What gets scored: the six signals
What it is
Senior interviews test whether you can solve a problem. Staff and lead interviews test whether you can pick the problem, frame it, and get other people to solve it with you. Different bar, and the scoring rubric reflects it in six dimensions.
Signal Senior looks like Staff/lead looks like
------------------------------------------------------------------
SCOPE owns a service owns a problem space
across teams
AMBIGUITY given a spec, executes given a vague goal,
produces the spec
JUDGMENT picks a good option explains three options,
why the others lose, and
the reversal criteria
INFLUENCE convinces their team convinces peers and leads
who do not report to them
MULTIPLIER delivers others got faster because
of them
RISK handles known risk names the risk nobody
else named
And two track-specific additions:
TEAM LEAD adds performance conversations, the hiring bar,
delivery predictability, morale under
pressure, PM and stakeholder negotiation,
headcount and prioritisation trade-offs
STAFF adds technical strategy over 12 to 24 months,
cross-org standard setting, deprecation and
migration leadership, design review
authority, making the hard call in a room of
senior people
What this is confused with: seniority as difficulty. A staff design round is not a harder version of a senior one. It asks a different question, and answering the senior question extremely well is the most common way to fail it.
Also confused: the six signals as a checklist to recite. They are the axes an interviewer scores on, so the useful work is making each one visible in a story you were going to tell anyway.
The problem it solves
The single most common failure is telling senior-level stories in a staff-level interview.
The story is about WHAT YOU BUILT, not about HOW THE ORG
CHANGED.
"I designed and built the new ingestion pipeline. It
handles 40,000 events a second, we cut p99 from 800ms to
120ms, and I wrote most of the core."
That is an excellent senior answer. Every clause is about
personal execution.
An interviewer scoring SCOPE hears one service.
Scoring INFLUENCE hears nobody outside the team.
Scoring MULTIPLIER hears one person's output.
Scoring AMBIGUITY hears an implementation of a decided thing.
The candidate is strong, the story is true, and it argues for
the level they already have.
The same project, told against the six signals:
"Three teams were each building their own ingestion, and
the reason was that nobody owned the question of what
'an event' meant here [SCOPE, AMBIGUITY]. I wrote the
proposal for a shared pipeline and spent about six weeks
getting the two teams who disagreed with it to a position
they'd defend, which meant giving up the schema registry
design I preferred [INFLUENCE, JUDGMENT]. The thing I got
wrong was the migration order: I sequenced by technical
difficulty rather than by which team could absorb the
change, and we stalled for a month [RISK, and a real
regret]. After it landed, the two consuming teams stopped
maintaining their own ingest, which is about a person and
a half of ongoing capacity [MULTIPLIER], and p99 went from
800ms to 120ms."
Same project. Same facts. The numbers are still there, at the
end, where they belong.
The technical result moved from the opening to the closing, and that reordering is most of the difference.
Mechanics
Making each signal visible
Scope. The question the interviewer is answering is "how big was the thing this person owned."
Weak: "I owned the payments service."
Strong: "I owned payment correctness as a problem, which meant
the service, the reconciliation with finance, the
contract with the two teams that wrote to our ledger,
and the on-call for all of it."
The tell is whether the boundary you describe is a REPOSITORY
or a PROBLEM. Repositories are assigned; problems are claimed.
Ambiguity. The question is "what was the state of the problem when you got it."
Weak: "We were asked to reduce checkout latency."
Strong: "The complaint was 'checkout feels slow', which
turned out to be three different things: a slow p99
for large carts, a rendering delay nobody had
measured, and a perception problem with the spinner.
I split them because two of them had different
owners."
The signal is producing the specification, not receiving it.
And the strongest version includes a problem you decided NOT
to solve, with the reason.
Judgment. The question is "does this person reason about options, or advocate for one."
The complete shape has four parts, and most candidates give
two:
1. the options, at least three
2. what each costs
3. why the ones you rejected lose, SPECIFICALLY
4. the REVERSAL CRITERIA: what would have to be true for
you to change your mind
Part 4 is the one almost nobody offers, and it is the
strongest single signal on this axis, because it demonstrates
that the decision was made against a model rather than a
preference.
"I'd revisit the single-writer design if write volume
passed about 40,000 a second or if we needed sub-50ms
writes in Europe, and both were far enough away that I
took the simpler thing."
Influence. The question is "can this person move people who do not report to them."
Weak: "I convinced the team to adopt it."
Strong: "The platform team's lead thought it was the wrong
layer, and he was partly right. I rewrote the
proposal to move the boundary he objected to, which
cost us a quarter, and he then co-signed it, which
is what got the other two teams to move."
The signal is a NAMED DISAGREEMENT and what it cost you. A
story where everyone agreed is a story with no influence in
it, and interviewers hear that.
Multiplier. The question is "did other people get faster."
Weak: "I mentored two juniors."
Strong: "Two teams stopped maintaining their own ingest,
which is roughly a person and a half of recurring
capacity, and the runbook I wrote is why the on-call
rotation could expand from four people to seven."
Quantify the OTHER PEOPLE'S output, not your own. Mentoring
counts when you can say what the mentee did afterwards.
Risk. The question is "does this person see what others miss."
Weak: "We had a rollback plan."
Strong: "The risk nobody had named was that our two consumers
both cached the old schema for up to 24 hours, so a
cutover looked instant to us and was a day long for
them. That's why the migration ran a dual-write
window of 48 hours rather than the two we planned."
The framing that scores is "the risk nobody else named",
which requires you to have been the one who named it, and to
say what changed because you did.
The audit, which takes an hour
Take your five best stories. For each, mark which of the six
signals it evidences, honestly.
A typical result before any work:
story scope ambig judg infl mult risk
ingest yes - yes - - yes
latency work - - yes - - -
migration yes yes yes yes - yes
the outage - - yes - - yes
mentoring - - - - yes -
INFLUENCE and MULTIPLIER are the two columns that are usually
empty, and they are the two that most distinguish the level.
The fix is not to invent stories. It is to notice that the
influence and multiplier content is ALREADY IN the stories
you tell and is being left out because it feels like it is
not the interesting part.
The influence and multiplier content is almost always present and almost always omitted, because the engineering is the part that felt like the work.
What the interviewer writes down
The scoring artifact is a written assessment read by people
who were not in the room. So the useful question is not "did
they understand me" but "what can they WRITE DOWN".
Give them quotable lines:
"owned the event schema standard across four teams"
"the two consuming teams retired their own ingest, about
1.5 engineers of recurring capacity"
"changed position after the platform lead's objection, and
that is what unblocked adoption"
Vague warmth does not survive the transcription. A sentence
with a boundary and a number does.
A worked example: two candidates, the same project
Two engineers from the same team, interviewing at the same company for staff, describing the same eighteen-month migration.
Candidate A's deep dive, compressed:
Spent 35 of 45 minutes on the technical design: the sharding
scheme, the consistency model, the dual-write mechanism, the
backfill's throughput tuning. All of it correct and
genuinely deep. Answered every follow-up.
The last 10 minutes covered the outcome: 4x throughput, p99
down 60%, cost down 22%.
Debrief comments:
"Very strong engineer. Deep systems knowledge."
"I'd hire, at senior."
"Scope: I heard one system. Ambiguity: the problem was
well-defined when they got it. Influence: didn't come up."
Outcome: offer at senior. Declined.
Candidate B's deep dive, same project:
Spent 8 minutes on the technical design, at a level of detail
that invited follow-up, and answered every follow-up as deeply
as A did when asked.
Spent the rest on:
- why the migration existed at all, which was a
reliability argument the candidate had made to a VP with
an error-budget number attached [SCOPE, AMBIGUITY]
- the three approaches considered, what each cost, and the
condition under which they would have chosen differently
[JUDGMENT]
- the two teams who resisted, what their objection was, and
the design change that resolved it [INFLUENCE]
- the sequencing mistake: ordered by technical difficulty
rather than by which team could absorb the change, which
cost a month [RISK, regret]
- what the org could do afterwards that it could not
before: two teams retired their own pipelines, and the
schema standard became the default for new services
[MULTIPLIER]
Same numbers, at the end.
Debrief comments:
"Owned the problem, not the project."
"Named a real mistake with a real cost."
"Changed the design in response to an objection, and could
say why the objection was right."
Outcome: offer at staff.
Candidate A was, by the team's own assessment, the slightly stronger engineer. The difference in outcome was entirely in which questions the 45 minutes answered.
What A did differently the second time, six months later:
Same project, same facts, one structural change: the
technical design moved from 35 minutes to 8, with an explicit
offer ("I can go as deep as you like on the sharding scheme,
and I'd rather spend the time on why we did this and what it
cost us to get agreement").
Two interviewers took the offer and went deep. A answered at
the same depth as before, in the follow-up rather than in the
opening.
Offer at staff.
A's own recorded assessment: "I thought the depth WAS the
signal. It's the price of entry. The signal is what I did
with the room's attention once I'd paid it."
"The depth is the price of entry, not the signal" is the compressed lesson, and it is why the reordering works without removing anything.
The signal audit, run on both candidates' story sets beforehand, would have predicted this:
A's 5 stories B's 5 stories
scope 2 4
ambiguity 1 3
judgment 5 5
influence 0 4
multiplier 1 3
risk 3 4
A's judgment column was full and their influence column was
empty. Both had the same underlying career; one of them had
noticed which parts of it the rubric asks about.
Production evidence
Published engineering career ladders (Rent the Runway's, CircleCI's, Dropbox's, Square's, and the collection at progression.fyi) all express the senior-to-staff transition in terms of scope, ambiguity and influence rather than technical difficulty, which is the structural basis for the six signals: the interview rubric follows the ladder.
Will Larson's Staff Engineer and Tanya Reilly's The Staff Engineer's Path both identify the same transition, with Larson's archetypes (tech lead, architect, solver, right hand) all defined by the shape of the scope rather than by depth, and Reilly's treatment of influence without authority as the distinguishing capability.
Google's hiring committee model, where a packet is read by people who were not in the room, is the documented reason quotable, written-down scope statements matter more than the impression left on any one interviewer.
Amazon's requirement that leadership-principle answers carry data, and its Bar Raiser role, are the clearest instance of a loop scoring on named dimensions rather than on a general impression.
Meta's published distinction between E5 and E6 impact is the same claim in different words: E5 is demonstrated by what you built, E6 by change across an organisation.
Structured-interview research consistently finds that defined dimensions with behavioural evidence predict performance better than unstructured impressions, which is why loops converge on rubrics and why speaking to the rubric is not gaming, it is answering the question asked.
The debate
Is technical depth still what matters? It is the price of entry and not the differentiator. A staff candidate who cannot go deep fails immediately, and a staff candidate who goes only deep gets a senior offer. The resolution is to demonstrate depth quickly and explicitly offer more, then spend the room's attention on the signals depth does not cover.
Should you consciously speak to the rubric? Yes, and it is not gaming. The rubric describes what the role requires, so an answer that omits influence and multiplier is not a modest answer, it is an answer to a different question. The failure mode is reciting the signals rather than evidencing them, and the distinction is whether the story contains a named disagreement and a number about someone else's output.
Do you need a story per signal? No, and trying produces six thin stories. Three or four strong stories that each carry three or four signals is the right shape, and the audit's purpose is to find the empty columns, which are almost always influence and multiplier.
Is the difference between senior and staff real, or is it politics? It is real and it is not about technical ability. The observable difference is whether the org changed, and that requires convincing people who do not report to you, which is a genuine skill with a genuine learning curve. The politics critique lands on the calibration process rather than on the distinction.
Should you admit a mistake? Yes, with the cost, and it scores on the risk and judgment axes rather than against you. A story with no regret in it reads as either unreflective or sanitised, and the strongest version names the mistake, its cost, and what you changed, which is what the candidate in the worked example did with the sequencing error.
Is it dishonest to reframe an existing story? No, as long as the facts do not change. The influence and multiplier content is usually already true and simply not mentioned, because the engineering is what felt like the work. Adding it is accuracy, not spin.
Follow-up Q&A
"What is the difference between a senior and a staff interview?"
The question being asked. Senior asks whether you can solve a problem; staff asks whether you can pick the problem, frame it, and get other people to solve it with you. That shows up on six axes: scope, from owning a service to owning a problem space across teams; ambiguity, from executing a spec to producing one; judgment, from picking a good option to explaining three and the reversal criteria; influence, from convincing your team to convincing peers who do not report to you; multiplier, from delivering to making others faster; and risk, from handling known risk to naming the one nobody else named.
"What is the most common way strong candidates fail a staff loop?"
Telling senior-level stories, extremely well. The story is about what they built rather than how the organisation changed, so an interviewer scoring scope hears one service, scoring influence hears nobody outside the team, and scoring multiplier hears one person's output. Every clause is true and the story argues for the level they already have. In one case two engineers described the same eighteen-month migration; the slightly stronger engineer spent 35 of 45 minutes on the technical design and received a senior offer, and the other spent 8 and received a staff offer.
"How do you make influence visible in a story?"
With a named disagreement and what it cost you. "I convinced the team" evidences nothing; "the platform lead thought it was the wrong layer and he was partly right, so I moved the boundary he objected to, which cost a quarter, and he then co-signed it, which is what got the other two teams to move" evidences all of it. A story where everyone agreed contains no influence, and interviewers hear that immediately.
"What is the strongest single thing to add on the judgment axis?"
The reversal criteria. Most candidates give the options and the choice; almost nobody says what would have to be true to change their mind. "I would revisit the single-writer design if write volume passed about forty thousand a second or if we needed sub-fifty-millisecond writes in Europe, and both were far enough away that I took the simpler thing" demonstrates that the decision was made against a model rather than a preference, which is the distinction the axis is measuring.
"How do you find the gaps in your own stories?"
Take your five best stories and mark which of the six signals each one evidences, honestly. It takes an hour. The two columns that are usually empty are influence and multiplier, and the fix is not to invent stories: that content is almost always already true and simply left out, because the engineering is the part that felt like the work. Quantify other people's output rather than your own, and name the disagreement you resolved.
"Why does what the interviewer can write down matter?"
Because the scoring artifact is a written assessment read by people who were not in the room, and at companies with a hiring committee or a calibration panel the packet is what is evaluated. Vague warmth does not survive transcription. A sentence with a boundary and a number does: "owned the event schema standard across four teams", "two consuming teams retired their own ingest, about one and a half engineers of recurring capacity". Give them quotable lines rather than a good impression.
Common misconceptions
"Staff is harder senior." It is a different question. Answering the senior question exceptionally well is the most common way to fail a staff loop.
"Depth is the signal." Depth is the price of entry. Demonstrate it quickly, offer more explicitly, and spend the remaining attention on what depth does not cover.
"Speaking to the rubric is gaming." The rubric describes the role. An answer omitting influence and multiplier is answering a different question, not being modest.
"I need a story for each of the six signals." Three or four stories carrying three or four signals each is the right shape. Six thin stories is worse than four strong ones.
"Admitting a mistake is risky." A story with no regret reads as unreflective or sanitised. The mistake with its cost and what you changed scores on judgment and risk.
"My influence content isn't interesting enough to mention." It is the content the level is defined by, and it is the column most commonly left empty by strong candidates.
Interview delivery note
Say this verbatim, when asked what you would bring at this level: "Senior is whether I can solve the problem. Staff is whether I can pick it, frame it, and get people who don't report to me to solve it with me. So the story I'd tell about that migration is less about the sharding scheme and more about the two teams who disagreed and the design change that resolved it." It states the distinction and then demonstrates that you know which story to tell.
The senior-versus-staff separator here is recursive and worth stating plainly: it is offering the depth rather than defaulting to it. Saying "I can go as deep as you like on the sharding scheme, and I'd rather spend the time on why we did this and what it cost to get agreement" gives the interviewer the choice, proves you have the depth available, and reallocates the room's attention to the axes that are actually being scored. In the worked example that single change, with no new content, moved an offer from senior to staff.
The second signal is the reversal criteria. Naming what would make you change your mind, with a threshold, is the cheapest and least-used way to demonstrate that a decision was reasoned rather than preferred, and it converts a judgment answer from an advocacy into an analysis.
Further reading
- Will Larson, Staff Engineer, and Tanya Reilly, The Staff Engineer's Path, for the scope-and-influence definition of the level.
- Published engineering career ladders (Rent the Runway, CircleCI, Dropbox, Square) and the progression.fyi collection, for the rubric the interview rubric follows.
- The architecture deep dive page, which is where these signals are most often won or lost, and the SCOR, STAR and the scar-tissue story page for the story structure that carries them.