How do you decide what to work on?

Asked in: behavioral round at every senior level.

The L4 candidate names interesting topics. The L6 candidate describes a process for choosing among many possible projects, with specific recent examples of trade-offs they made.

What an L4 answer sounds like

“I work on whatever my manager prioritizes, plus side projects that I find interesting.”

True for L4. Junior ICs are not expected to set their own direction. you’re being honest, but you’re at L4.

What an L5 answer sounds like

“Three factors:

User and business value. Will this measurably improve a metric the team cares about? Is the impact-per-effort ratio favorable?

Technical leverage. Does this build a foundation other things depend on? Or is it a one-off?

My role on the team. Where am I uniquely positioned to add value vs delegating? What’s the team’s skill gap I can fill?

When several projects are candidates, I’ll write a one-paragraph proposal for each and discuss with my manager. The conversation often reframes the choice.”

This is L5. You have a framework, you involve your manager, you’re aware of your position on the team.

What an L6 answer adds

“…some additional dimensions:

Long-term vs short-term. Some work pays off quickly (a model improvement that ships in a quarter); some pays off over years (a foundational infrastructure investment, a new research direction). I balance the portfolio explicitly: roughly two-thirds short-term work that ships, one-third longer-term investment.

Counterfactual impact. The right question isn’t ‘is this project valuable?’ but ‘is this project more valuable than the next-best thing I could do?’ For senior engineers, the cost of a wrong choice isn’t the project that failed; it’s the project that wasn’t done.

Strategic alignment. Is this in line with where the team and the broader org are going over the next 12-18 months? Even valuable projects in the wrong strategic direction tend to die or get re-routed.

Team multiplier effects. Some projects empower other engineers (shared infrastructure, eval frameworks, internal tools); some only benefit me. At senior levels, multiplier work is often the highest impact even when individual project metrics are smaller.

Honest assessment of my own engagement. If I’m dragging through a project I should be excited about, that’s a signal. Either the project isn’t actually well-scoped, or I’m not the right person, or my interests have shifted. Worth surfacing before I burn months on it.”

flowchart TB
	accTitle: A counterfactual process for choosing a portfolio of work
	accDescr: Candidate projects first pass user-value and strategic-alignment gates. Surviving proposals are compared with the next-best use of the same time using evidence, effort, technical and team leverage, ownership fit, and engagement. Selected work enters a deliberate mix of near-term delivery and longer-term investment; declined work keeps a written rationale and can return when evidence or strategy changes.
	A["Candidate projects<br/>one-paragraph proposals"] --> B{"Advances a user outcome<br/>and team strategy?"}
	B -->|"no"| C["Decline now<br/>record the reason"]
	B ==>|"yes"| D["Name the next-best use<br/>of the same time"]
	D --> E["Compare evidence + value<br/>against effort + risk"]
	D --> F["Compare technical leverage<br/>and team multiplier"]
	D --> G["Check ownership fit<br/>and sustained engagement"]
	E --> H{"Best counterfactual<br/>use of this person's time?"}
	F --> H
	G --> H
	H -->|"no"| C
	H ==>|"yes"| I["Select with manager<br/>and affected partners"]
	I --> J["Near-term work that ships<br/>roughly two-thirds"]
	I --> K["Longer-term investment<br/>roughly one-third"]
	C -. "new evidence or strategy" .-> A
	class A,D viz-input
	class B,H viz-focus
	class C viz-warning
	class I,J,K viz-output
	class E,F,G viz-state
	class A viz-wide

Read it this way: do not ask only whether a project is valuable. First remove work that misses user outcomes or strategy, then compare each survivor with the next-best use of the same person's time. Select a portfolio rather than a single winner, so near-term delivery and longer-term leverage both receive deliberate capacity. The two-thirds/one-third split is the answer's example, not a universal rule. Original synthesis checked against NASA's decision-analysis guidance, OpenStax on opportunity cost, and Cooper, Edgett, and Kleinschmidt on project portfolios.

Tells that get you a strong-hire vote

You have an explicit framework, not just a list of considerations.
You bring up counterfactual impact (vs the next-best alternative).
You name strategic alignment as a separate axis from technical merit.
You discuss multiplier effects of platform / shared work.
You can give a specific recent example of a project you chose vs one you declined.

Tells that get you down-leveled

“I work on what’s most interesting” with no business framing.
No mention of strategic alignment.
No examples of trade-offs you made.
Treating prioritization as your manager’s problem alone.

Common follow-up

“Tell me about a time you decided not to do a project that was assigned to you.”

The L6 answer should describe:

The project and why it was assigned.
The reasoning for declining (better alternative, scope creep, mismatch with team strategy).
How you raised the concern with the assigning party.
The eventual decision and what shipped instead.
Any reflection on whether you’d handle it the same way again.

The signal is whether you can productively push back, not whether you’ve ever said no.

How do you decide what to work on?

Practice before you read