Every stakeholder says their item is urgent. How do you decide what goes into next quarter?

What the framework is actually for

The weak answer to this question is a list of framework names. The strong answer starts from why a framework helps at all, and it is not that the arithmetic is wise. RICE cannot tell you whether enterprise single sign-on matters more than checkout latency, because you supplied every input. What a framework does is force the disagreement into named variables, so an argument that would otherwise be "mine is urgent" against "no, mine is" becomes "we disagree on reach" or "you think confidence is high and I think it is a guess". That is a solvable argument. Say this out loud early, because it is the difference between a candidate who has run a real quarterly planning cycle and one who has read about them.

The second thing a framework buys you is a record. When the same stakeholder returns in week five, the value is not the score, it is that the trade-off was written down in front of everyone and nothing has changed since.

I use RICE by default, and cost of delay divided by duration when the portfolio is genuinely date-sensitive, because that is the case where a purely value-per-effort ranking hides the fact that a cheap item's value evaporates if it slips.

Ring-fence before you rank

Nothing gets scored until the immovable work is out of the pool, because scoring a regulatory deadline against a dashboard request is a category error that discredits the whole exercise. Take the quarter's engineering capacity, then subtract in this order: work with an externally fixed date, contractual commitments already signed, and a standing reliability and debt allowance. Whatever is left is the discretionary budget, and only that gets a score.

Concretely, on a payments platform with one squad, call the quarter twelve engineer-months. A twenty percent standing allowance for reliability and debt takes 2.4. The KYC re-verification work carries a regulator's date and takes 4. That leaves 5.6 engineer-months of genuinely discretionary capacity, and that number is the single most useful sentence in the entire planning conversation.

The worked scoring

Reach is merchants affected per quarter, impact uses the conventional 3 / 2 / 1 / 0.5 / 0.25 scale, confidence is a percentage, effort is engineer-months. Reach must use the same unit and the same period for every row or the scores are not comparable, which is the most common way teams break RICE without noticing.

Item	Reach	Impact	Conf.	Effort	Score
Checkout latency reduction	12,000	0.25	70%	2	1050
Bulk refunds self-service	900	0.5	100%	2	225
Enterprise SSO (sales-blocking)	40	3	80%	3	32
Executive analytics dashboard	15	1	50%	2	3.8

Against 5.6 discretionary months, latency and bulk refunds fit at 4 months. SSO does not fit whole. So it gets sequenced first in the following quarter with a two-week discovery slice now, which is a real commitment the sales lead can take to a customer rather than a vague reassurance. The dashboard does not make the cut on this arithmetic at all.

Notice what the table exposes: SSO scores low only because forty accounts is a small reach, and the sales lead's real argument is that those forty accounts carry disproportionate revenue. That argument is legitimate, and it is an argument about whether reach should be weighted by contract value. The framework did not settle it, but it located it precisely, which is the whole point.

The escalated executive request

An executive asking for something is data, not noise. Before treating that dashboard as a score of 3.8, assume the exec knows something you do not, and go and ask: is this for a board conversation, a specific decision, an investor question. Nine times in ten the underlying need is narrower than the request, and can be met with an existing BI export or a scheduled query for a day of work rather than two engineer-months of product surface. Satisfying the need while declining the build is the move that keeps the relationship and the capacity.

Where the need is genuinely strategic and you are simply wrong about its value, update the score and say so. Being seen to change your mind on evidence is what makes the other ninety percent of your noes credible. What you must not do is silently absorb it, because unfunded executive work does not displace nothing, it displaces whatever the team quietly drops, and you will not find out which until the quarter fails.

The hard-dated compliance item

This one is not prioritised, it is scheduled, and the distinction should be explicit in how you talk about it. You do not defend it on value; you state the date, the consequence of missing it, and the capacity it consumes, and then you plan the rest of the quarter around the remainder. The failure mode here is optimism: teams treat the compliance estimate as fixed when it is usually the least well understood item on the list, and it grows. Start it early rather than last, front-load the unknowns, and set a checkpoint at which you would escalate for external help rather than discovering the shortfall in the final fortnight.

The reliability and debt reserve

Reserving capacity is the item stakeholders most reliably attack, because it has no requester in the room. Defend it as insurance against velocity loss, not as engineering hygiene: the reserve exists so that the next four quarters can still deliver at this rate. It also needs to be a real allocation with named work and visible outcomes, because a reserve that is never spent on anything legible becomes the first thing cut next planning cycle. If it keeps getting raided mid-quarter, that is the signal to raise, not to quietly abandon it.

Making the no survive

A no that survives has four parts, and candidates typically produce only the first. State the decision. State what it would displace, by name, because "not this quarter" invites renegotiation while "this would push bulk refunds out, which your support team asked for" makes the requester weigh the trade-off themselves. Offer the cheapest thing that addresses the underlying need. Then give a re-entry condition: what would have to be true for this to win next time, and when it will next be considered.

Sample phrasing that does all four:

We are not building the dashboard this quarter. Taking it would displace the bulk refunds work, which came from your own support leads. I can get you the three numbers you need as a weekly export by Friday. If the deal pipeline still needs a self-serve view at the October review, it goes in on the strength of that, and I will bring the numbers.

Finally, put the whole list in one place everyone reads, including the items that did not make it and why. Noes delivered in private one-to-ones do not survive because each stakeholder only sees their own rejection and reasonably concludes they were singled out. Published together, the list is visibly a capacity constraint rather than a personal judgement, and it does most of the defending for you.

The trap

The trap is arguing the score. Once you are defending why enterprise SSO is a 32, you have accepted that a better-argued input beats a worse one, and any determined stakeholder can move a confidence percentage or an impact multiplier until their item wins. The argument you can actually win is about capacity: there are 5.6 engineer-months, here is what is in and what each addition would remove. Scores order the list; capacity is what makes the no hold. Candidates who have only read about prioritisation defend the number, and candidates who have lived it defend the constraint.

The framework's output is not a decision, it is a shared vocabulary for the trade-off. What makes a no stick is a fixed capacity, a named thing it displaces, and a stated condition under which it could change.