team cycles in parallel with one added phase: after proposals are drafted but before comment windows close, each lead posts their draft plan to a shared page and flags cross-team dependencies explicitly — "our reporting item assumes platform ships the events API by week 6." Dependency conflicts then get sharpened and routed to a single cross-team session (same 90-minute format, leads only) that runs before the team sessions lock their plans. The sequencing matters: teams that lock first and reconcile later end up relitigating their own finished plans.

OKR-driven organizations. The ritual slots underneath company OKRs cleanly: strategic context (input #3) becomes the relevant company objectives, and proposals must name which objective they serve — or argue explicitly for a "keep the lights on" allocation, which honest plans always contain. Resist the temptation to make every commitment a key result; operational work that serves no OKR still has to happen, and plans that pretend otherwise get quietly padded.

Client-driven teams and agencies. Half your capacity isn't yours to plan. The adaptation is in the capacity math: subtract contracted client work first, using the scoreboard's actuals for how much client work really consumed last quarter (it's always more than the contracts say), and run the ritual on the remainder. A plan that budgets the whole team as if clients don't exist is the agency version of capacity theater.

Failure modes to watch for

• The shadow offsite. The 90-minute session metastasizes into a half-day because leaders "just want to walk through everything." Hold the line: reading happens before, deciding happens during. If someone hasn't read, the session is not the remedy.
• Proposals as fait accompli. If the lead writes all the proposals and the team just comments, you've built a consultation, not a planning process. Healthy cycles have proposals from at least half the team.
• Capacity theater. The scoreboard says the team finishes 60% of what it plans, and the room budgets 100% anyway "because this quarter is different." It isn't. Budget to demonstrated capacity and let over-delivery be the surprise, in that order.
• The plan nobody can find. If "where's the quarterly plan?" has more than one answer, the artifacts failed. One canonical page, linked from the team's front door.
• Skipping the broadcast. The plan works internally but adjacent teams keep planning against your dropped items, because nobody told them. The broadcast is thirty minutes of work that prevents a quarter of cross-team friction.

Your next steps

Four to five weeks before next quarter starts:

1. Book the skeleton now: publish date for the inputs page, proposal deadline, comment window, and the 90-minute session. The dates are the process; announce them in one message.
2. Write the scoreboard first. Last quarter's commitments and what actually happened, with honest notes. This is the input that changes behavior most.
3. Ship the proposal template with the displacement question intact — it's the line that turns advocacy into planning.
4. Run the session with the three-outcome rule (in, out, cut down) and end with the commitment question.
5. Publish all three artifacts within 48 hours, wire the plan into your weekly status ritual, and calendar the week-seven checkpoint before you forget.

If you want the whole cycle under one roof: an Openbook space handles it with a Docs room for inputs and proposals with threaded comments, a Project Status room for the plan of record and its weekly RAG history, and the Feed for the broadcast — one address for the argument, the decision, and the follow-through. See how the rooms compose at /product. Then cancel the offsite venue and spend the two days shipping something from the plan instead.