An interviewer plays your CFO. The company cut 15% of engineering last quarter and you want eighteen months of platform investment. They point at the November 2022 Eras Tour onsale — Ticketmaster reported about 3.5 billion system requests at four times its previous peak and cancelled the next day's public sale — and say "so what exactly am I buying?" Walk me through the case you make.

What the interviewer is testing

Three things. Whether you can express an architectural ask in a number the CFO already forecasts. Whether you understand what a 15% cut removes from your argument — headcount is no longer available as a currency, so "we need six engineers" is not an ask, it is a decline. And whether you can read the Ticketmaster example correctly, which most candidates cannot: they hear a capacity story when the documented outcome was a commercial one.

The clarifying questions that change the answer

• Which revenue event in the next eighteen months carries the largest single-day exposure, and how much?
• Is the cut a permanent structural reduction or a hiring freeze with a reopening date?
• What is infrastructure cost per order today, and is it rising faster than orders?
• Which number does the CFO already report to the board that this could move?

Without the first and the fourth you are guessing, and an investment case built on a guess is spotted in one question.

A strong answer's arc

Lead with the event, not the system. "Our largest onsale next year is in March and carries about X of the quarter's revenue." That sentence puts the ask in the CFO's own model.

State the failure mode in commercial language. The documented Ticketmaster outcome was not a latency figure: after 3.5 billion system requests at four times the previous peak, with about 3.5 million people registered through Verified Fan and only a limited number of them issued access codes while the rest were waitlisted, the next day's public sale was cancelled. More than two million tickets had sold in a single day. A cancelled sale is a revenue, brand and political event — it drew a letter from a United States senator — and none of that appears on an availability dashboard.

Then reframe the ask. The naive version is capacity for 4× peak, which is bought once and idle for 364 days, and a CFO who has just cut 15% will say no correctly. The useful version is a demand-admission mechanism: the ability to decide how much demand to let in, in what order, with what the excluded customer is told. Ticketmaster already had the shape of this in Verified Fan invite codes. The architecture ask is for the admission control, the queue the customer can see, and the honest message to the people who do not get in.

Close with reversibility and a kill number. "Twelve weeks, one squad, no new headcount. At month three I will show you admitted-buyers-per-dollar under load test; if it is not above the threshold we stop and you keep the remaining budget."

Common weak answers

• Asking for headcount. The cut was the answer to that question.
• Leading with the technology. A queueing system named in sentence one and the CFO has stopped listening.
• Quoting an availability target. 99.95% does not describe a cancelled onsale, and the executive knows it does not.
• Promising the 4×. You cannot, and promising it makes the next conversation about your credibility.

What a strong answer adds

The second-order cost of the cut itself: on-call depth fell by 15% too, so the expected duration of the next incident rose, which raises the value of the admission mechanism rather than lowering it. And one sentence that changes the decision: the cheapest version of this is not more capacity, it is the right to turn demand away in a way the brand survives.

When this is the wrong answer

If the business has no large concentrated demand event — steady traffic, no onsales, no launches — the admission framing is borrowed from someone else's problem and will read as such. Then the honest case is unit cost: cost per order, with a curve and a date, and no reference to anybody's outage.