Presales and Solution Architect Interviews: The Complete Guide

What it is

A presales interview is an audition for a job whose entire output is other people's confidence. Nothing you produce in the role ships. You write no code that runs in production, you own no service that pages you, and the artefacts you leave behind are a demo somebody remembers, a proposal somebody signs, an estimate somebody is later held to, and a set of promises that an implementation team discovers nine months after you made them. Because the output is confidence, the interview cannot be an examination. It has to be a simulation, and it is graded the way a live customer meeting is graded: by what the people in the room believe when you stop talking.

That distinction explains almost everything strange about these loops. A backend engineering interview can hold a candidate at arm's length and ask them to solve a bounded problem, because the job is bounded problems. A presales loop cannot, because the job is unbounded problems asked by people who are not obliged to be reasonable, in a room where you have no time to look anything up. So the loop puts you in front of a panel playing customer roles, hands you an incomplete brief, and watches what you do when the brief turns out to be wrong. The technical questions are still there, but they are diagnostic rather than the point. What is being sized is your behaviour when the ground moves.

Every question in this guide sits somewhere on one path, which is the deal as the presales person owns it. The interview questions cluster at the gates rather than in the stages, because a gate is where a deal is lost by something a solution architect said or failed to say.

flowchart TD
  accDescr: The deal path drawn around its gates, qualification either walking away for no budget or no access or proceeding to discovery, then demo and hard questions and proof of value with exit criteria, and a security and compliance review whose control gap or blocked answer sends the deal back to the demo, while clearing it leads to estimate and internal sign-off, proposal and procurement, and signature and handover.
  Q{Qualify} -->|no budget or no access| WALK[No-bid or walk away]
  Q -->|proceed| DISC[Discovery]
  DISC --> DEMO[Demo and hard questions]
  DEMO --> POV[Proof of value with exit criteria]
  POV --> SEC{Security and compliance review}
  SEC -->|control gap or blocked answer| DEMO
  SEC -->|cleared| EST[Estimate and internal sign-off]
  EST --> PROP[Proposal and procurement]
  PROP --> CLOSE[Signature and handover]

Look at the two edges that do not point forwards. Deals die at the security gate and at qualification, and both of those are stages where the technical work was already finished and correct, which is why an interviewer who only tests your architecture is not testing the role.

The title varies more than the job does: sales engineer, solutions engineer, solution consultant, presales consultant, solution architect, principal architect, bid architect, field CTO. The work behind those titles differs along two axes. The first is whether you sell a product or sell delivery capacity, since a product vendor's sales engineer demonstrates something that exists and defends its edges while a services company's solution architect designs something that does not exist yet and defends a number attached to building it. The second is how far into delivery the role reaches. Some presales roles hand off at signature and never see the consequence; others stay accountable through the first phase, and that difference changes what you are permitted to promise and therefore what the interview probes.

The roles this guide covers, and where they diverge

A product-vendor sales engineer spends their week in demos, technical discovery, proof-of-concept scoping, security questionnaires and competitive displacement. The recurring hard moment is being asked in public whether the product does something it does not, so the interview is demo-heavy and will contain a live curveball.

A services or systems-integrator solution architect converts fuzzy requirements into an architecture, an estimate, a delivery shape and a written proposal that becomes a contract schedule. The recurring hard moment is a budget that cannot buy the scope, or a number delivery will not sign, so the interview is estimation-heavy and tests whether you understand that an estimate is a commercial commitment rather than an engineering opinion.

A bid architect works inside formal procurement: RFIs, RFPs, compliance matrices, clarification windows, red-team reviews, submission mechanics. The recurring hard moment is a scoring scheme that will not let you explain yourself, and the interview probes process discipline and the ability to allocate scarce senior time against published weights rather than against personal discomfort.

Many real roles are a blend, and a good interviewer tells you which blend before the loop starts. If they do not, ask. "Roughly what proportion of the role is demo and POC work against written proposal and estimation work?" is a fair question in the recruiter screen, and the answer tells you which half of this guide to weight.

The shape of the loop

Presales loops run four to six stages, and the middle one decides the outcome.

The recruiter screen establishes territory, deal size, quota model and whether you have carried a number before. Two things get graded: whether you can describe a deal in a way a non-technical person understands, and whether your account of why you are leaving is stable. Sales organisations are alert to candidates who present the customer as the problem.

The hiring-manager conversation tests technical credibility and commercial instinct together. Expect to be walked through a deal end to end and interrupted; the interruptions are the assessment, and a smooth uninterrupted narrative tells them nothing.

The demo or presentation round is the centre of the loop. You are given a scenario, sometimes a real product to learn, sometimes a brief to build a pitch against, and typically a preparation window between two days and a week. A panel plays the customer and is instructed to be difficult in specific, pre-agreed ways. Most rejections happen here.

A panel round splits the axes deliberately. An engineer probes depth, a sales counterpart probes whether you would be a partner they trust in front of a customer, and sometimes a delivery lead probes whether you would hand them something buildable. Their scores are compared afterwards, and disagreement between them is itself informative: a candidate the engineer loved and the account executive found evasive is a specific failure mode.

A closing conversation with a leader tests judgement rather than skill, with questions chosen to be big and unsafe: when would you refuse a deal, what would you do if you disagreed with your account executive in front of a customer, what does a bad quarter look like from your seat.

Why we need it

This interview exists in the form it does because the role sits on a fault line between two organisations measured on incompatible things, and the person on that fault line can do enormous damage quietly.

Sales is measured on booked revenue this quarter. Delivery is measured on whether what was sold can be built at the margin it was sold at. Presales is the only function that speaks both languages, and it is structurally pulled towards one of them, because it usually reports into sales and often carries a quota. A weak presales hire resolves that tension by telling the customer what wins the meeting and letting the consequence land elsewhere. The damage is invisible for two or three quarters and then arrives at once, as an unbuildable statement of work, a reference customer who will not take calls, a security answer quietly softened to unblock a deal, or a delivery team who distrust every estimate out of the field.

That damage is expensive in a way an underperforming engineer's is not. An engineer who writes poor code produces a defect that gets found. A presales person who overcommits produces a contract, and a contract is enforceable. This is why the interview is weighted so heavily towards behaviour in the awkward moment rather than towards knowledge. An interviewer can teach you the product in six weeks. They cannot teach you to say "it does not do that" out loud when four people are looking at you and the number is large.

The second reason is that the work is unobservable until it is too late. Almost every other technical function leaves an audit trail: commits, tickets, incidents, dashboards. Presales leaves a trail of meetings nobody else attended. Hiring managers therefore over-invest in the simulation, because it is the only direct evidence they will ever get. A candidate's CV can tell them the deals closed; only the demo round can tell them what the candidate does when the demo breaks.

The third reason is the one candidates most often miss. The role is a communication role with a technical entry requirement, not a technical role with a communication bonus. The applicant population skews towards engineers who want out of delivery, and that population is systematically stronger on the axis the job weights least. So the loop is designed to find the ceiling on the axis engineers rarely rehearse: what happens when the right answer is unwelcome, when the question is political rather than technical, and when the most useful thing you can say is a number you do not want to say.

Why it simulates instead of asking

Presales loops rarely ask "how would you handle X" and much prefer to make X happen, because every candidate can describe the correct handling of a mid-demo gap and very few perform it under load. The gap between described and performed behaviour is the entire predictive content of the interview, so a well-run loop closes off the described version. The consequence for preparation is severe: rehearsing answers is necessary and nowhere near sufficient, and you need to rehearse them out loud, at speed, while being interrupted.

The three axes interviewers grade

Almost every presales scorecard reduces to three axes, and how they interact matters more than being strong on any one.

Technical depth

Depth is graded as a floor, not as a maximum. The panel needs to know you cannot be embarrassed in front of a customer's architect, that you will recognise when a question is beyond you, and that you will not invent an answer. Past that floor, additional depth stops earning marks and can start costing them, because depth deployed at the wrong moment reads as an inability to judge the audience.

The engineer on the panel is looking for three things: whether your architectural vocabulary is load-bearing or decorative, tested by asking why rather than what; whether you know the boundaries of your own knowledge and mark them out loud; and whether you can hold a technical position against pressure when you are right, because a presales person who folds under a customer architect's scepticism is useless in a competitive evaluation.

The failure mode graded hardest is fabrication. An interviewer who catches an invented capability, benchmark or compliance certification will usually stop the loop, because the same reflex in front of a customer creates liability. "I do not know, and here is how I would find out by Thursday" is a full-credit answer, and it is worth rehearsing so that it sounds like confidence rather than retreat.

Commercial and business acumen

This is the axis that separates candidates, because it is the one engineers arrive without and the one the role cannot function without. It is not finance terminology. It is whether you reason in the customer's units and in your own company's units without being prompted.

The customer's units means describing a benefit as something on their profit and loss or their operational scorecard. "Deployment takes four minutes instead of forty" is a technical property. "Your release window stops needing the Saturday maintenance slot, which is where the change-failure incidents cluster" is the same fact in their units, and interviewers listen for whether the translation happens spontaneously.

Your own company's units means understanding that not all revenue is equal, that a heavily discounted deal with a bespoke commitment can consume more delivery capacity than it funds, and that some deals should be declined. A candidate who never mentions margin, delivery risk, or the cost of the response itself is scored as commercially naive regardless of how good the architecture was. The bid/no-bid decision is the cleanest test, and treating no-bid as the timid option rather than as capital allocation is a recognised tell.

Communication under load

The third axis is measured continuously, in every round, including the ones nominally about something else. It has four observable components, graded separately even when the scorecard collapses them into one line.

Structure is whether your answer has a shape the listener can follow: an answer first, then the reasoning, rather than a chronology that arrives at the point in the last sentence. Interviewers notice this within the first two answers and it colours everything that comes after, because a person who buries the conclusion in a low-stakes interview answer will bury it in a customer meeting too.

Concision is whether you can stop. Presales people who cannot stop talking lose deals by filling the silence a customer was using to decide, and by continuing past the point where they had won the argument into territory where they had not. A panel will sometimes leave a deliberate pause after your answer to see whether you fill it. Let it sit.

Audience calibration is whether you change register when the room changes. If the panel contains somebody playing a chief financial officer and somebody playing a platform engineer, and you answer both in the same words at the same depth, that is a scored miss even when the content was correct, because half the room was not served.

Handling of the unwelcome is the hardest component and the most predictive. It covers saying no, delivering a number nobody wants, contradicting the person who might become your boss, and staying composed when a demo dies. There is no substitute for having the sentences ready.

How the weighting shifts

The shift with seniority is the one to internalise. At entry and mid level you are graded on whether you can be trusted in a room; at senior and above, on whether you can be trusted to say the thing your own organisation does not want to hear, which is why the walk-away scenario appears reliably in principal-level loops and almost never in graduate ones.

The demo round

The demo round is the highest-variance stage in the loop and where preparation returns the most, and candidates routinely prepare for the wrong thing. It is not testing whether you can operate the software. It is testing whether you can build a narrative a specific audience recognises as their own problem, hold it under interruption, and behave well when something goes wrong. Product knowledge is table stakes and the panel knows you learnt it last week.

Structure that survives interruption

A demo that runs as a feature tour cannot survive interruption, because a tour has no argument and therefore no thread to pick back up. A demo built as an argument can. The structure that holds is: a one-paragraph statement of the problem in the customer's own words, confirmed back to them; the single moment in the workflow where their pain is worst; the product doing that one moment; the consequence in their metric; then, and only then, the surrounding capability.

Two rules make the difference under pressure. First, know which three minutes of your demo carry the argument, because when you lose half your time you will have to cut everything else and the panel will watch what you choose to keep. Second, narrate intent before action: "I am going to show you the approval step, because that is the part your team said takes two days." A narrated demo can be followed even when the screen does something unexpected, and an unnarrated one becomes incomprehensible the moment it deviates.

Interviewers also score something candidates neglect entirely, which is whether you opened by confirming the brief. Saying "before I start, let me check I have understood what matters to you" and then visibly correcting your own agenda based on the answer is a strong signal, because it is the behaviour of somebody who has previously demoed the wrong thing for forty minutes and does not intend to do it again.

The mid-demo question about something the product does not do

This will happen in your interview because it happens in the job, and the panel has agreed in advance who will ask it. The scored moment is the first four words out of your mouth.

The answer is that it does not do that, said plainly, before any mitigation. Everything that softens the no is read as a soft yes, and the full treatment of this scenario explains why: nothing said in a demo stays in the demo. It travels into the evaluation scorecard, into the requirements matrix that becomes a contract appendix, and into a delivery backlog owned by somebody who was not in the room and cannot renegotiate. "Yes, with some configuration" is the single most expensive sentence in presales.

What separates an adequate answer from a strong one is that the next move is a question rather than a workaround. You want to know what the capability is for, what happens today without it, and where the requirement came from, because those three answers distinguish a genuine blocker from a line copied out of an incumbent vendor's datasheet. Only then, if there is a credible path, do you describe it with its cost and its owner, and only if you can name somebody running it in production. A workaround invented on the spot is a soft yes in different clothing.

Because this moment is graded on wording rather than on strategy, it is worth having the wording. The left column below is what candidates say under pressure, and each of those phrasings has been written down by a customer as a commitment at some point.

Three sentences worth having ready in the exact words:

"It does not do that today. Can I ask what you would be doing with it, so I can tell you whether there is a sensible path or whether this is a genuine gap for you?"

"Not natively. There is a pattern customers use instead, and I would rather show you it properly at the end than rush it now. Is it all right if I finish this flow first?"

"I am not going to guess at a date for you. What I can do is get you in front of the product manager who owns that area, and if this decides the deal, get you an answer in writing from someone who can commit it."

That last one matters because the second reflex after the soft yes is the roadmap promise, and a date you did not set is a date you cannot honour. The related question on a feature you do not have treats the implied commitment as the defect, because it loses the account at renewal rather than at signature, which is a worse place to lose it. And get back to the demo afterwards: a gap question arrives in the middle of a narrative you built deliberately, and chasing it across four screens and never returning is scored as losing control of the room.

When the product fails live

If your interview involves a live environment, there is a reasonable chance it will break, and a non-trivial chance the panel arranged it. The scenario page on a live failure sets out the sequence, and the part carrying the score is naming the failure out loud before doing anything about it. Clicking silently and hoping is what everybody does, and it converts a thirty-second technical problem into a five-minute credibility problem, because the room watches you pretend. "That is not what should have happened. Give me one attempt and if it does not come back I will show you this a different way and we will not lose the meeting to it" costs you nothing and buys the room's patience.

The second graded element is what you say about the cause, and it requires knowing which of three things happened. A defect is the product behaving wrongly, and it needs a ticket and a date. An environment problem is your own setup, and it is embarrassing rather than serious. A gap you were hoping not to reach is a capability question wearing a technical failure's clothes, and describing it as an environment problem when the panel knows better is the answer that ends the interview. Being able to say "that was my environment, not the product, and I will prove it to you on Thursday" is only available to somebody who established which it was before speaking.

Discovery and qualification

Discovery separates senior candidates from competent ones, because a competent candidate collects requirements and a senior one qualifies while collecting them.

The order of the questions is the assessment

If an interviewer asks you to run a discovery call, they are listening for sequence. Business outcome first, then current process, then constraint, then decision process. Candidates who open with technology are marked down not because technology is irrelevant but because opening with it means the solution is already chosen and the rest of the call is confirmation bias. The walkthrough of a technical discovery call sets out the full sequence, including the two questions that do the most qualification work with the least friction: what changed to make this the quarter they act, and what would have to be proven for them to proceed.

Two behaviours in that round earn disproportionate credit. Declining to demo when asked to demo prematurely, done gracefully rather than stubbornly: "I could show you something now and there is a decent chance it would be the wrong thing. Give me ten minutes on how it works today and the demo will be about you." And closing with a written recap that names commitments on both sides, because that is the artefact that turns a pleasant call into a qualified opportunity.

Qualifying while you collect

The distinction interviewers are listening for is that discovery questions do two jobs. Each one gathers a requirement and simultaneously returns a qualification signal, and a candidate who can say what each question tells them is demonstrating the thing that separates the two levels. The value is rarely in the answer itself; it is in what the answer rules out.

The last row is the one candidates omit and senior interviewers wait for, because a technically won deal that has never touched procurement is the most common form of slipped forecast.

The political layer that discovery has to surface

Two scenarios recur because they are the ways discovery goes wrong that no requirements-gathering technique fixes.

The first is a stated requirement that contradicts everything else you hear. The full scenario separates the three cases that look identical from outside, and the one that catches technically strong candidates is the political commitment. If a director has already told their leadership the answer is a particular platform, being right about the architecture does not help you and being right in public makes you an enemy. The strong move tests the requirement against something already decided rather than arguing with it, and puts both readings in writing with a price on each, so the customer can change direction using their own facts rather than yours.

The second is an enthusiastic technical contact who cannot get you in front of anybody who can spend money, which is both the most common real situation in presales and a reliable interview scenario. What is graded is whether you can distinguish enthusiasm from a buying signal, and whether you have a stopping rule at all. The treatment of this case sets out three underlying situations that all resolve to the same move: make one specific ask that requires other people, and treat the refusal or the silence as the qualification answer rather than as a setback. Candidates who cannot name the point at which they would stop investing are describing a pipeline full of pleasant conversations, and every sales manager has already managed that person once.

Estimation, pricing and internal sign-off

For services-side solution architect roles this is the heaviest part of the loop, and where candidates most often give an answer that is technically sensible and commercially illiterate.

Estimating and pricing are two activities

Conflating them is the first thing graded. Estimating is a bottom-up statement of the effort a body of work takes, produced against comparable work you have delivered. Pricing is a commercial decision about what to charge, made with competitive position, margin targets and strategic value in view. A candidate who slides from one into the other in the same breath is telling the panel they will let the desired price bend the estimate, which is how a company ends up committed to something it cannot build. The full page on sizing under uncertainty sets out the discipline: decompose until the pieces resemble work you have delivered before, carry a three-point range with contingency as a visible line rather than as padding folded invisibly into the tasks, and let the commercial shape absorb the uncertainty you cannot remove.

That last idea is what senior interviewers listen for, and it is the point at which an engineer's instinct and a commercial instinct diverge. Faced with scope that cannot be pinned down, the engineering reflex is to widen the estimate, and a widened estimate is simply a larger number with no explanation attached, which a competitor will undercut without having solved the uncertainty either. The commercial response is to change the structure: price a discovery phase on its own, run a time-and-materials first increment with a fixed-price follow-on once the unknowns have collapsed, or cap an envelope and attach change control. Reaching for contract shape rather than for a bigger number is a senior signal, and it is one of the few answers in this discipline where the better response is also the one that protects the customer.

Effort is not a date

The second graded distinction is between effort and duration. Two hundred person-days is not four months, and the conversion is where dates are lost: productive capacity per person per week after meetings, support and leave; ramp-up before a new joiner contributes; and the sequence constraints that mean the critical path is frequently bound by the customer's dependencies rather than by your effort. A candidate who converts effort to duration by dividing by team size, with no productive-capacity factor and no look at sequence, has produced a date that is wrong in a predictable direction. Turning an estimate into a defensible date covers both halves, including what to do when the sales lead has already promised an earlier one, where the answer that scores is committing a smaller scope to the promised date rather than the same scope to a compressed plan.

Who owns which part of the number

Then the question that separates people who have been burnt from people who have not: who inside your own company signs off before the number reaches the customer. An estimate nobody in delivery has accepted is a number your company will be held to and nobody will defend, and the page on internal sign-off is blunt that a distribution list is not sign-off and a named person who will run the work is.

Being able to break the number into parts and say who owns each is the fastest way to show an interviewer you have done this in an organisation rather than in a spreadsheet.

Interviewers listen for whether you volunteer this unprompted, because doing so is a reliable proxy for having lived through a project sold on a number delivery never accepted. The related set piece is a budget that cannot buy the scope, where the strong answer shows the arithmetic once in front of the customer and converts the gap into a choice they own rather than a negotiation you conduct.

RFPs, compliance matrices and security questionnaires

Formal procurement is where presales becomes documentary, and the questions here are less about persuasion than about discipline under a scoring scheme you do not control.

Deciding to bid

The framing that scores on bid or no-bid is capital allocation. A serious response consumes days of your most expensive engineering time, and those days had an alternative use, so the qualification is about access, authorship and timing. Can you get to the stakeholders or are you restricted to written clarifications, does the requirement wording suggest an incumbent drafted it, is the timeline realistic, and do you have a relationship or are you column fodder making somebody else's procurement compliant. The full qualification pass covers the signals that suggest a wired deal and how to decline in a way that leaves the relationship intact, because declining badly costs you the next one as well as this one.

Triaging a bid that is already late

Once committed, the scenario interviewers reach for is the bid that will not be finished, and it tests allocation. The graded behaviour is triaging against the published scoring weights rather than against what feels most uncomfortable, and the full page adds the two points panels listen for. Freeze the solution and the price on day one, because a solution still moving on day four means every written section is describing a slightly different system and an evaluator will notice the contradictions before you do. And reserve the final day for a red-team read by somebody who wrote none of it, plus submission mechanics. Bids are lost on upload deadlines, unsigned forms and missing attachments with a frequency that embarrasses everybody involved.

The compliance matrix with nowhere to explain

This is the question that best distinguishes candidates who have run a bid from candidates who have read about one. The matrix wants a yes or a no against a requirement you partly meet, and there is no comments column.

The reframe that earns the marks is that the box is not a scoring field, it is a representation. A yes is an assertion your company can be held to if you win, which means the question is not "how do I score well here" but "what am I authorised to assert", and a bare no scores zero. The full answer works through the clarification-window move, the qualified marking with a cross-reference, and who in your organisation is allowed to say yes at all.

The options are worth laying out, because the interview answer that fails is the one that treats this as a scoring optimisation with no downstream consequence.

The follow-on candidates miss is that the matrix outlives the bid. It becomes an appendix, and the wording you chose under time pressure in week three is the wording that governs a dispute in month fourteen, read by a commercial manager who was not there and has no interest in how tight the deadline was.

The security questionnaire blocking a deal

The first graded insight here is that the person blocking you is not evaluating your product; they are discharging a duty against a control framework, and treating them as an obstacle rather than as a reviewer with a job is the behaviour that ends deals. The second is that three different blockers wear the same clothes, and the full treatment separates the real control gap, the wording problem where you hold the control but described it in language their framework does not recognise, and the policy that cannot bend and therefore needs reporting as a qualification answer.

The bright line the panel is checking for is that you never edit an answer to make it pass. A questionnaire response is a representation with the same standing as a contract schedule, and a sales lead asking you to soften one is the moment the role's integrity is tested. Have the sentence ready: "I cannot change that answer, because we would be representing something we do not do. What I can do is document the compensating controls and get our security lead on a call with theirs."

How much design to commit to in a proposal

The proposal question is where solution architect candidates most often reveal that they have never watched their own document become a contract. Two readers want opposite things from it. The evaluator rewards specificity, because specificity is how they tell a real solution from a brochure, and a vague proposal loses on scoring. Procurement and legal convert that specificity into obligations, and every named component, every number and every diagram box becomes something you can be held to in month fourteen when the architecture has changed for good reasons. Writing for one reader and forgetting the other is the failure.

The resolution is not to be vaguer. It is to split the document along the line of what you genuinely control. Commit to interfaces, outcomes, non-functional requirements and constraints, because those you can hold regardless of how the internals eventually look. Illustrate the inside, clearly labelled as indicative, with an explicit route for changing it. The full page on this makes the point that specificity is cheap in some places and ruinously expensive in others, the difference being whether the detail constrains how you build or what you deliver. A named message broker constrains how; a stated throughput and recovery objective constrains what, and only the second is a promise you can keep through an architecture change.

The other half is what makes a document survive procurement at all, and the end-to-end treatment covers it: two options and a recommendation, priced exclusions, acceptance criteria, change control, and a document written for a reader you will never meet. The part interviewers probe hardest is the assumptions register, because it is the load-bearing element and the one candidates treat as boilerplate. Every estimate rests on assumptions. An assumption written down and accepted by the customer converts into a change request when it turns out to be false, which is a commercial conversation with a defined outcome. An assumption held only in the estimator's head converts into an argument about what was reasonable to expect, and that argument is lost by whoever cannot produce a document.

A sentence worth being able to say out loud when a customer objects to the register: "These are not caveats to protect us. Each one is something we could not verify in discovery, and each one has a cost if it turns out otherwise. I would rather we both know which ones matter now than discover them in month three."

References, POCs and the moments that go wrong in public

Three late-stage scenarios recur in interviews because they are where a deal that looked won stops being won, and each tests whether you will protect the relationship or the quarter.

A reference call going wrong is the sharpest. Your existing customer, on a call with your prospect, starts describing how painful the implementation was, and nearly every candidate's instinct is to intervene. Intervening confirms the worst possible reading of what they were about to say. The scenario page sets out what to do instead, and panels grade the reframe hard, because a candidate who sees an unhappy reference as a sales problem rather than a customer problem is a candidate whose accounts churn.

A POC is the most expensive instrument in presales and the most frequently misused, and scoping one properly is a decision-instrument exercise rather than a trial: two or three claims genuinely in doubt, written criteria signed off by the person who will decide, a hard time box, and named customer resources committed before any work starts. The commonly missed element is the last one, and it is the one interviewers push on. A POC where the customer has committed no people is a POC you will run alone, and a result you produced alone is a result they have no reason to believe, because nobody on their side watched it happen or can defend it internally when somebody asks. Its companion, where the POC passed and the deal was lost anyway, is the more revealing of the pair, because it forces the candidate to account for the commercial track nobody was running.

The third is the competitive moment, where the answer begins by not disputing the claim, since you have no idea what it includes and disputing it makes the customer defend it on the competitor's behalf. Most unmatchable claims turn out to be matchable claims with a different scope boundary or an unstated assumption underneath, so the move is to decompose the claim against your own basis in front of the customer and let the difference surface. Where the competitor is genuinely better, concede cleanly: a concession is the cheapest credibility available and it purchases belief in everything else you claim. What you never do is match a number you cannot deliver, because that converts a lost feature comparison into a delivery failure with your signature on it.

What interviewers ask

Interviewers in this discipline ask fewer questions than an engineering loop and grade each on more dimensions, so it helps to know which dimension each stage feeds.

The hiring manager's opening walkthrough of a deal is scored on three things at once. Whether the narrative has a shape, meaning you can deliver it in four minutes rather than fourteen. Whether commercial context is present without being asked for, meaning you know the deal size band, the competitor, the decision process and why it closed or did not. And whether you own the parts that went badly. The observable tell is pronoun use under pressure: candidates who say "we won" and "they went dark" are describing events with no learning in them, and experienced interviewers probe exactly there.

The technical panellist is scored on floor and on boundary marking, which is why the questioning drills down until you stop knowing. Reaching that point is expected and is not the failure; continuing to answer past it is. The strong response states what you do know, marks the boundary explicitly, and says how you would resolve it and by when. A candidate who has never once said "I do not know" across a ninety-minute technical panel has almost certainly bluffed something, and the panel's shared suspicion that they cannot identify which answer it was tends to sink the whole submission.

The sales counterpart is scored on partnership. They are asking whether they would put you in front of their largest account: whether you would tell them the truth about a deal they want to believe in, whether you make them look competent in front of a customer or correct them in public, and whether you understand that the quarter is real. The tell they watch for is contempt. Engineers moving into presales sometimes leak disdain for sales as a function, in jokes about overpromising or a tone shift when describing account executives, and it is disqualifying because the role cannot work without that relationship.

The delivery-side panellist, where there is one, is scored on whether you would hand them something buildable. They probe assumptions, sign-off, and what you tell delivery on day one. Answering with "that is a delivery problem" ends the round.

The demo panel is scored on the four communication components and on the two set pieces, the gap question and the failure. They also grade one thing candidates never anticipate: whether you asked them any questions. A demo delivered without a single question to the audience is scored as a presentation, and presenting is not the job.

The executive round is scored on judgement, with questions chosen to have no safe answer. When would you refuse a deal. Tell me about a time you were overruled. What would you do if your account executive told a customer something untrue while you were in the room. What they want to see is whether you have a position and can hold it without being difficult, and whether you know the difference between escalating a risk and refusing to be a team player. The walk-away question is the archetype, and the strong answer includes the part most candidates omit: what you do when you lose the argument, which is to get the risks recorded and then stop objecting.

One cross-cutting signal worth naming, because it appears on scorecards under different words in every round: whether you distinguish what you know from what you believe from what you were told. Presales work is conducted almost entirely on incomplete information, and a person who reports all three in the same register is a person whose deal reviews cannot be trusted. Candidates who say "the customer told me the budget is approved, though I have not seen it and I would not forecast on it yet" are demonstrating the single most valuable habit in the discipline.

Questions

The questions below are phrased the way interviewers put them, and each answer names what the panel grades, because that is frequently not what the question appears to be about.

Walk me through a deal you worked from first contact to close.

Tell it in four minutes with a shape: what the customer was trying to do, what was in doubt, what you did about the doubt, what the commercial context was, and how it ended. Name the deal size band, the competitor, the decision process and the moment it could have gone the other way. If it was lost, say so early and say what you would do differently, in one sentence rather than five.

The grading is on whether commercial context arrives unprompted and whether the story contains a moment where you were wrong. Narratives in which the candidate was correct throughout and the customer was the variable are marked down for containing no evidence of learning. The second thing graded is compression: a candidate who cannot tell a deal story in four minutes will not brief an executive in two.

How would you open a first technical call with a prospect you know almost nothing about?

Set an agenda in the first minute, including how long you intend to take and what you want to leave with, then ask about the outcome before the technology. What are they trying to change, what does the process look like today, who else is affected, what has to be true for them to act this year. Confirm what you heard before offering anything, and close by naming a next step and a commitment on both sides.

The graded signal is question order, and opening on technology or current stack is marked down because it means the solution is already selected. The second is whether you protect the call's purpose against a premature demo request, declining gracefully rather than capitulating or lecturing them about process.

The prospect asks in the middle of your demo whether the product can do something it cannot. Go.

"It does not do that today." Then ask what they would be doing with it, what happens now without it, and where the requirement originated. If there is a genuine path, describe it with its cost and name somebody running it in production. If there is not, say the gap is real and ask whether it is a blocker, because the answer changes whether you are still in the deal. Then return to the demo thread you left.

Everything is graded on the first sentence, and softening it into "with some configuration" fails the question outright, because the panel knows that sentence propagates into a requirements matrix and eventually into a contract. Secondary credit for asking what the capability is for rather than solving the feature as stated, and for resuming the demo rather than abandoning it.

The customer says your competitor has committed to delivering this in eight weeks. You cannot.

Do not dispute it. Ask what the eight weeks includes: which environments, whose data, how much of the integration, who is doing the migration, what is assumed about their side. Almost every unmatchable commitment is a matchable commitment with a different scope boundary or an assumption nobody has stated. Decompose it against your own basis in front of them, and where you find the difference, make sure they know which question to put to the other vendor. If the competitor is genuinely faster, say so and say why, and then compete on what happens in week nine.

The grading is on composure and on whether you match the number. A candidate who reflexively promises eight weeks has told the panel they will do it in the job, and one who attacks the competitor's honesty has told them they will make the customer defend a rival. Conceding a real advantage cleanly scores positively, which surprises candidates who assume the exercise is to win the exchange.

Your demo environment breaks in front of twelve people. What happens in the next thirty seconds?

Say out loud that it is broken and that you are not going to pretend otherwise. Take one attempt, narrating what you are trying. If it does not come back, move to the alternative you prepared, which might be a recorded walkthrough, a second environment or the whiteboard, and protect the purpose of the meeting rather than the demo script. Afterwards, send a same-day note saying what broke, whether it was a defect or an environment problem, and when you will show them properly.

The graded behaviour is the first sentence, because silence while clicking is what everybody does and naming it converts a technical problem into a demonstration of composure. The second element is whether you had an alternative at all, since a candidate with no fallback has told the panel exactly how they prepare for customer meetings. The third is honesty about the cause: describing a capability gap as an environment problem ends the interview if the panel knows better, and in a real deal it ends the relationship rather later and more expensively.

How do you decide whether to respond to an RFP?

Treat it as spending scarce senior engineering days rather than as an opportunity to be seized. Establish whether you can get access to stakeholders or are restricted to written clarifications; who authored the requirements, since wording lifted from a vendor's documentation tells you the shape of the incumbent's involvement; whether the timeline is compatible with a serious answer; whether you have any relationship at all; and whether there is a win theme you can name in one sentence. If the answer is no on access and no on relationship, decline in a way that keeps the relationship, or bid conditionally on a subset you can win.

The signal is whether no-bid is available to you as a real option and whether you cost the response. Candidates who treat every RFP as worth answering are describing a team whose best people spend their quarter writing documents that lose. The second signal is win-theme articulation: if you cannot say in one sentence why you win this specific bid, you do not have a reason to spend the days.

The compliance matrix wants yes or no and you partly comply. There is no comments column. What do you write?

First, try to change the question. If the clarification window is open, ask whether qualified compliance is acceptable and where the explanation should be placed, because that question is frequently answered generously. If the window has closed, use a qualified marking with an explicit cross-reference to the section carrying the explanation, and put that explanation where evaluators read rather than in an annexe. A bare yes is a representation you cannot support after award; a bare no scores nothing. And check who in your organisation is authorised to make the assertion, because in a formal bid it is often not the person filling in the cell.

The grading is on whether you recognise that the cell is a contractual representation rather than a scoring field. Candidates who optimise purely for score, reasoning the detail can be sorted out later, have described how a company ends up in a dispute about an appendix. Additional credit for knowing that the matrix survives the bid and becomes a schedule, and for asking who signs it off rather than assuming the answer is you.

The bid is due in five days, three technical sections are blank, and the clarification window has closed. How do you spend the five days?

Get the scoring weights in front of you and allocate against them, not against which sections feel worst. Freeze the solution and the price on day one, because sections written against a moving solution contradict each other. Spend the middle days on unanswered high-weight requirements and on the compliance matrix, which is usually pass or fail rather than scored. Answer a section you lack information for by stating the assumption explicitly rather than writing something vague. Keep the last day for a red-team read by someone who has written none of it, and for submission mechanics.

The graded behaviour is triage against published criteria and the discipline of not writing on the final day. Bids are lost on portal deadlines, unsigned forms and missing attachments often enough that reserving that day is a recognised marker of experience, and a candidate who plans to write until the deadline has told the panel which failure they have not yet had.

Their security team has blocked the deal over one of your questionnaire answers. The business wants to buy. What now?

Find out what specifically failed, because there are three different problems here. A genuine control gap is an engineering conversation and possibly a compensating-controls conversation with evidence attached. A wording problem, where you hold the control but described it in language their framework does not recognise, is a mapping exercise done with the reviewer. A policy that cannot bend is a qualification answer and needs reporting upwards as one. Then get your security lead talking to theirs, because a reviewer will accept from a peer what they will not accept from a vendor's field team.

The bright line is that you do not edit an answer to make it pass, and the panel will sometimes apply pressure to see whether you will, usually by playing a sales lead who suggests the wording was overly conservative. Practise the refusal out loud, because it has to sound like a fact rather than an objection. The second graded element is whether you work with the reviewer or around them, since routing round security via an executive sponsor wins the battle and creates an enemy who reviews your renewal.

Who signs off an estimate inside your own company before it goes to the customer?

A named person who will be accountable for delivering it, not a distribution list and not silence taken as consent. In practice that means the delivery lead or practice lead who would staff the work accepts the effort and the assumptions, commercial accepts the price and the margin, and where there is unusual risk, whoever is authorised to accept that risk sees it explicitly. What delivery receives on day one should be the same document that was signed off, including the assumptions register and the exclusions.

The grading is on whether you have internalised that an estimate becomes a commitment the moment it leaves the building. Candidates who describe estimation as their own analytical output, with no internal acceptance step, are describing the mechanism by which delivery teams come to distrust the field.

The delivery lead refuses to own the number you need to submit. It is due tomorrow.

Establish what specifically they disagree with, because "that is too low" is not actionable and "we cannot get a Kafka engineer for eight weeks" is. If it is a difference about scope or assumptions, fix the assumptions and re-estimate honestly. If it is a genuine difference of professional judgement, do not average the two numbers, because an averaged number is wrong on both bases and owned by nobody. Write the disagreement down, both positions and what each rests on, and put it in front of the person who can accept the risk. Then submit what they decide.

The graded elements are the refusal to average and the willingness to escalate in writing, which is a document rather than a confrontation. Candidates who quietly submit their own number because the deadline was tomorrow have described the behaviour that makes the role dangerous.

How do you estimate work when the scope is genuinely unknown?

Decompose until the pieces resemble work you have delivered before, and estimate those against what they took. Carry a three-point range, keep it visible in the document rather than collapsing it to a single figure, and put contingency on its own line so it can be discussed rather than silently spent. Then choose a commercial shape that matches the residual uncertainty: a separately priced discovery phase, a time-and-materials first increment with a fixed-price follow-on, or a capped envelope with change control. If the uncertainty is too large for any shape, you are being asked to price discovery and should sell discovery.

The signal is reaching for contract structure rather than for a larger number. Padding is what a junior estimator does with uncertainty, and padding is invisible, unexplainable when challenged, and the first thing a competitor undercuts. Naming a range and then choosing a shape that carries it is the senior move.

You are told to take fifteen per cent out of the number. What do you take out?

Scope, phasing or team mix, never the same work re-estimated downwards. Concretely: defer a workstream to a later phase, reduce the environment count, move a portion offshore or to a mixed team with the seniority consequences stated, remove a discretionary item you can name, or narrow the acceptance criteria with the customer's agreement. Every one of those is a decision somebody can take. Re-estimating identical work at a lower figure is not a decision, it is a fiction that delivery discovers in month two.

What is graded is whether the reduction is traceable to something that changed. "I would sharpen the assumptions and get to twelve per cent", with no corresponding change in what is delivered, is a candidate who will absorb commercial pressure into engineering estimates for the rest of their career.

The budget is half what the scope costs. What do you put in front of the customer?

Show the arithmetic once, plainly, so the gap is a shared fact rather than your opinion. Then convert it into a choice: what your budget buys as a coherent first phase that delivers one outcome completely, what the full scope costs, and which specific outcome is being deferred, described in their terms. If the gap cannot be bridged even in phases, say so and say what would have to change. Do not agree to the full scope at the reduced figure.

The graded reframe is turning a negotiation into a decision the customer owns. The second element is what you cut, because it reveals experience faster than anything else in the answer. Cutting scope is honest and visible: the customer knows what they are not getting and can decide whether they mind. Cutting quality, testing, environments or non-functional work to preserve the feature list is invisible at signature and arrives in production, which is how the account becomes a reference you cannot use and a delivery team you cannot staff twice.

How much of the design do you commit to in a proposal?

Commit to what you can hold regardless of how you build it: interfaces, integration contracts, outcomes, non-functional requirements, constraints and acceptance criteria. Illustrate the internals to win the evaluation, but label them as indicative and write in the mechanism for changing them before anybody needs it. Be specific where specificity demonstrates command without constraining implementation, and be careful with named products, named versions and component counts, because each is a line procurement will lift into a schedule.

The signal is whether you know the document has two readers with opposite interests, and whether you have watched one of your own diagrams become an obligation. Candidates who answer purely from the evaluator's side produce beautiful proposals delivery cannot honour; candidates who answer purely from the legal side produce proposals that lose. The strong answer holds both and says which parts of the document are which.

Your only contact loves the product and cannot get you to anyone with budget. What do you do, and when do you stop?

Work out which of three situations you are in: there is no project, there is a project owned by somebody else, or there is a project and he is not somebody his organisation listens to. Then make one ask that requires other people, such as a working session with the platform owner and someone who holds the budget line, or a one-page summary he takes to his director with a date for feedback. Multi-thread openly rather than behind him. Set a stopping point before you start and honour it.

The grading is on whether enthusiasm is being mistaken for a buying signal, and on whether a stopping rule exists at all. A candidate who would keep nurturing indefinitely has described a pipeline of pleasant conversations, and every sales manager has managed that person once.

Everything the customer says they want contradicts what you think they need. What do you do?

Treat the contradiction as information. Work out whether the stated requirement is stale, a proxy for something nobody will say aloud, or a commitment somebody has already made publicly and cannot retract. Test it against something already decided rather than arguing with it: ask how it fits a constraint they have stated, and let the conflict surface from their own facts. Then put both readings in the proposal with a price on each, so the choice is theirs and nobody has been contradicted in front of their leadership.

The graded skill is raising a disagreement without making anyone wrong, because being right about the architecture and wrong about the politics loses deals. The second signal is whether you would comply anyway when the requirement is a political commitment, with the consequences documented.

Scope me a POC for a customer who says they need to see it working before they can buy.

First establish what is genuinely in doubt, because a POC that proves things nobody doubted is expensive theatre. Pick two or three claims, write pass and fail criteria specific enough to be unambiguous, and get them signed off by the person who will make the decision rather than by the engineer who will run it. Put a hard time box on it. Name the customer resources you need and get them committed before any work starts, including data, access and a person who will be in the sessions. Agree in advance what happens if a criterion fails.

The signal is whether the criteria are agreed by the decision-maker, since criteria agreed with a friendly engineer are not agreed criteria. The second signal is whether you would refuse an unscoped POC, and the third is whether you treat a failure as information to report rather than as something to manage around.

You passed every POC criterion and they bought from the competitor. What went wrong?

Most likely one of three things. The commercial track stood still while the technical one ran, so at the end there was a validated solution and no business case, no budget owner and no procurement path. Or the criteria measured what you are strong at rather than what the buyer needed to decide, which usually means they were drafted by you and accepted by somebody who was not deciding. Or the decision was never technical, and the POC was a compliance step in a process already leaning elsewhere. The deliverable now is a debrief that establishes which, because the same design will lose the next one.

The graded element is whether you accept that a POC removes an objection and does not create a buyer. Explaining the loss as customer irrationality fails; identifying that you ran one of two necessary tracks and did not check the other demonstrates the commercial awareness the level requires.

You are on a reference call and your customer starts saying the implementation was painful. What do you do?

Let them finish. Interrupting confirms the worst reading of what they were about to say, and the prospect will conclude that you knew. Then use the call to get specifics, which are more useful to you than the general impression: what went wrong, when, whether it was resolved, and what they would want done differently. Afterwards, follow up with the prospect on the facts, including what has changed since. And treat the existing customer's unhappiness as a problem to fix rather than an incident to contain, because they were willing to take the call, so the relationship is recoverable.

The graded reframe is whose problem this is. A candidate who describes damage control to the prospect and nothing about the existing customer has told the panel how their accounts will churn, and one who asks who at their own company knew this reference was unhappy has identified the real defect.

When would you tell your own company to walk away from a deal?

When the conditions for delivering it successfully do not exist and cannot be created inside the deal: a scope and budget that cannot meet even in phases, a mandatory requirement the product genuinely does not satisfy, a timeline bound by the customer's own dependencies, or a commercial structure that transfers unpriced risk. The way to make the case is a one-page recommendation, not a veto: what you believe, what it rests on, what the exposure is, and the specific conditions that would change your view. Take it to whoever is authorised to accept the risk, because that is the decision being made. If they proceed anyway, get the risks recorded, then stop objecting and help win it.

Two things are graded. Whether you separate what is technically true from what you would prefer, since many walk-away arguments are an architect's distaste for an unpleasant project wearing a risk assessment's clothes. And what you do when overruled, because the answer that matters is that you become useful again. A candidate who would keep relitigating a decision that has been taken is describing themselves as a problem their manager will have to manage, and no amount of being right about the deal offsets that.

Your account executive tells the customer something about the product that is not true, while you are in the room.

Do not correct them in front of the customer unless the untrue thing is material to a decision being taken in that meeting. If it is material, correct it as a clarification rather than a contradiction: "Let me add some precision to that, because the way it works is slightly different and it matters for your migration." If it is not material, let the meeting finish, raise it with them directly and privately, and then make sure the accurate version reaches the customer in writing, because the record is what survives.

The graded judgement is the materiality test, and both extremes fail. Never contradicting your account executive tells the panel a customer can be misled on your watch; correcting every inaccuracy immediately in public tells them you will make your partner unsafe in front of customers, which ends the working relationship the role depends on.

The customer's architect is openly hostile and clearly wants the incumbent to win. How do you handle the room?

Take the objections seriously and answer on the technical merits, because the rest of the room is watching whether your product survives scrutiny from the person who understands it best. Do not attempt to win them over socially; that reads as evasion. Where they are right, agree specifically and quickly, since a conceded point costs almost nothing and denies them the argument. Where they are wrong, hold the position calmly and with evidence rather than volume. And privately, work out with your account executive what this person's incentive is, because someone defending an architecture they chose is a different problem from someone protecting a team's jobs.

The signal is whether you can be disagreed with without becoming either defensive or deferential. The panel is simulating the moment their product is attacked by the one person in the room qualified to attack it, and they need to know you will not fold, because folding in front of a customer's technical authority loses everybody else in the room at the same time. Secondary credit for reading the incentive rather than treating hostility as a personality trait.

The customer asks for a discount you know your company will not approve. They ask you directly, in the room.

Do not price in the room and do not signal that there is room. Say plainly that you are not the person who sets it and that you will not guess, then move the conversation to the thing that determines the answer: what shape the deal would need to take for a different number to be justifiable, meaning term, volume, scope, timing in the quarter, or reference commitments. Take it back to whoever owns pricing with those details attached, because a discount request with a rationale gets a decision and one without gets a no.

The graded discipline is not committing what is not yours, which is the same reflex as not committing a roadmap date. Interviewers watch for the candidate who wants to be liked in the room and creates an expectation their commercial team must then withdraw, which is worse than the original refusal.

What does your first ninety days look like here?

Learn the product to the depth where you can be attacked and not embarrassed, which means using it against real scenarios rather than reading documentation. Sit on calls before you run any, specifically on losses. Learn the top three competitors from the position of somebody who has to concede a real advantage credibly. Get to know the delivery or support side early, because they hold the truth about what the product does in production. Then take a small deal end to end rather than shadowing a large one indefinitely.

What is graded is whether the plan is about learning to be useful or about being visible. The most valuable item is the one candidates rarely name: the relationship with delivery, which separates presales people whose commitments are believed internally from ones whose numbers are quietly re-estimated after they leave the room.

Tell me about a technical claim you made that turned out to be wrong.

Choose a real one where the consequence landed on somebody, describe how it was discovered, and describe what you did in the hour after you knew. What matters is the sequence: you told the customer before they found out, you told your own delivery team, and you changed something about your process so the class of error does not recur. If the change was a habit rather than a policy, say which habit.

This is graded almost entirely on whether the example is real, and interviewers ask a second and third layer of detail to test whether the story holds. The related signal is whether you disclosed proactively, because self-disclosure of an error is the behaviour that makes a presales person's other claims believable.

What is the hardest thing about this job that people outside it do not see?

The good answer is specific and structural rather than emotional. Holding two organisations' incompatible measures in your head at once and being honest with both. Being the person whose commitments are enforceable but who controls neither the product roadmap nor the delivery capacity. Carrying the memory of what was promised across a deal cycle long enough that everybody else has forgotten. Or the fact that the work leaves no audit trail, so a reputation gets built on rooms nobody else was in and is defended with nothing but consistency.

The panel is checking whether you understand the shape of the role rather than the tasks in it. An answer about long hours or difficult customers is answering a different question; an answer about the fault line between sales and delivery tells them you have stood on it.

We give candidates a scenario and a week to prepare a presentation. What would you want to know before you start?

Who is in the room and what each of them cares about. What the customer's current situation is, particularly what they use today and why they are looking. What decision the meeting is meant to move, since a first-call deck and a shortlist presentation are different artefacts. How long you have, and whether the panel wants interruptions or a clean run. And whether there is anything they have already been told, so you do not contradict your own account team.

That set of questions is itself the answer, and asking it is scored more highly than any deck you could build. A candidate who takes the brief as given and prepares in silence has demonstrated exactly the behaviour that produces a technically excellent presentation about the wrong problem. This is the cheapest mark available in the entire loop and most candidates leave it on the table.