Start with the arithmetic, not the ambition
Q4 looks like three months. It is not. Take 65 weekdays between 1 October and 31 December, subtract public holidays and the dead week between Christmas and New Year, and most European and UK teams are left with roughly 58 to 61 working days. Then apply the discount everyone forgets: annual leave that must be used before it expires, the interview loops that always land in November, and the support or incident load that eats 15–25% of an engineering or ops team's week.
A worked example. Ten people × 61 days = 610 person-days. Remove 8 days of average carried-over leave each (80 days), 18% recurring operational load (95 days) and 5% for planning, reviews and budget work (26 days). You are left with about 409 usable person-days, or 67% of the headline figure. That is the number you plan against. Any quarter plan that assumes 610 days is not a plan, it is a wish.
Force the list down before you score it
Scoring frameworks are useful, but they are slow and they invite lobbying. Do a crude pass first. Every candidate item must have a named owner willing to be accountable, a finish line you could photograph ("invoicing runs without manual export" rather than "improve invoicing"), and an estimate in person-days. Anything failing one of those three tests leaves the list immediately. In practice this removes 40–60% of a typical backlog in an hour.
Only then do you rank. Here is how the common methods compare when you have days rather than weeks to decide.
| Method | Time to run | Best for | Weakness |
|---|---|---|---|
| Forced ranking (1..n, no ties) | 60–90 minutes | Small teams, high trust, obvious trade-offs | Hides reasoning; hard to defend upward |
| RICE (reach × impact × confidence ÷ effort) | Half a day | Product backlogs with usage data | Reach is guessed for internal work; false precision |
| Weighted scoring (3–5 criteria) | Half a day plus a criteria debate | Cross-functional lists with competing goals | Weights get gamed once people learn them |
| WSJF (cost of delay ÷ duration) | Half a day | Anything with a hard date or expiring opportunity | Cost of delay is subjective outside contractual deadlines |
| Buy-a-feature / fixed budget | 90 minutes | Stakeholder groups who all "need everything" | Needs a facilitator; poor for technical dependencies |
For a fourth quarter, WSJF and forced ranking usually beat RICE. Q4 is defined by deadlines — renewal dates, regulatory changes, seasonal demand, budget expiry — and cost of delay captures that better than reach estimates do.
Commit to four things, publish the rest as parked
The failure mode is not choosing badly, it is choosing too much. With about 400 usable person-days, four commitments at 60–90 days each consumes 240–360 days and leaves genuine slack for the unplanned. Five or six commitments means every one runs at 80% resourcing and all of them slip into January in a half-finished state, which is worse than not starting.
Publish the parked list anyway, in the same document, under a heading that says "Not resourced this quarter". Visibility kills most of the lobbying. People can accept being fifth if they can see they are fifth.
Freeze scope, but define the exception route
A scope freeze without an escape hatch gets ignored by week three. Write the rule down: new work enters the quarter only if it is a customer-impacting incident, a legal or compliance requirement, or a revenue opportunity above a stated threshold — and it enters by displacing something named. One in, one out, in writing, signed off by the same person who owns the quarter.
Arguments for a hard freeze
- Finishing rate goes up because context switching drops
- Estimates stay meaningful, so January planning is faster
- Owners can say no without escalating every request
Arguments against
- Q4 opportunities are genuinely time-bound and can be worth more than the plan
- Rigid freezes push urgent work into shadow projects you cannot see
- Small teams may not have a swap of equivalent size to trade
Weekly cadence that keeps it honest
Thirty minutes a week, same slot, four questions per commitment: is the finish line unchanged, are we ahead or behind in person-days, what is blocked, and what would you cut if you lost five days. Track only two metrics on the page — days remaining and percentage of the finish line reached. Burn-up charts are optional; the honest verbal answer to question four is not.
In the final fortnight, switch the meeting from progress to landing: what has to merge, ship, be signed or be invoiced before the last working day, and who is available on the days either side of the holiday. Anything that cannot land by then should be deliberately stopped at a clean checkpoint rather than left mid-flight.
Quick checklist before you publish the plan
- Capacity calculated in person-days, with leave and support load deducted
- Four or fewer commitments, each with an owner and a photographable finish line
- Parked list visible in the same document
- Written exception rule with a one-in, one-out trade
- Named final landing date, at least four working days before the holiday shutdown
Once the priority list is stable, the next dependency is money. Move straight into annual budgeting while your actuals are fresh, and use the tools comparison to check you are not paying for three systems that do the same job.