Review this deck for a 25-minute architecture review: slides 1 to 6 cover the landing zone and network layout, 7 to 18 walk through each microservice, 19 to 26 show the data model, 27 to 31 give the test strategy, and 32 to 34 ask for four engineers for two quarters. The approval has been deferred twice. What would you cut, what would you change, and what would you leave alone?

What is actually required

The meeting has one output: yes, no, or yes-with-conditions on four engineers for two quarters. Everything that does not move the audience toward that output is costing time that belongs to the decision.

The deck is ordered bottom-up, so the ask arrives at minute 23 of 25, into an audience whose attention is spent. Worse, slides 1 to 31 contain nothing the audience can disagree with. Faced with 31 slides of unfalsifiable structure and a number at the end, the only safe response available to a cautious approver is "come back with more detail" — which is what happened twice.

What I would cut and why it is safe

• Slides 7 to 18, the service-by-service walk. Twelve slides of boxes answer a question nobody in the room is accountable for. Replace with one slide of the container view.
• Slides 1 to 6, the landing zone, unless the four engineers are being asked for because of it. Network layout belongs in the appendix.
• Most of 19 to 26. Keep the one schema decision that is hard to reverse — usually the tenancy key or the identifier design — and drop the entity-by-entity tour.

A reasonable budget for a 25-minute decision slot is about one substantive slide per two minutes, so roughly 12 slides, because each substantive slide draws a question that runs 60 to 120 seconds. Thirty-four slides in 25 minutes is a plan to be interrupted.

The one change that matters

Put the ask on slide 1 with three numbers: what it costs, what it delivers, and what breaks if it is deferred a quarter. The third number is the one that gets decisions, because deferral is otherwise free. A line like "each quarter of delay adds roughly £180k of manual reconciliation and pushes the vendor exit past its contract date" makes doing nothing a visible choice rather than a default.

What I would leave even though it looks odd

The test strategy. It looks like engineering detail in a funding conversation, and it is the slide a cautious approver uses to believe the estimate. Cut it to one slide and keep it. Keep the appendix too, and say out loud that the 22 cut slides are in it: a visible appendix converts "they have not thought about it" into "they have thought about it and are respecting my time."

When this is the wrong answer

If the audience is the engineering group that will build it, invert the whole critique: they need the mechanism, the rejected alternatives and the parts you are unsure about, and the funding ask is the slide to cut. And if this is a genuine design review rather than a funding gate, the service walk is the content — but then it needs a pre-read and 90 minutes, not a 25-minute slot.

Common weak answers

• "Add an executive summary." A summary bolted onto a bottom-up deck gets read and then contradicted by the next 31 slides.
• "Cut words and add diagrams." The defect is the order and the missing ask, not the density.
• "Book a longer meeting." Two deferrals are not a time problem. A deck that cannot get a decision in 25 minutes will not get one in 50.