You are convinced the company should walk away from a deal that everybody else wants to win. How do you make that case, and what do you do if you lose the argument?

First, check which kind of objection you have

Before making the case, be honest about which of two things you are holding. One is a feasibility or commercial judgement: the scope cannot be delivered inside the price at any staffing we can field, a mandated requirement is one we cannot meet, the customer's dependency cannot be secured, or the contract shape puts unbounded risk on us. The other is a preference: the technology is one you dislike, the work is unglamorous, the customer's team is difficult, the margin is thinner than you would like.

Only the first justifies escalation, and conflating them is how architects lose the standing to be listened to on the first kind. An interviewer is listening closely here, because a candidate who describes every commercially awkward deal as undeliverable is describing themselves as an obstacle.

Make it a recommendation with conditions, not a veto

A no delivered as a position invites the organisation to route around you. A no delivered as a decision paper with named conditions puts the decision where it belongs and, usefully, is much harder to dismiss. Keep it to one page and lead with the recommendation.

RECOMMENDATION - Northgate WMS replatform
For: deal board, 30 July.  From: solution architecture.

Recommendation: do not bid in the current form. Bid conditionally
if the three conditions below are met by 8 August.

Basis
1  Scope against price. Bottom-up sizing is 610 days, £560k cost
   including risk-weighted contingency. Their stated ceiling is
   £400k, which at target margin funds 366 days. We would be
   committing to 610 days of scope on 60% of the funding.
2  Mandated requirement. Section 4.1 requires a certification we
   do not hold and cannot obtain inside the programme window. We
   would be non-compliant on a pass or fail criterion.
3  Dependency we cannot control. The plan needs their mainframe
   test environment from October. They have declined to commit a
   date twice, and the migration workstream cannot start without it.

Conditions that would change this recommendation
C1  Scope reduced to the phase-one boundary in annexe A, at £400k,
    with the workbench contracted separately.
C2  Written confirmation that 4.1 is scored as desirable, not
    mandatory, or a named partner accepted for that element.
C3  A dated commitment on the test environment, with day-for-day
    schedule relief if it slips.

If we bid without C1 and C3 my estimate of the outcome is a
project underfunded by roughly £160k that is late for reasons we
cannot recover, and a reference customer we will not be able to
name.

Cost of not bidding: the 22 presales days already spent, and the
relationship, which I propose to protect with a paid two-week
assessment offer.

The conditions are what make this usable rather than obstructive. They convert your judgement into something the sales lead can act on, they give the deal board a middle option between bidding blind and walking, and they make you the person who found a way rather than the person who said no. They also protect you if the conditions are met, because you have publicly committed to supporting the bid in that case.

Escalate to whoever can accept the risk

You are not entitled to stop a deal, and claiming that authority is usually a mistake. What you are entitled to is that the risk is accepted knowingly, at the level that carries it. A margin decision belongs to whoever owns the profit and loss; a delivery risk of this size belongs to the delivery director who will resource it; an unbounded contractual liability belongs to whoever signs.

So the escalation is not upward against the sales lead, it is sideways to the people whose problem this becomes. Show the sales lead the paper first, in advance, and take their corrections — half the time a fact you have is wrong, and the other half you learn something about the deal that changes the picture. Then table it where the decision is made. An organisation is allowed to take a bad deal deliberately, for a strategic reason, and sometimes should. What it must not do is take one accidentally because nobody wrote the arithmetic down.

Losing the argument well

If the decision goes against you, the professional behaviour is a single clean transition: object once, in writing, then stop objecting and start reducing the damage. Continuing to relitigate makes you the reason the deal is difficult, and it is the fastest way to be excluded from the next one — including the ones where you would have been right.

Reducing the damage is concrete work. Get the assumptions and the dependency relief into the proposal, so that the risk you named at least has a contractual mechanism attached to it. Get the customer resource commitment named with an owner and a date. Make sure the delivery lead has read the sizing and the conditions before signature rather than after, because before signature the information is a veto and afterwards it is only news. Put the risks on the register with your name against them and a date, which is not defensive book-keeping but the thing that makes the next escalation credible.

And then deliver it wholeheartedly, because a team that has been handed a deal someone considers unwinnable will deliver it as though it were unwinnable, and an architect visibly waiting to be proved right accelerates that.

The credibility economy this sits inside

The reason to be sparing with this is that it draws on a finite balance. An architect whose caveats have historically mattered is listened to and can stop a deal with a sentence; one who caveats everything is politely thanked and worked around within two quarters. That balance is built by being specific, by naming conditions that could be met rather than objections that cannot, and by being visibly pleased when a deal you doubted is fixed and won.

The other half of the economy is what happens when the deal fails as you predicted. The temptation is to point at the paper. Doing so buys one moment of vindication and costs the relationship with the sales team you need for the next five pursuits. The better use of the paper is quiet and structural: take it to the next qualification review as evidence for a standing gate — for instance, that no fixed-price bid proceeds where the sizing exceeds the customer's stated budget by more than thirty per cent without a named executive accepting it — so the argument does not have to be won personally the next time.