HR, screening and offer fundamentals

Four things, none of which is technical depth. Whether your experience matches the level the requisition was opened at, because a recruiter who advances a candidate two levels off spec wastes an engineer's afternoon and their own credibility. Whether your compensation expectation is inside the band, since that is the cheapest disqualifier available and the one they are measured on. Whether the logistics work — notice period, location, visa, hybrid days. And whether you can hold a coherent conversation about your own work, which is a proxy for how you will present to the hiring manager. Everything you say is being written into a summary that the next interviewer reads instead of your CV, so the screen is where your framing is set for the whole loop.

Ninety seconds is roughly two hundred words, which is far less than candidates assume, so the shape has to be decided in advance.

Target: 90 seconds. Four beats. Present, evidence, why here, handover.

NOW - one sentence locating you at a level.
  "I'm a backend engineer with eight years in payments, currently a
   senior engineer at Wexford, where I own the settlement service."

EVIDENCE - one or two specifics with a number, chosen to match the ad.
  "The two things I'd point at: I led the move of settlement off a
   nightly batch onto an event stream, which took reconciliation
   breaks from about forty a month to under five. And I've been the
   engineer on call for that system for three years, so I've run the
   incidents as well as designed it."

WHY THIS ROLE - specific to them, one sentence, no flattery.
  "I'm looking at you because the role is on the ledger team and it's
   the same problem at a much larger volume, and because it's an
   individual-contributor role rather than a step into management."

HANDOVER - give them the next question.
  "Happy to go deeper on the settlement work or talk about what I'm
   looking for next - whichever is more useful."

WHAT IS NOT IN IT
  - chronology from university onwards
  - the words "passionate", "journey", "fast-paced"
  - anything the recruiter cannot assess or repeat to the manager

The reason to script this and no other answer is that it is the only question you are guaranteed to be asked, and it sets the frame everything after it is judged against. A candidate who opens with a chronological walk from graduation spends the ninety seconds on their least relevant decade and arrives at the current role as the clock runs out.

The evidence beat is where the answer is won or lost, and the test for it is whether a recruiter could repeat it to the hiring manager. "Took reconciliation breaks from forty a month to five" survives that retelling; "worked on large-scale distributed systems" does not, because it contains nothing to carry. Pick the specifics from the advert rather than from your own sense of what was hardest.

The handover matters more than it looks. Offering the recruiter two directions turns the screen into a conversation you are partly steering, and it prevents the awkward pause where they scramble for a follow-up and land on salary earlier than you wanted.

The omissions are deliberate. Anything the recruiter cannot assess is wasted breath at this stage — architecture detail, framework opinions, the reason your last migration was hard. Save it for the round where someone can appreciate it.

The audience changes, so the currency changes. In a screen you are being matched against a requisition, so you speak in scope, domain and level: what you own, how big it is, how long you have owned it, and the outcome in terms a non-engineer can repeat. In a technical round the same question is an invitation to nominate the system the interviewer will interrogate, so you name one thing with real depth behind it and stop, because whatever you mention becomes the next twenty minutes. The commonest error is running the recruiter version in the technical round, which sounds like a sales pitch to an engineer, and the technical version in the screen, which the recruiter cannot use and will summarise as "seemed strong, hard to follow".

State it accurately and immediately, including the part that is inconvenient. If your notice is three months, say three months and then say what is actually negotiable — whether unused leave can be offset, whether your employer has released people early before, whether a partial handover start is possible. Vagueness here is expensive because a hiring manager planning a start date will discount a candidate whose availability is unknown, and a notice period discovered late in the process feels like something you concealed. Do not offer to break your notice, because it tells the new employer exactly how you will treat them on the way out. If the notice is genuinely a problem for their timeline, better to learn it in week one than after four rounds.

The questions only they can answer, which are mostly about the process and the band. Ask what the stages are and how long the whole loop takes, who you will meet and in what order, whether the level is fixed or determined by the loop, what the compensation band for the level is, why the role is open, and how long it has been open. Those answers are cheap for a recruiter to give and expensive for you to guess. Save the team-health and technical questions for the hiring manager, who has the information; asking a recruiter about the team's architecture wastes a turn and produces a rehearsed answer. Ask what would make the hiring manager say yes, because recruiters often tell you.

By naming what you are moving towards rather than what you are escaping, and by being specific enough to be believed. "I have owned the same service for three years and I want a system at a volume where the design decisions change" is concrete and flattering to nobody. Some negative reasons are perfectly sayable if stated as structure rather than grievance: a reorg that removed the work you were hired for, a company that has stopped investing in your area, a promotion path that does not exist at your employer's size. What loses you a screen is naming a person, describing your manager, or listing several complaints, because the recruiter's only data point on your judgement is how you talk about people who are not present.

Three things, all avoidable. Naming a compensation expectation far outside the band with no flexibility signalled, which closes the file that afternoon regardless of your answers. Being unable to say anything specific about the company or the role, which reads as a mass application and lowers the effort anyone downstream will spend on you. And treating the recruiter as an obstacle — short answers, visible impatience, correcting their terminology — when they are the person who will argue for your level and carry your case to the panel. The fourth, less obvious one is contradicting yourself between your CV and the call on dates, titles or scope, because the recruiter reconciles the two while you talk and cannot advance a candidate whose account does not line up.

It buys you a phone call and nothing else, and it is read for about a minute by someone deciding whether an hour is worth spending. That means its job is to make the match obvious in the top third: current level, domain, the technologies that appear in the advert, and two or three outcomes with numbers on them. Later in the loop its role changes completely — it becomes the source of the questions you will be asked, because interviewers with no time to prepare pick a line and push on it. Both facts point the same way: put on it only what you would like to spend twenty minutes defending, and put the thing you most want to be asked about where it will be seen first.

Unevenly, which is what catches people. Most lines are never mentioned; a small number are examined in detail, and you cannot predict which. Anything phrased as leadership will be probed for what you personally decided, any number will be asked how it was measured, and any technology listed is fair game for a question about a failure mode. The pattern to expect is not an accusation but a drill: what was your part, who else was involved, what did you do when it went wrong. Claims that survive are ones where you can name the mechanism and the messy detail. Claims that collapse are the ones inherited from a team's achievement, and the collapse is audible, because the answer thins out at exactly the third follow-up.

The same work three times, and only the third is worth a question.

RESPONSIBILITY - describes the job description, not you
  "Responsible for the payments settlement service and its
   integrations with third-party providers."

  Every person who has ever held the role could write this line. It
  contains no decision, no outcome and nothing to ask about, so it
  is skimmed.

ACTIVITY - describes effort, still no result
  "Worked on migrating the settlement service from batch processing
   to an event-driven architecture using Kafka."

  Better, because there is a decision in it. But "worked on" hides
  your part, and there is no evidence it succeeded.

EVIDENCE - decision, mechanism, measured outcome, your part
  "Led the migration of settlement from nightly batch to an event
   stream, cutting reconciliation breaks from ~40/month to under 5
   and month-end close from 3 days to 4 hours. Designed the
   idempotency and replay model; ran the dual-run cutover across
   two quarters with no customer-visible incident."

  Now there are five questions an interviewer can ask, and you want
  all five: how did you measure breaks, what did dual-run cost, how
  does replay avoid double payment, what went wrong, what was your
  part versus the team's.

The progression is from describing the role to describing what changed because you were in it. Recruiters and hiring managers read hundreds of the first kind, so the first kind carries no information — it tells them the title, which they already have from the job history.

The number is doing two jobs. It makes the claim assessable, and it demonstrates that you knew what you were optimising before you started, which is a seniority signal in itself. A candidate who cannot say what improved usually could not say what the target was either, and interviewers infer that quickly.

The last clause is the one candidates omit and the one that protects you. Separating what you designed from what the team delivered pre-empts the "what was your part" drill and makes the whole line more credible, not less. A line that claims everything invites the interviewer to find the seam; a line that draws the seam yourself does not.

Accurately, and with your contribution named narrowly, because a narrow true claim is worth more than a broad one you cannot hold. "I was one of four on the migration; I owned the idempotency model and ran the cutover" is a stronger sentence than "I led the migration", partly because it is true and mainly because it survives the third follow-up. Interviewers are not looking for sole authorship — almost nothing at senior level has any — they are looking for whether you can distinguish your judgement from your team's output. The candidate who claims the whole project and then cannot explain a decision made by someone else has done more damage than the one who claimed a quarter of it, because now every other claim is suspect too.

Either keep it and be ready to say when you last used it, or move it into a clearly labelled earlier section. What you must not do is leave it looking current, because an interviewer picking a keyword to open with will pick that one, and "I haven't used it since 2023" arriving after two vague answers reads as inflation rather than rust. The clean handling is pre-emptive: "that was my main stack until 2023 and I'd need a fortnight to be useful in it again, though the concepts are the same and here is what I'd expect to have changed." Stated first, that is honest self-assessment; extracted after you have struggled, it is a discovered gap, and the difference is entirely in who says it.

Put dates in months rather than years so nothing looks concealed, and give the short item a one-line reason on the CV itself. A nine-month role labelled "contract, fixed term" or "role eliminated in restructure, team of 30 to 8" stops being a question. A gap labelled "career break — caring responsibilities" or "six months, deliberate break after four years on call" does the same. The mechanism is that unexplained anomalies get filled in by the reader with something worse than the truth, and once the reader has invented an explanation you spend the screen dislodging it. One line, factual, no defence, no apology — and the same wording verbally, so the account matches.

Use your official title, then translate it in the line beneath if it is misleading. Internal titles are often meaningless outside the company — grades, invented ladders, "engineer II" where the market says senior — so a candidate who silently upgrades themselves is not lying about capability but is creating a mismatch with the reference check and the background check, which will pull the official record. The safe construction is the real title followed by the scope: "Engineer II — equivalent to senior; owned the settlement service and mentored two engineers." That gets you levelled on scope, which is what you wanted, and leaves nothing for verification to contradict. Level disputes are winnable on evidence and unwinnable once the paperwork disagrees with you.

Knowing which decision is being made in the room you are in changes what you say in it.

flowchart TD
    A[Recruiter screen<br/>level, band, logistics] --> B[Hiring manager call<br/>is this the role you want]
    B --> C[Technical rounds<br/>can you do the work]
    C --> D[Design or depth round<br/>at what level]
    D --> E[Debrief and levelling<br/>hire, and at which band]
    E --> F[Offer construction<br/>compensation inside the band]
    F --> G[References and background<br/>verification only]

The important property of this chain is that the level is decided before the number is, at the debrief, and the number is then constrained by the band the level carries. That is why arguing about compensation at the offer stage has a ceiling that was set two stages earlier, and why evidence of scope belongs in the design round rather than in the negotiation.

The hiring manager call is the stage candidates waste. It is the only point where someone with real information will answer questions about the work, the team and what the first year looks like, and it is usually treated as a soft round to be survived. Coming with three specific questions changes both what you learn and how you are remembered, because a candidate who interrogates the role reads as someone with options.

The final stage is verification and not assessment, which matters for a practical reason: by the time references are taken the decision is made, so an inconsistency found there does not cost you a debate, it costs you the offer. That is the whole argument for the CV and the screen matching the record exactly.

The stage that is missing from most candidates' mental model is the debrief. Nobody in it has met you since their own round, so what circulates is the written scorecards plus the recruiter's summary from the screen. Your framing outlives you in that room, which is the strongest reason to get the ninety seconds right.

Read the verbs and the ownership, not the title or the years. "Contribute to", "work with" and "support" describe a mid-level role however senior the title says; "define", "own", "set the direction for" and "across teams" describe a staff-shaped one. Then check for the tell-tale inflation: a "senior" role that lists a specific framework in the first requirement and no ambiguity anywhere is usually a mid-level role with a title used as a substitute for money. The other direction exists too — a "senior engineer" advert that expects you to hire, set strategy and own a domain is a staff or lead role written by someone who could not get the headcount approved at that level, and the salary band will confirm which it is if you ask.

That something in the requisition does not clear the market, and it is worth finding out which thing before you invest four rounds. The benign explanations are a genuinely narrow skill combination, a slow internal process, or an approval that lapsed and was reissued. The costly ones are a band set below the level being asked for, a hiring manager who cannot decide, a bar that no candidate has met because the loop is broken, or a role that already burned through a hire who left inside a year. Ask the recruiter directly how long it has been open and how many candidates reached final stage, then ask the hiring manager what has been missing from the people they have seen. Both will usually tell you, and the answer is diagnostic.

The value is not the question, it is knowing in advance what a bad answer sounds like.

1  "Why is this role open - is it growth or a backfill?"
   good     new team, funded scope, someone promoted internally
   bad      "the last person moved on" with no elaboration, or
            three people have held it in two years
   tells    turnover in the seat, and whether the role is real

2  "What does the first ninety days look like, concretely?"
   good     a named first project, an owner to learn from, a
            definition of done
   bad      "hit the ground running", "you'll shape it yourself"
   tells    whether the work exists yet, or you are being hired
            into an idea

3  "What has stopped this team shipping in the last six months?"
   good     a specific dependency, a migration, a named constraint
   bad      "nothing really" or a blame-shaped answer about another
            team or a previous engineer
   tells    self-awareness, and how blame moves here

4  "How does on-call work, and when were you last paged?"
   good     rota size, compensation, and a real answer with a date
   bad      "it's pretty quiet" from someone not on the rota
   tells    whether the manager shares the load they are describing

5  "How were the last two promotions on this team decided?"
   good     names the ladder, the evidence, the calibration forum
   bad      "we don't really do levels" or "when the business allows"
   tells    whether progression is a process or a favour

6  "What would make you regret hiring me a year from now?"
   good     a specific risk about the role's ambiguity or the domain
   bad      a compliment, or an answer about attitude
   tells    whether they have thought about fit or just about filling

7  "How much of the team's time goes to unplanned work?"
   good     a rough percentage and a reason
   bad      no idea, or "we're agile so it varies"
   tells    whether anyone is measuring the thing that will eat
            your quarter

The design of the list is that every question is answerable from the manager's own recent memory. Questions about culture and values produce recruitment copy, whereas "when were you last paged" and "how were the last two promotions decided" cannot be answered from the brochure, so you get either a real answer or a visible absence of one — and the absence is itself the information.

Two of these are load-bearing. The one about the last six months is where you learn how failure is discussed, and a manager who reaches for another team's fault in a first conversation with a stranger will do the same when it is your project. The promotion question is where you learn whether the level you accept is a ceiling, because a company that cannot describe its last two promotions does not have a mechanism, whatever the ladder document says.

The regret question does more work than its phrasing suggests. It is hard to answer with flattery, and the good version tells you the risk the manager has privately identified in the role — which is exactly the thing you want to weigh before accepting, and the thing nobody volunteers.

Ask three, not seven. The list is for choosing from; asking all of them turns a conversation into an audit and burns the goodwill you were building.

Individually these are noise. The value is in the mapping, because each one points at a specific defect you can then test for.

observed in the process                signals
-------------------------------------  ---------------------------------------
Scheduling slips twice, no apology     the role is not a priority, or the
                                       team has no operational discipline

Nobody can describe the role the       no agreed scope; you will be hired
same way twice                         into a negotiation between managers

Interviewers have not read your CV     you are volume, not a candidate; the
or the JD                              debrief will be thin and arbitrary

"We move fast" used to explain a       absence of process is being sold as
missing process                        a value

Pressure to decide in 24 hours         they expect to lose you on reflection,
                                       or a competing internal candidate
                                       exists

Offer verbal only, written one         approval is not actually in place, or
delayed a week                         terms are still being argued internally

Level or title left vague until         levelling will land at the bottom of
after you accept                       the band and be revisited "later"

Salary band will not be disclosed       band is below market, or they price
at any stage                           per candidate rather than per level

Current employees will not be made      the reference works only in one
available to speak to                  direction

Everyone you meet has been there        either a real retention problem or
under 18 months                        recent hypergrowth. Ask which

Manager cannot name the last person     high churn being smoothed over
who left the team and why

Questions about hours or on-call        the answer is bad and known to be
answered with a joke                   bad

The reason to hold this as a table rather than an instinct is that each row has a follow-up attached. A flag is a hypothesis, and the correct response is a question rather than a withdrawal: ask directly why scheduling slipped, ask the hiring manager and the skip-level to describe the role separately and compare, ask what the band for the level is and what determines position within it.

Two rows should end a process on their own in most cases. A twenty-four hour deadline on a life decision is a technique rather than a constraint, and the honest test is to say you need until the end of the week — a real offer survives that and a manufactured urgency does not. Refusing to put a level or a title in writing before you accept is the other, because everything you were told about progression depends on where you land.

The rows about churn are the ones candidates rationalise. A team where nobody has been present for two years has either grown very fast or lost people steadily, and the two are trivially distinguishable by asking how many of the team's current members are backfills. The question is not rude, and the reluctance to answer it is the answer.

Ask for the mechanics rather than the experience, because mechanics cannot be softened. How many people are on the rota, how often does a shift come round, what is the paging volume out of hours in a typical week, is it compensated in pay or time, who is the escalation, and what happens to your sprint commitments during a shift. Then ask the manager when they were last paged themselves. Volume answers like "it's quiet" mean nothing without a rota size, since quiet across twelve people is a different job from quiet across three. Asking this carefully is not a negative signal; it reads as someone who has carried production before, and a team with a healthy rota enjoys answering. Discomfort is the finding.

Ask for descriptions of recent, specific events rather than for judgements. What was the last incident and what changed afterwards; what did the team ship last quarter that they were pleased with; how long does a small change take from merge to production; what is currently the most annoying part of working here. Each is a question about a fact, so the answer is either concrete or evasive, and both outcomes are useful. The framing matters: asked as curiosity about how the work happens, these are the questions of a candidate who has options, which is how you want to be read. Asked as a list of tests with no reaction to the answers, the same questions read as an interrogation and the room closes.

The hiring manager decides to hire, the panel advises through written scorecards, and a levelling or calibration forum usually decides the level, which is the decision that sets your band. The recruiter decides almost nothing about the outcome and controls almost everything about the process — scheduling, framing, what gets escalated, and how your case is argued when it is not in the room — which is why treating them as an ally rather than a gatekeeper pays. Finance or a compensation team owns the exceptions, so an out-of-band request is a different approval from a within-band one and takes longer. Knowing this tells you where to spend effort: evidence of scope goes to the panel, and the negotiation goes through the recruiter to whoever owns the exception.

Because it is the cheapest way to price you below what the role is worth, and because it filters out candidates the band cannot reach. Both motives are real and only one is adversarial. A recruiter working from your current number will construct an offer as a percentage uplift on it, which anchors your pay to your last employer's decisions rather than to the value of this role, and compounds for years because every future raise is a percentage of that. In several jurisdictions the question is now unlawful for exactly this reason, and in others it is simply asked out of habit. The productive response is neither to answer nor to refuse, but to redirect to expectations for this role, which is information they genuinely need.

Same question, three replies, and the ranking is by what each does to the number you end up with.

Q: "What are you earning at the moment?"

WORST - answer it plainly
  "I'm on £78,000 plus a 10% bonus."

  outcome  the offer is now built as £78k + 8-15%. You have priced
           yourself using your last employer's budget. If you are
           underpaid, you stay underpaid, and no evidence you
           present later moves the anchor much.

MIDDLE - refuse without redirecting
  "I'd rather not share that."

  outcome  legitimate, and slightly costly. The recruiter still
           needs a number to check you against the band, so the
           conversation stalls and they either ask again or move on.
           Reads as guarded rather than prepared.

BEST - redirect to the role, with a researched range
  "I'd rather not anchor on my current package - it's a smaller
   company and a different scope. For a role at this level I'm
   looking at £95,000 to £110,000 base, and I'd flex on where I
   land depending on the level and the rest of the package. What's
   the band you're working with for this one?"

  outcome  gives them what they actually need - a range to check
           against the band - without giving them your history.
           Ends with a question, so the next number spoken is
           theirs. If your range is above the band you find out in
           the first call rather than the fifth.

IF PRESSED A SECOND TIME
  "I understand you need to check I'm in range - my expectations
   are the £95k to £110k I mentioned, and that's the number that
   matters for this role. Is that workable?"

  Say it once more, calmly, and do not apologise. Repeating the
  same sentence without escalating is what makes it a boundary
  rather than a negotiation.

The ranking is about anchoring rather than about honesty. All three answers are truthful; they differ in which number is on the table when the offer is built, and the offer is nearly always constructed from the first number that was said out loud with any authority behind it.

The middle answer is worth understanding because well-prepared candidates land there and think they have won. A flat refusal protects your history and leaves the recruiter unable to do their job, so the conversation either loops back or you are quietly deprioritised in favour of someone whose fit is known. The fix is not to give more, it is to give something different — expectations rather than history.

The best answer works because it supplies the function of the question without its substance. The recruiter needs to know whether you are affordable; they do not need to know what you earn. Ending on "what band are you working with" matters as much as the range itself, since a candidate who asks that question is routinely told the band, and the band is the single most useful number in the whole process.

The second-pressing script exists because the question usually comes twice, and the second time is where people fold. Repeating your position in the same words with no new justification is the whole technique. Adding reasons invites argument, and apologising signals that the boundary is available for negotiation.

Triangulate from three kinds of source, because each is biased differently. Aggregated self-reported data — Levels.fyi, Glassdoor, local equivalents — is skewed towards the well-paid and the recently hired, so read the distribution rather than the average and discount the top decile. Advertised bands for the same level at comparable companies are more reliable where salary transparency laws apply, and are worth collecting for the specific title. Then talk to two or three people at the level you are targeting, since human sources are the only ones that tell you where a band actually sits versus where offers land inside it. Adjust for company stage, location and whether the total includes equity. The output you need is a range with a floor you would decline below, not a single number.

Whoever has better information, which is usually the employer, since they know the band, the level and what the last three hires accepted. That is why the standard advice to make them speak first is sound as a default. But it inverts once you have researched properly: if you know the market for the level, naming a well-supported range anchors the construction of the offer in your favour, and waiting merely hands the anchor to the party who was going to open low anyway. The practical rule is to try to get the band first, and if that fails, give a researched range rather than a point, with a reason attached to the top of it. What loses money is naming a number you have not researched, in either direction.

Give a range whose bottom is a number you would genuinely accept, because the bottom is what you will be offered. Attach a basis to it — the level, the market data, the scope you have been hearing about — and attach a condition, so the number is not final: "based on what I know about the level, £95,000 to £110,000 base, and I would want to see the full package before settling on where in that I land." That construction gives them something to work with, protects you from being held to a point estimate before you know what the equity and bonus look like, and leaves the level argument open. Never state a range whose bottom you would resent, and never say "at least" — it turns the floor into the target.

Base salary is the only fully reliable component, because it is contractual, paid regardless of performance, and forms the basis of every future raise. A bonus is semi-reliable, and the question that determines which is whether it is individual, company-performance-dependent or discretionary, plus what it has actually paid out for the last three years. Equity ranges from near-cash for public restricted stock units to lottery ticket for private-company options. Pension or provident contributions are real money and frequently ignored, and a few percentage points of employer contribution can outweigh a base difference. Allowances, insurance, learning budgets and sign-on bonuses are one-off or in-kind and should be counted at what you would actually pay for them, which is often less than the stated value.

Public equity is close to cash: restricted stock units vest on a schedule, convert to shares with a known market price, and can be sold in a window. The only real risks are the share price and the vesting period. Private equity is a claim on a future event that may never happen. Options give you the right to buy at a strike price, which means you pay to exercise, often owe tax at exercise on a valuation nobody can sell at, and hold an illiquid asset until an acquisition or listing. The paperwork also matters more than the number: liquidation preferences, participation, the size of the preferred stack ahead of you and the exercise window after you leave can all take the value to zero while the company is still described as successful.

The offer says the grant is worth £120,000. The arithmetic behind that number is where the whole answer lives.

The offer letter
  20,000 options, strike £2.00, current preferred price £8.00
  "£8.00 - £2.00 = £6.00 x 20,000 = £120,000 over 4 years, £30k/yr"

WHAT THAT NUMBER ASSUMES
  1  you stay 4 years           -> 25% of it is real at the cliff
  2  the £8.00 holds or grows   -> it is a preferred-round price,
                                    not a market price
  3  you can exercise           -> 20,000 x £2.00 = £40,000 cash,
                                    from you, up front
  4  no tax at exercise         -> in many regimes the spread is
                                    taxable at exercise, on paper
                                    gains you cannot sell
  5  common ranks with preferred -> it does not
  6  a liquidity event happens   -> median outcome is none

SAME GRANT, PLAUSIBLE OUTCOME
  Company raises £60m at higher valuations. Preferred stack with a
  1x non-participating preference now totals £150m.
  Acquisition price: £140m. A good outcome. Press release says so.

  proceeds to preferred    £140m  (preference paid first)
  proceeds to common         £0
  your 20,000 options        £0
  your exercise cost, if paid £40,000 -> lost

ANOTHER PLAUSIBLE OUTCOME
  You leave after 2.5 years. Vested: 12,500 options.
  Post-termination exercise window: 90 days.
  To keep them: £25,000 cash within 90 days, plus tax on a
  £75,000 paper spread. Most people cannot, so the options lapse
  and the 2.5 years produced nothing.

WHAT TO ASK BEFORE VALUING IT AT ANYTHING
  strike price and current valuation, and the date of that valuation
  total shares outstanding, fully diluted - percentage, not count
  total preferred raised and the preference terms
  post-termination exercise window
  whether secondary sales have ever been permitted

The first arithmetic is the one recruiters do, and it is not dishonest so much as conditional on six things that are rarely stated. Each assumption is individually plausible and jointly unlikely, which is why the expected value of a private grant is a small fraction of its quoted value even at a company that succeeds.

The preference stack is the assumption candidates have never heard of and the one that most often zeroes the number. Investors are paid before common shareholders, so a company sold for less than the money raised against it can produce a celebrated exit and nothing at all for employees. Asking what has been raised and on what terms is a reasonable question, and a company that will not answer has told you how to value the grant.

The exercise window is the second silent killer, because it converts your equity into a demand for cash at the worst possible moment — when you have just resigned. A ninety-day window on options with a large spread is functionally a retention device rather than compensation, and a ten-year window is a genuinely different and much better offer with the same headline number.

The practical conclusion is not to ignore equity but to price it as a lottery ticket with a known cost, and to decide the offer on base plus reliable cash. Saying so out loud in a negotiation is also effective: "I value the grant at close to zero for planning purposes, so I need the base to stand on its own" is an argument recruiters at private companies hear regularly and can act on.

Vesting is where a headline number turns into an actual entitlement, and the first year is not linear.

Grant: 4,000 RSUs. 4-year vest, 1-year cliff, quarterly thereafter.
Grant date 1 Sep 2026. Share price at grant £25 -> headline £100,000.

date          event                     vested   cumulative  if you left
------------  ------------------------  -------  ----------  -----------
1 Sep 2026    grant                          0           0%          £0
1 Mar 2027    6 months in                    0           0%          £0
31 Aug 2027   CLIFF - 12 months          1,000          25%     £25,000
1 Dec 2027    quarter                      250          31%     £31,250
1 Mar 2028    quarter                      250          38%     £37,500
1 Sep 2028    24 months                      -          50%     £50,000
1 Sep 2029    36 months                      -          75%     £75,000
1 Sep 2030    fully vested                   -         100%    £100,000

WHAT THE SHAPE MEANS
  leaving at month 11    -> £0. Eleven months of the grant, gone.
  leaving at month 13    -> £31,250 already yours
  a new grant only starts vesting from its own grant date, so
    switching jobs at month 11 forfeits a year twice over: the
    unvested grant here, and the cliff at the new employer

REFRESH IS THE PART NOT IN THE LETTER
  Year 1 income includes 25% of this grant. So does year 4.
  With no further grants, total comp is flat then falls off a
  cliff in year 5. Ask: is there an annual refresh grant, what
  did people at my level receive last year, and is it a
  percentage of salary or discretionary?

  A £100k grant with annual refreshes is worth roughly double a
  £100k grant without them over four years. The offer letter
  looks identical.

The cliff is the part that changes decisions rather than just arithmetic. Twenty-five per cent arriving in a single day at month twelve means the value of staying is wildly non-linear, and a candidate weighing a move at month ten is often being offered less than they are forfeiting. Knowing your own cliff date before you start interviewing is basic hygiene.

The quarterly steps after the cliff are the reason "how much unvested equity am I walking away from" is a question with a moving answer. Some employers will buy out a forfeited grant as a sign-on award, and asking is normal — but the ask has to be specific and evidenced, which means bringing the schedule rather than a claim.

The refresh question is the one that separates people who understand equity from people who read the headline. Two offers with identical grants can differ by a factor of two over four years depending on whether refreshes exist, and the offer letter will not say. Ask what someone at your level received as a refresh last year, and treat a vague answer as a zero.

The share price column is illustrative only for public stock and fictional for private, which is worth stating in your own head each time you read a schedule. Units vest; value does not.

Because the number is set once and the level compounds. Your level determines the band you sit inside, the size of your annual increase, the size of any equity refresh, the work you are given, and how long it takes to reach the next level — so a candidate who accepts the top of a mid-level band is behind a candidate at the bottom of the senior band within about two years, and the gap widens. Down-levelling with a promise of quick review is the specific trap, because promotion requires evidence of the next level's scope and you will be given the current level's work. If you believe the level is wrong, argue it before the debrief with evidence of scope, since afterwards you are asking to reopen a decision several people already made.

Negotiating supplies a reason and asks once; haggling repeats a demand and splits differences. The distinction is observable to whoever is on the other side: a negotiation gives them something to take to whoever approves exceptions — market data for the level, a competing offer, unvested equity being forfeited, scope beyond the job description — while haggling gives them nothing but your preference, and preference is not an approval case. The practical difference is in the number of rounds and the ending. A negotiation is one considered ask, possibly one adjustment, then acceptance of the answer. Haggling is three successive nudges, which spends the goodwill you will need in your first month and reliably produces the impression that you are difficult before you have started.

More than candidates assume, because the levers with the least approval friction are the non-base ones. A sign-on bonus is the easiest yes, since it is one-off and does not move your position in the band or affect anyone else's parity. A buy-out of forfeited unvested equity is a recognised category with a specific evidence requirement. Start date, notice-period accommodation, remote or hybrid days, an equity grant size, a guaranteed first-year bonus, relocation support, a learning or conference budget and an early review date are all commonly moved. Level is negotiable before the debrief and hard afterwards. Job title is often free. What is rarely moved is anything company-wide: pension percentage, holiday allowance, notice terms and benefits structure.

Written rather than spoken, because a written ask can be forwarded to whoever approves it.

Subject: Re: Offer - Senior Engineer, Ledger team

Hi Priya,

Thanks for sending this through, and for how straightforward the
process has been. I want to say clearly up front: I want to accept,
and I'm expecting us to land this. There's one part I'd like to work
on before I sign.

The offer is £96,000 base with a 10% target bonus and a £40,000 RSU
grant over four years.

On base, I've been working from three things. Advertised bands for
senior backend roles at comparable London firms this quarter are
£105,000 to £120,000. I'm also leaving £18,000 of unvested RSUs
behind - my cliff is in March and I'm walking away from it. And the
scope we discussed in the design round is broader than the advert:
the ledger reconciliation work sits across the payments and finance
teams, which was the level of ambiguity we talked about as the
staff-adjacent part of the role.

What would make this straightforward for me is £108,000 base. If
base is constrained by the band, a £15,000 sign-on to cover the
forfeited equity plus £102,000 would work equally well - I'm easy
about which route, and I'd rather give you options than a single
number.

Everything else in the offer is fine as it stands, and I'm not
looking to reopen anything else.

Happy to talk it through on a call if that's easier. I'd like to get
this signed this week.

Best,
Aran

Four properties make this a negotiation rather than a demand. It opens by stating the intention to accept, which removes the recruiter's main fear and changes what they will risk for you. It gives reasons that can be forwarded — a market range, a specific forfeited amount, an observation about scope from a round they can check. It offers two routes to the same outcome, which lets the recruiter pick the one their approval process actually permits. And it closes everything else, so nobody worries that agreeing will produce a second ask.

The equity forfeiture is the strongest of the three reasons because it is exact and verifiable, and because it maps to a category that already exists in most compensation policies. Vague appeals to market value get discounted; a number with a date attached gets an approval request written.

Naming a single figure rather than a range at this stage is deliberate and the reverse of the advice for the screen. Early on, a range keeps you in the process; at offer stage a range is read as its bottom, so the ask is a point with an alternative construction beside it.

The tone is doing real work. This email will be read by a hiring manager who is deciding how much political capital to spend, and a candidate who is warm, specific and closing is a much easier case to argue than one who is aggrieved. Nothing in it threatens to walk, because you cannot walk credibly here and a bluff spent now cannot be recovered.

Most candidates negotiate from a single offer, and the belief that this leaves you powerless is what produces the accepted first number.

Situation: one offer. No other process live. You want the job and
you have no alternative to point at.

WHAT NOT TO SAY
  "I'm speaking to a few other companies" - if untrue it can be
  checked more easily than people think, and if they ask which,
  the negotiation ends and so does your credibility.
  "I need more" with no basis - nothing to approve.

THE ASK
  "I'm going to accept this, so I'll be straightforward: I'm not
   running another process, and I'd rather tell you that than
   imply otherwise. What I'm asking for is based on the level
   rather than on an alternative.

   Two things. The advertised range for this level at comparable
   firms sits at £105k to £120k and the offer is £96k, which is
   below the bottom of that. And in the second round we agreed the
   role includes the reconciliation work across finance, which is
   wider than the advert.

   £106,000 makes this an easy yes for me today. If the band caps
   below that, I'd take £100,000 with the level set at senior two
   rather than senior one, or £100,000 with a £12,000 sign-on.
   Any of the three works."

WHY THIS HOLDS WITHOUT LEVERAGE
  - the reasons are external to you: a band, a scope, a market
  - three routes means the recruiter can find the one that is
    approvable rather than having to say no
  - honesty about having no alternative buys credibility that a
    thin bluff destroys
  - the real leverage is unstated and genuine: they have spent
    five rounds and would rather pay 6% more than restart

REALISTIC OUTCOMES
  best      £106,000, or £100k plus the level. Common.
  likely    £100,000 to £102,000, or the sign-on. Very common.
  floor     "£96k is the band ceiling for the level." Then ask
            what the review cadence is and get it in writing.
  feared    offer withdrawn. Effectively unheard of for a single
            reasoned ask made warmly.

The premise to reject is that leverage means a competing offer. Your actual leverage is the cost of not hiring you, which by the offer stage is five rounds of engineering time, a hiring manager with an unstaffed roadmap and a requisition that may lapse. Nobody restarts that over a few per cent, which is why the feared outcome sits where it does.

Declaring that you have no alternative is counter-intuitive and usually correct. A bluffed competing offer invites one question — from whom, and at what — that you cannot answer, and once the recruiter concludes you are inventing leverage they will hold the number and you will accept it. Trading the bluff for credibility makes the reasons you do have land harder.

Three routes is the mechanism that most often converts a no into a yes. A recruiter told "£106,000" can only relay a request that may be outside their authority; a recruiter given base, level and sign-on options can find the door that is already unlocked, and level is frequently the cheapest of the three for them and the most valuable of the three for you.

The floor response matters as much as the ask. If the band genuinely caps, the useful next move is not to push but to convert the conversation into a written commitment about the first review — date, criteria, who decides — which is worth more than the difference you were arguing over and costs them nothing today.

The order matters more than the wording, because two of these steps are irreversible once passed.

flowchart TD
    A[Offer received<br/>get it in writing first] --> B[Thank them<br/>ask for a few days]
    B --> C{Is the level right}
    C -->|no| D[Argue level with scope evidence<br/>before discussing money]
    C -->|yes| E[Price the whole package<br/>base, bonus, equity, pension]
    D --> E
    E --> F[One reasoned ask<br/>with two or three routes]
    F --> G{Response}
    G -->|moved| H[Accept and stop<br/>confirm in writing]
    G -->|firm at band ceiling| I[Get review date and criteria<br/>in writing, then decide]

The first box is the one candidates skip. Negotiating against a verbal offer means negotiating against a number that can quietly change, and asking for the written version is normal, so there is no cost to waiting for it. Everything you agree after this point should end up in the same document.

The level branch comes before the money branch for a structural reason: the level sets the band, so winning the level argument often moves more money than winning the base argument, and it cannot be revisited once the offer is built. Evidence here is scope from the rounds you have just done, which is why the design round is part of your negotiation whether you treated it that way or not.

The single-ask box is the discipline that keeps this a negotiation. One well-reasoned ask with alternative constructions gives the recruiter room to manoeuvre; a sequence of separate asks over four days makes them defend rather than advocate, and the second ask is where goodwill starts being spent.

The right-hand terminal is the one worth rehearsing, because it is the commonest real outcome. If the answer is a genuine band ceiling, the value left on the table is the review — a date, written criteria and a named decision-maker — and a company that will not put any of that in writing has told you what the informal promise was worth.

State it factually, once, without threatening. "I have another offer at £112,000 base; I would rather be here and I am asking whether you can get closer" gives the recruiter a fact to take upward and leaves you a route to accept either way. Do not name the other company if you would rather not, but be prepared to confirm the number is real, because inflated figures are checkable through a surprisingly small industry. Never use a competing offer as a deadline unless the deadline is genuinely theirs and you will honour it. The tactic that backfires is running an auction: going back a third time with a revised competing number converts a negotiation into a bidding process, and the company that wins it often does so resentfully.

Because a deadline is a commitment to act, and one you abandon destroys the only leverage you had. "I need an answer by Friday or I will accept the other offer" followed by you still negotiating on Tuesday teaches the recruiter that nothing you say has a consequence, and the rest of the conversation is priced accordingly. It also removes their flexibility: an artificial deadline inside someone's approval cycle can force a no that a week would have turned into a yes. The honest construction is to state the real constraint and its source — the other company has asked for a decision by the eighteenth — and to ask what is achievable in that window. Real constraints are respected; manufactured ones are tested.

One ask, one adjustment, then accept the answer. Beyond that the returns turn negative, because you are no longer moving the number, you are shaping the impression your future manager forms before you arrive. The exception is when new information appears — a competing offer that genuinely lands, a discovery that the level was set below what was discussed — in which case say explicitly that something has changed rather than presenting it as another push. What signals inexperience is the salami approach: agreeing base, then asking about sign-on, then about start date, then about title, each in a separate message. Put everything in one email, name what you are not asking about, and let the recruiter close it in a single approval.

When the thing that is wrong is structural rather than numerical. A band that caps below your floor after a genuine attempt, a level you believe is wrong that they will not revisit, a refusal to put the title, level or reporting line in writing, or a process that has shown you how the company behaves under mild pressure — those are reasons, and they do not improve after you join. A number you are mildly disappointed by is usually not a reason on its own, provided the level is right and the review mechanism is real. Decide your floor before the offer arrives, in writing, to yourself, because the moment of the offer is the worst time to work out what you will accept. Then decline warmly and specifically enough to leave the door open.

The headline base differs by nine thousand in one direction and the package differs by twenty-two in the other.

                                    OFFER A          OFFER B
                                    Series B, 60 ppl  listed, 4,000 ppl
----------------------------------  ---------------  ----------------
Base                                    £105,000          £96,000
Bonus target                             none              15%, paid
                                                           4 of last 4 yrs
Equity, annual value                    £30,000 stated    £18,000 RSUs,
                                        private options    public, liquid
Equity refresh                          none confirmed     ~£12,000/yr,
                                                           standard at level
Pension, employer                        3%                10%
Private medical, family                  self-funded       included
                                         ~£2,400/yr

HEADLINE TOTAL                          £138,000          £143,900
                                        looks close        looks close

RE-PRICED ON RELIABILITY
Base                                    £105,000          £96,000
Bonus at historical payout                     £0          £14,400
Equity at defensible value              £0 - £3,000       £18,000
  A: options, £40k to exercise,
     preference stack above common,
     90-day window. Priced near nil.
  B: vested quarterly, sellable.
Refresh, year 2 onwards                        £0          £12,000
Pension                                   £3,150           £9,600
Medical                                        £0           £2,400

RELIABLE YEAR-1 TOTAL                   £108,150         £140,400
RELIABLE YEAR-2 TOTAL                   £108,150         £152,400

Difference in year 2: £44,250 in favour of the lower base.

The arithmetic is not the point; the re-pricing rule is. Each component is restated at what it will probably pay rather than at what the letter claims, and almost the entire gap comes from three lines candidates habitually ignore: pension, bonus history and equity liquidity. Seven percentage points of pension alone is £6,450 of real money that never appears in a comparison of base salaries.

The equity line is where the offers actually separate. Offer A quotes a paper value that requires £40,000 of your cash to realise, sits behind a preference stack and expires ninety days after you leave, so pricing it near zero is not cynicism but arithmetic. Offer B's grant vests in tradable units on a known schedule. The two numbers are not the same kind of object and adding them into a single "total compensation" figure is the error the comparison exists to prevent.

The refresh row is what makes year two decisive, and it is invisible in most offer comparisons because it is not in either letter. Ask for it explicitly at both companies. A company that grants annually at your level has a compounding package; one that does not has a package that falls in year five, at which point you are negotiating from inside rather than outside.

None of this settles the decision, and it is worth saying so in an interview context. Offer A may still be right for the scope, the learning or a genuine view on the company's trajectory — but it should be chosen as a £108,000 job with a lottery ticket, not accepted as a £138,000 one.

Money is one row of about eight, and writing the weights down before the offers arrive is what stops the most recent conversation deciding it.

Weights set BEFORE either offer arrived. Score 1-5.

factor                     weight  A     B     A wtd  B wtd
-------------------------  ------  ----  ----  -----  -----
Reliable comp, year 2         25%    2     5    0.50   1.25
The work itself               20%    5     3    1.00   0.60
Manager and how they          15%    4     3    0.60   0.45
  handled the process
Level and progression         15%    2     4    0.30   0.60
  is the level right, is
  there a real ladder
Learning and the people       10%    5     3    0.50   0.30
  you would learn from
Sustainability                10%    2     4    0.20   0.40
  on-call, hours, unplanned
Risk of the company            5%    2     5    0.10   0.25
  runway, funding, layoffs
                                                ----   ----
TOTAL                                           3.20   3.85

Tie-breaks, used only if within 0.2
  which one would you regret not taking in two years
  which one is easier to leave from - CV legibility
  which one did people who left speak well of

HOW TO USE IT
  Set the weights before you see the numbers, or the weights
  will be reverse-engineered from the offer you already prefer.
  If the result surprises you, the weights are probably wrong -
  fix the weights rather than the scores, and say what changed.

The mechanism is ordering, not arithmetic. Setting the weights first forces you to decide what this move is for — money, scope, learning, stability — and that decision is much easier in the abstract than while holding two offers and a deadline.

The two rows that carry more than candidates give them are progression and sustainability. A job you cannot be promoted out of is a two-year job whatever it pays, and a rota that pages you four times a week will decide how you feel about everything else within a quarter. Both are knowable in advance from the questions asked of the hiring manager, which is the practical reason those questions matter.

The instruction about surprise is the honest part. A scorecard is a thinking tool rather than an oracle, and if the total contradicts your instinct, the useful move is to interrogate the weights out loud — often you discover the real weight on "the work itself" is forty per cent, which is a legitimate answer and worth knowing about yourself.

Note what has no row at all: the brand name, the office, and how much you liked the recruiter. All three move decisions more than they should, and giving them no weight is a deliberate choice rather than an omission.

Four, in rough order of impact. Who you report to, because a manager decides your scope, your feedback and whether your promotion case is ever written. The work itself, meaning whether the problems are ones you get better at solving rather than ones you merely absorb. On-call and unplanned work volume, since that sets whether the job fits inside the week. And whether the company is growing, because growth creates scope and contraction removes it regardless of your performance. Everything else — office, brand, perks, technology choices — has a shorter half-life than candidates expect. The reason this matters at offer stage is that all four are researchable before you sign and effectively fixed afterwards, while the number is the one thing that can still be revisited later.

Separate the internal level from the external label, because they do different jobs. The internal level governs your band, your work and your promotion path, so that is the thing to argue about. The external title governs how recruiters read your CV in three years, which is not nothing — a title that inflates you sets up an awkward interview later, and one that deflates you costs you screens. Watch for titles used as substitutes for money at small companies, where "Head of Engineering" over four people is a genuine trap in both directions: it may be real scope or it may make your next move harder to level. Ask what the internal level is called, what the band is, and what the title maps to elsewhere.

The things that are expensive to discover later. Base, bonus mechanism and whether it is discretionary, equity type with grant size, strike price, vesting schedule and post-termination exercise window, pension percentage, start date, notice period, probation length and what notice becomes during it, and the reporting line and level in writing. Then look for the clauses candidates skip: a clawback on a sign-on bonus if you leave inside a year or two, a training bond, a non-compete or non-solicit with its duration, garden-leave provisions, and anything making the offer conditional. Confirm in the document anything agreed verbally, especially a review date or a level. An offer letter that omits a promise made on a call is the promise disappearing.

Verification and risk, not assessment, because the hiring decision was made before it was requested. What is being confirmed is that you held the role and the dates you claimed, that nothing disqualifying is known, and occasionally that a specific doubt from the debrief is not real. Which means the risk to you is almost entirely about consistency: a reference who describes a different scope or a different title from your CV creates a problem that no argument fixes at that stage. Choose referees who can speak to the work you actually described, tell them what the role is and what you claimed, and never list someone you have not spoken to. In many jurisdictions employers give only factual references by policy, so the absence of a glowing one means nothing.

Typically identity, right to work, employment dates and titles, education, and for regulated roles a criminal-record and financial check. The surprises are mundane rather than dramatic. Employment dates that disagree with your CV by a couple of months, because you counted from your offer rather than your start. Titles that differ from the official record. A degree classification or an awarding year that does not match. A gap you smoothed over. Any of these can trigger a discrepancy process even when the underlying facts are innocuous, because the checker's job is to report differences rather than to judge them. The protection is to write your CV from the record rather than from memory, and to flag anything odd yourself before the check runs.

Three to five working days is normal and asking for a week is rarely refused, particularly if you say what you are doing with the time. Longer is possible with a reason — a live final round elsewhere, a decision involving relocation or a partner — and honesty about it is more effective than vagueness, because a recruiter who knows your constraint can work with it. What you cannot do is go quiet: silence during a decision window is read as a candidate about to decline, and the hiring manager begins reopening the pipeline. If you are genuinely waiting on another process, say so and give a date. And treat a demand to decide within twenty-four hours as a data point about the company rather than as a constraint you must meet.

Short, factual and unarguable, because its only job is to start the notice period cleanly.

Sent after the new offer is signed and any conditions are cleared.
Told to your manager in a conversation first; this is the record.

Subject: Notice of resignation

Hi Marcus,

Following our conversation this morning, this is to confirm my
resignation from my role as Senior Engineer. My notice period is
three months, so my last day would be 27 October 2026.

I'd like to make the handover as complete as possible. I'll put
together a written list of what I own that nobody else does -
settlement, the reconciliation jobs and the two provider
integrations - with a proposal for who picks up each, and I'd
suggest we pair on the settlement runbook over the next few weeks.
Happy to be guided on the priorities.

Thank you for the last four years, and particularly for backing
the settlement rewrite when it was not an obvious call. I've
learned a great deal here.

Best,
Aran

WHAT IS NOT IN IT
  - where you are going, unless you choose to say
  - why you are leaving, or any comparison of the two roles
  - anything about pay, the counteroffer, or a grievance
  - conditions, hedging, or "unless something changes"

The conversation happens before the email and the email exists so the date is not disputed. Resigning in writing without speaking first is remembered, and resigning verbally without confirming in writing leaves the last day open to interpretation at exactly the moment when a fortnight matters.

The handover paragraph is the part that produces your reference and your reputation in a small industry. Naming the specific systems only you touch and proposing owners is worth more than an offer to "help with the transition", because it demonstrates that you are managing the risk your departure creates rather than leaving it for someone to find in November.

Everything in the omissions list is omitted for the same reason: nothing in this message should be negotiable or quotable. A resignation that names a grievance invites a conversation about the grievance, one that hedges invites a counteroffer, and one that says where you are going hands your manager and their network information you may not want circulating before you start.

The thanks are genuine and specific, and this is not sentiment. A one-line specific thank-you is the difference between a factual reference and a warm one, and the person you are writing to is likely to appear in your professional life again.

A counteroffer arrives when you have most leverage and least judgement, so the decision is worth having made in advance.

flowchart TD
    A[Counteroffer arrives] --> B{Why were you leaving}
    B -->|money only| C[Rare. Money fixed<br/>and nothing else was wrong]
    B -->|scope, manager, growth| D{Is that fixed in writing<br/>with a date}
    D -->|no| E[Decline. The offer buys<br/>time, not change]
    D -->|yes| F[Ask why it took a resignation<br/>to arrange it]
    C --> G[Consider, and expect to be<br/>seen as a retention risk]
    F --> G

The first branch is the whole decision. If you were leaving for money and only money, a counteroffer can genuinely solve the problem, and that case exists — it is just much rarer than the number of counteroffers accepted would suggest. Almost everyone leaves for scope, a manager, growth or exhaustion, and a counteroffer addresses none of those while feeling like it addresses all of them.

The written-and-dated test on the second branch is what separates a real change from a promise made under pressure. A new title, a named project, a different reporting line with a start date is a change; "we were about to promote you anyway" and "let's revisit in six months" is the same job with a raise and a manager who now knows you were looking.

The question in the fourth box is uncomfortable and clarifying. If the money was available all along, it was withheld until you forced it, which tells you how the next two years of raises will be decided. If it was not available and appeared overnight, someone has spent a favour to plug a hole, and favours are repaid.

The final box is the part nobody mentions. Whatever is agreed, you are now categorised as a flight risk, and that categorisation quietly affects what you are trusted with and who is first considered when a reduction comes. That cost is real, unmeasurable and worth pricing into the decision.

Because the money moves and the reason you were leaving does not. Most people resign over scope, a manager, a promotion that never arrived or unsustainable load, and a raise changes none of them — so the same dissatisfaction returns within months, now without the offer you turned down. Two further costs are structural. Your loyalty has been repriced as conditional, which affects what you are trusted with and who is considered first in a reduction. And your bargaining position is spent, since the next raise conversation cannot be repeated with the same lever. The industry pattern is consistent enough to be worth stating plainly: a large share of accepted counteroffers end in departure inside a year, usually on worse terms.

It is your employer keeping you on payroll and away from the work for part or all of your notice, typically to protect client relationships or unreleased information. Practically it means you are paid, still employed and therefore still bound by your contract — so you cannot start the new job, and often cannot work at all — which matters enormously to your start date. Check whether your contract permits it before you resign, because a three-month notice with garden leave is a three-month wait that the new employer must be told about. It is also a lever: employers will frequently shorten notice or release you early in exchange for a clean handover, and asking is normal. Get any early release in writing.

It varies enormously by jurisdiction and the honest answer is that you cannot resolve it from general knowledge. Some jurisdictions void them outright or restrict them to senior roles; others enforce narrow, short, compensated ones and strike down broad ones. What you can do without a lawyer is read what yours actually says: how long, what activity, what geography, whether it names competitors or a whole industry, and whether continued payment is required. Then check the non-solicit separately, since clauses about approaching clients and colleagues are more commonly enforced than clauses about working at all. If the new role sits inside the restriction, tell the new employer before you sign — they are usually relaxed and occasionally will indemnify you, and discovering it later is far worse.

A genuine handover and no sabotage, which is both the decent position and the self-interested one, since your reference and your reputation in a small industry are decided in these weeks. Concretely: a written list of what only you know, runbooks for the systems you are on call for, paired sessions on the parts that cannot be documented, an honest state-of-play on anything in flight, and no quiet coasting. What you do not owe is a promise to stay, a full account of your reasons, or your new employer's details. Do not take code, data or documents, and do not recruit colleagues during notice if your contract restricts it. The way you leave is the version of you that circulates, and it circulates for years.

Usually two things, neither of which they intended. That they had not researched the market for the level, which reads as the same lack of preparation an interviewer would penalise in a design round. And that they will accept a stated position without testing it, which is a mild negative signal for a senior role where the job includes pushing back on scope, estimates and priorities. The practical cost is separate and larger: the number becomes the base for every future percentage increase, and internal parity conversations later will place you at the bottom of the band with no way to correct it. One warm, reasoned ask costs you nothing and is expected — recruiters usually hold room for it, and an unused allowance is not returned.

Because it hands the decision to the party whose job is to minimise it, and it supplies no information they can act on. Fairness is not a number; the recruiter will convert your answer into the bottom of the band and record you as easy, because that is the only reading available. It also fails the seniority read: a candidate at senior level is expected to know what their scope is worth, and declining to say suggests either that you have not looked or that you are uncomfortable with the conversation — and discomfort in a negotiation reads across to how you will handle a stakeholder. The strong version is a researched range with a basis and a genuine flexibility about composition rather than about amount.

Because it is checkable, unnecessary and asymmetric in cost. The follow-up questions — which company, what level, what was the base — arrive immediately, and a fabricated answer collapses under two of them. Industries are small enough that a recruiter often knows the band at the company you named, and the moment they conclude you are inventing leverage, the number freezes and everything else you claimed is re-examined, including the parts on your CV. The asymmetry is what makes it a bad bet: the upside is a few per cent and the downside is a withdrawn offer or a story that follows you. Everything the bluff was meant to achieve is achievable with market data, forfeited equity and the cost of not hiring you.

It is a question about whether your trajectory fits the role, and it is failed in two opposite directions. An answer aimed at management when the role is a long-term individual-contributor position, or at architecture when the team needs someone to own delivery, tells the manager you will be dissatisfied inside a year. The opposite failure is an answer so vague — "growing, learning, taking on more" — that it says nothing and reads as either evasion or absence of direction. The workable answer names a direction rather than a title, ties it to what this role would build, and admits what you do not yet know: deeper technical ownership of a domain, having led something end to end, and an honest open question about whether management is for you.

"What are you looking for in your next role, and what would make you turn this one down?" It resists preparation because the second half requires you to name a real constraint, and it exposes whether you have thought about the decision or are simply available. A strong answer states two or three specific things — ownership of a system rather than tickets, a domain you want to go deeper in, a level with a described path — and then names a genuine dealbreaker calmly: an on-call rota below a certain size, a role with no scope for the reconciliation work discussed, a package that cannot reach a floor you can justify. It is concrete enough that the interviewer can tell you now whether the role fits. A weak answer is entirely positive, agrees that everything about the role sounds ideal, and has no dealbreaker — which tells the interviewer that you will accept anything, will negotiate from nothing, and may well leave when the thing you never articulated turns out to be missing. The underlying test is whether you are choosing a job or hoping to be chosen.