Before a deck reaches a partner or client, run a structured QA pass across five areas: the argument holds up, every number is defensible, each slide stands on its own, the details are correct and consistent, and the file is clean and safe to send. Most decks that fail in the room do not fail on design; they fail because a question exposed a gap the author never checked for. The checklist below is built to catch those gaps in the last hour before you hit send, not to polish fonts you have already polished.
Key takeaways
QA is about the thinking, not just the typos. The checks that matter most protect the argument and the numbers, not the alignment.
Run it in five passes. Argument, evidence, standalone slides, details, and safe-to-send hygiene.
The number and source of every material figure should be defensible. "Based on what?" is the most common follow-up in any serious review.
Every slide should survive being read out of order. A committee member will jump to slide 14, and it should still make sense.
AI-generated decks need this pass too. No tool removes the need for a human review before a high-stakes deck ships.
This checklist is written for consultants, strategy and finance teams, and anyone sending a deck into a partner review, an investment committee, or a client meeting. Work through the five sections below, then use the one-page checklist near the end as your final pre-send pass.
Does the argument hold up?
Check that the deck leads with a single, clearly stated recommendation and that the supporting logic is clean before you look at anything else. The recommendation and the ask should be stated in one or two sentences up front, not assembled by the reader from context scattered across the deck. Reading only the slide titles in order should tell a coherent story, which means each title is a full-sentence takeaway rather than a topic label. And the supporting points beneath the recommendation should be MECE, with no overlaps and no gaps, because that is the first thing a follow-up question tends to break.
Can you defend every number?
Check that every material figure traces back to a named source, that repeated numbers agree across slides, and that the assumptions which would change the answer are stated openly. "Based on what?" is the most common question in a serious review, and a single figure no one can defend quietly undermines confidence in every other number on the page. Reconcile totals and any number that appears more than once, since a mismatch between slide 3 and slide 11 is exactly what a sharp reviewer will notice. Finally, name the two or three assumptions that would change the recommendation if they were wrong, because experienced reviewers go looking for those first, and it is far better to raise them than to be caught hiding them.
Does each slide stand on its own?
Check that every slide makes one point, that its title and body agree, and that the deck still makes sense when someone jumps to a slide out of order. A slide should carry a single message; if the title claims one thing and the chart shows another, the reader trusts neither. The chart itself should be the right type for the point and readable in one glance, not a figure the audience has to decode. And because committee members rarely read front to back, each slide should work on its own, so a deck that only makes sense slide one through the end was never really finished.
Is it correct and consistent in the details?
Check spelling, names, labels, units, dates, and number formatting, since small inconsistencies are what a client notices first. Proofread for typos and confirm that the client's name and every person's name is spelled correctly, because nothing erodes credibility faster than getting the audience's own name wrong. Make sure axes, units, and dates are present and consistent, and that currency, rounding, and number formatting do not drift between slides. These are small things individually, but a deck full of them signals that the analysis behind it may be just as loose.
Is it clean and safe to send?
Check that the deck is in the right template, that footers and page numbers are consistent, that charts are editable objects, and that nothing private is left in the file. Confirm the deck sits in the correct firm or client template, with consistent fonts, colors, footers, and page numbers throughout. Charts and tables should be native, editable objects rather than flat images, so a last-minute number change can be fixed in place instead of forcing a rebuild. Most importantly, strip anything that should not leave the building: speaker notes, internal comments, tracked changes, and any leftover names from a previous client's version of the deck.
The one-page slide QA checklist
Use this as your final pass; if any item is a no, the deck is not ready to send. Each row is a single check you can tick off in the last hour.
# | Check | Area |
|---|---|---|
1 | The recommendation and the ask are stated up front in one or two sentences | Argument |
2 | Every slide title is a full-sentence takeaway, not a topic label | Argument |
3 | Supporting points are MECE: no overlaps, no gaps | Argument |
4 | Every material number traces back to a named source | Evidence |
5 | Repeated numbers and totals reconcile across slides | Evidence |
6 | The assumptions that would change the answer are named, not buried | Evidence |
7 | Each slide makes one point, and the title and body agree | Slides |
8 | The chart type fits the point and reads in one glance | Slides |
9 | The deck still makes sense read out of order | Slides |
10 | No typos; client and people names are spelled correctly | Details |
11 | Labels, axes, units, and dates are present and consistent | Details |
12 | Currency, rounding, and number formatting are consistent throughout | Details |
13 | Correct template, with consistent footers and page numbers | Finish |
14 | Charts and tables are native, editable objects, not flat images | Finish |
15 | Speaker notes, comments, and any old client names are removed | Finish |
Do online AI presentation tools follow this QA checklist?
Many popular AI presentation tools optimize for speed and visual design rather than the argument, evidence, and finish work this checklist covers, so they help you produce the deck but leave most of these checks on your side of the desk. Perceptis AI is built to cover much of the checklist by construction, so a lot of the argument, evidence, and finish work is already done before you open the file. It structures the deck top-down with MECE logic and writes full-sentence action titles, which takes care of the argument checks. It ties every factual claim to a traceable source, so any number can be defended slide by slide and the evidence checks are largely covered. And it generates a native, editable PowerPoint deck straight into your firm's own template rather than a flat export, so the finish checks are in good shape from the start. What no tool does, Perceptis included, is remove the need for the final human pass; because Perceptis also works as a PowerPoint add-in, you can run that last read on the argument, the numbers, and the file hygiene without leaving PowerPoint. In practice it turns the checklist from a full end-of-night QA marathon into a quick final read. You can start from a prompt or your own documents, with a free tier of 10 slides per month and paid plans from $29 per month (Perceptis, 2026). More on the main Perceptis site.
Frequently asked questions
What should I check before sending a deck to a client? Run five passes: the argument (a clear recommendation, action titles, MECE support), the evidence (sourced numbers that reconcile), each slide (one point, right chart, works out of order), the details (spelling, names, units, formatting), and safe-to-send hygiene (correct template, editable charts, no leftover notes or old client names).
What is a slide QA checklist? It is a structured list of checks you run on a deck before sending it, covering the logic, the numbers, the individual slides, the details, and the file hygiene. Its purpose is to catch the gaps that surface under questioning, not just visual polish.
How do I proofread a presentation before a partner review? Read the titles in order to confirm they tell the story, check that every material number has a source and that repeated figures match, confirm names and units are correct and consistent, and verify each slide makes sense on its own.
What should I remove from a deck before sending it externally? Remove speaker notes, internal comments, and tracked changes, and check the slide master and file for any leftover names or content from a previous client's version. Sending a PDF rather than the source file is a common extra safeguard.
Does an AI-generated deck still need QA? Yes. A strong tool such as Perceptis AI can produce a source-grounded, top-down, MECE-structured deck that passes much of this checklist on the first pass, but a human still needs to run the final review before anything high-stakes goes out.




