In this article
Before you start
This page belongs to the behavioural interview guide; the general method for any story is in how to answer interview questions.
Why interviewers ask this
A candidate who cannot say what they dropped will drop things silently on the job. The question tests how you choose when everything is urgent, whether you tell people what you are not doing, and whether you respect who owns the decision. Candidates report questions of this kind in loops at large technology companies: how you rank several projects, how you ask for more capacity, and what a ruthless trade-off cost you. No public source here ties any one wording to a named company.
A weak answer says "I stay organised and work hard." A strong answer names a criterion, a cut and the person who agreed. A Staff answer adds the cost of the cut and what you changed afterwards. Related ground appears in the guides to Amazon Leadership Principles and to Google behavioural interview questions.
Prioritising under pressure from first principles
- Cut list
- A cut list is a short written statement of what you will not do, or will do later, because of a priority choice. It gives the cost of each cut and names who agreed to it.
In plain words
Picture a restaurant kitchen on a busy night. Orders arrive faster than one cook can finish them. A good cook does not rush every dish. The cook reads the tickets, starts what takes longest, and tells the waiter that one table will wait a little longer.
The analogy breaks in one place. A head cook owns the kitchen. An IC usually does not own the decision, so telling people is not enough: the owner must agree.
In one precise sentence, prioritising under pressure is choosing what to do first and what not to do from a stated criterion, then getting the person who owns the outcome to agree. The four steps are list and size, score, cut and name the cost, and agree with the owner.
Step-by-step description
- Every demand gets an owner, deadline and effort.
- One stated criterion gives an order.
- The cut list says what waits and why.
- The owner makes the final call.
- The four steps are list and size, score, cut and name the cost, and agree with the owner.
- The owner of the outcome makes the final call.
Step 1: List every demand and size it
Listing every demand comes first, because a priority call made from memory leaves things out. Input: every request, bug and promise. Output: one list in which each line shows who asked, the deadline, what happens if it slips and a rough effort. Intercom's RICE write-up estimates effort in person-months; a person-month is the work one team member can do in a month. In the running example, four demands cost 0.5, 1, 4 and 2 person-months, 7.5 in all, while one person has 2.0 person-months of capacity this quarter (illustrative).1 The gap, not any single request, is the problem to solve. You now have the facts needed to compare demands.
Step 2: Score the demands against one stated criterion
Scoring ranks the demands by a criterion you can say aloud. A common choice is the RICE score, which multiplies reach, impact and confidence and divides by effort (Intercom, 2018). Reach counts people affected in a period. The same source picks impact from 3, 2, 1, 0.5 or 0.25 and confidence from 100%, 80% or 50%; effort divides the total because more effort is worse. Input: the list from the first step. Output: an order. The four projects below are illustrative inputs run through that formula, giving 8,000 for the checkout fix, 2,000 for the latency work, 1,125 for the redesign and 320 for the dashboard.1
| Project | Reach per quarter | Impact | Confidence | Effort (person-months) | RICE score |
|---|---|---|---|---|---|
| Checkout fix | 2,000 | 2 | 100% | 0.5 | 8,000 |
| Latency work | 5,000 | 0.5 | 80% | 1 | 2,000 |
| Redesign | 3,000 | 3 | 50% | 4 | 1,125 |
| Dashboard | 800 | 1 | 80% | 2 | 320 |
Text description of the diagram
- The scores are 8,000 for the checkout fix, 2,000 for the latency work, 1,125 for the redesign and 320 for the dashboard.
- The checkout fix scores four times the latency work, so it goes first.
Step 3: Cut something and name what it cost
Cutting means choosing what does not happen and saying so. With 2.0 person-months of capacity, the plan funds the checkout fix (0.5) and the latency work (1.0): 1.5 person-months, 75% of capacity, which leaves 0.5 person-months, a quarter of capacity, free for urgent work (illustrative).1 The redesign and the dashboard are cut, which is 6.0 of the 7.5 person-months demanded.1
The cost goes in the cut list. For the redesign it reads: the 3,000 people per quarter who would have reached it wait at least a quarter longer (illustrative). When an interviewer asks what the trade-off cost you, this is the sentence they are asking for. Input: the ranked list and the capacity. Output: a cut list the owner can accept or change.
Step 4: Agree the decision with the person who owns it
The owner of the outcome makes the final call, and your job as an IC is to make that call easy. Bring the ranked list, the cut list, a recommendation and one or two options. The owner is usually a manager, a product owner or the service owner; ask who it is before the pressure starts. Do not decide alone, and do not give in silently.
For reliability trade-offs, the call should follow a rule agreed in advance, such as an error budget, which the Google Site Reliability Engineering book's chapter on embracing risk covers. Afterwards, tell everyone affected the reason and the new date, and note when you will re-score. Persuading an owner you do not report to is its own skill; see influencing without authority.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →The Staff-level answer
The answer below is an illustrative example, not a real person's account.
Tell me about a time you had to prioritise under pressure.
I had four requests and far less capacity than I needed. I listed each with its deadline, who asked and a rough effort, then ranked them by customers affected, impact per customer and effort, and I marked which inputs were guesses. A checkout bug and a latency fix came first. I recommended dropping the redesign and the dashboard and told my manager the cost: the redesign would slip a quarter. We had agreed customer impact would decide, and she agreed in writing. I kept some capacity free for urgent work. I re-scored monthly, and when the redesign's data firmed up, I moved it up.
Say it in this order: the demands, the criterion, the cut with its cost, who agreed, and what you reviewed. It is Staff-level because it names the criterion, the cost of the cut, the person who agreed and the follow-up.
The rubric
| Dimension | Weak answer | Strong answer | Staff answer |
|---|---|---|---|
| Criterion | "I did the most urgent thing." | Names one criterion, such as customers affected. | States the criterion, marks guessed inputs and shows what would change the order. |
| The cut | Says nothing was dropped. | Names what was deferred. | Names the cut and its cost in customer or time terms. |
| Authority | Decides alone or complies silently. | Takes a recommendation to the owner. | Agrees the decision rule beforehand and gets the call in writing. |
| Communication | Tells the team afterwards. | Tells affected people before acting. | Gives each person the reason, the new date and a way to challenge it. |
| Follow-up readiness | No review. | Reviews when something changes. | Sets a review date and says what would reverse the decision. |
This rubric is our own reading of what a strong answer shows, not any company's scoring sheet.
Three variants of this question
How do you prioritise multiple projects at the same time?
Name your criterion, show the order it produces, and say what waits. Limit active work to what capacity allows, because a long list of started projects is not a priority order.
How do you request additional resources when capacity is insufficient?
Show the gap between demand and capacity, for example 7.5 person-months demanded against 2.0 available (illustrative).1 Then bring options with their costs: add people, cut scope or move the date. Ask the owner to choose, rather than asking for more people with no alternatives.
A teammate insists on shipping a feature on time even though it regresses p99 latency past your SLO. How do you resolve it, and who makes the final call?
Settle it with numbers and the rule your team agreed, not with who argues harder. The final call belongs to the owner of the service under the agreed error-budget policy; if none exists, it goes to the manager or service owner, with you bringing the data and options such as shipping behind a flag or fixing first.
An example shows why the numbers matter. Suppose the SLO says 99% of requests must be faster than 200 milliseconds, over 10,000,000 requests in 30 days (illustrative).1 The error budget is then 100,000 slow requests. If the regression makes 3% of requests slow, that is 300,000 a month, and the budget is spent in 10 days.1
Text description of the diagram
- The error budget is 100,000 slow requests.
- If the regression makes 3% of requests slow, that is 300,000 a month, and the budget is spent in 10 days.
Common mistakes
- "I just worked harder and got everything done." (wrong) It hides the cut, so the interviewer hears no prioritisation. (right) "I could not do everything, so I cut the dashboard and told my manager what that cost."
- "I decided what to drop and told the team afterwards." (wrong) You may not own the call. (right) "I recommended the cut, my manager agreed, and then I told the team."
- "The score said so, so it was objective." (wrong) Impact and confidence are judgements (Intercom, 2018). In the example, lifting the redesign's confidence from 50% to 100% raises its score from 1,125 to 2,250, above the latency work's 2,000; the order flips once confidence passes about 0.89, a value between Intercom's 80% and 100% steps.1 (right) "The score ordered the options, and I flagged the two guesses that could change the order."
- "Everything was a priority." (wrong) It means nothing was ranked. (right) "I ranked them, and one item was last."
Follow-up questions
Can you tell me about a time you prioritised the customer over technical perfection?
Pick a case where you shipped a smaller correct thing, name the technical debt you accepted and say when you repaid it. The same four steps apply, with the customer cost of waiting as your criterion.
Can you tell me about a time you had to reprioritise when a deadline shifted?
Show that you re-ran the scoring with the new input, not that you worked later. Say which item moved out and who agreed.
What if your capacity drops?
Re-fit the plan to the new number. With 1.0 person-month instead of 2.0, only the checkout fix (0.5) fits and the latency work (1.0) no longer does, so the cut list grows (illustrative).1
What do you do when a stakeholder says everything is the top priority?
Ask what happens if each item slips by a week, then show the trade-off and ask them to choose what waits. For the people side, see interpersonal communication in interviews.
What was your most challenging project?
Choose a project where you made a real cut, then tell it in the order of the answer above. The common interview questions guide shows how to shape other stories the same way.
- Prioritising under pressure means ranking by a stated criterion, cutting on purpose and naming the cost.
- A RICE score orders options, but impact and confidence are judgements, so flag the guesses.
- As an IC you recommend; the owner of the outcome makes the final call.
- For a reliability clash, an agreed error budget settles it better than argument.
- A cut list with a cost is the sentence that separates levels.
Sources
- RICE: Simple prioritization for product managers — Intercom (Sean McBride), 5 Jan 2018
- Site Reliability Engineering, Chapter 3: Embracing Risk — O'Reilly (Beyer et al.), 1 Apr 2016
