Deck benchmark v1
The inputs every deck tool gets in a hands-on review – so you can check our scores and run the
same test yourself. These are fictional test fixtures, not review results.
Version apm-deck-benchmark-v1, published 2026-10-16.
Files
| File | Purpose | SHA-256 |
|---|---|---|
| brief.txt | Task 1 – 120-word brief | 4e805bbe04803923… |
| lumenfield-q3-2026-report.pdf | Task 2 – 4-page source report with one table and two charts | 687276ffa3320b02… |
| lumenfield-template.pptx | Task 3 – PowerPoint template (Georgia/Verdana, #1F4E79, 16:9) | 7e7f4da0c4e2ae91… |
| expected-facts.json | Every figure a correct deck must reproduce | 4498c17995c75ec1… |
| manifest.json | Version, date and SHA-256 checksums | – |
Lumenfield Analytics, its customers and all figures are invented for this test. A new version gets a new number; v1 files are never edited.
The three tasks
- Prompt to deck (brief.txt): Paste the 120-word brief unchanged and ask for the 10-slide sales deck it describes.
- Document to deck (lumenfield-q3-2026-report.pdf): Upload the 4-page report (one table, two charts) and ask for an 8-slide summary; every figure is checked against expected-facts.json.
- Brand template and export (lumenfield-template.pptx): Apply the template (Georgia/Verdana, #1F4E79), export to PPTX and open it in PowerPoint and Google Slides: editable text, native charts, fonts kept, speaker notes present.
Procedure
- Use a fresh account on the cheapest plan that offers the features being tested (record which plan). Note the date and the tool version if shown.
- Task 1: paste brief.txt unchanged. Accept the first complete result; one follow-up prompt for missing slides is allowed and recorded.
- Task 2: upload lumenfield-q3-2026-report.pdf with the instruction “Summarize this report in 8 slides for the board.” If the tool can’t take files, task 2 is recorded as not possible – not scored zero on other criteria.
- Task 3: apply lumenfield-template.pptx (or the closest brand-template feature), export to PPTX, open the file in PowerPoint and import it into Google Slides.
- Check every figure in the task 2 deck against expected-facts.json. Count invented figures, quotes and sources separately.
- Keep the generated files and screenshots of the deck and of the exported PPTX in PowerPoint; they are published with the review.
- Score each applicable criterion with the anchors below. Write one sentence of rationale per criterion.
Scoring anchors
0–5 per criterion; 2 and 4 sit between the anchors.
| Criterion | 0 | 1 | 3 | 5 |
|---|---|---|---|---|
| Slide design (20%) | Unusable: overlapping or cut-off text, broken layouts. | Generic layouts, inconsistent styles; most slides need rework. | Clean and consistent; a few slides need layout fixes. | Presentation-ready: clear hierarchy, consistent, on-brand without fixes. |
| Structure & storytelling (15%) | Ignores the brief: wrong slide count or missing required parts. | Required parts present but in a confusing order; no speaker notes. | Follows the brief with a sensible order; speaker notes thin or missing on some slides. | Clear narrative from problem to call to action; useful speaker notes on every slide. |
| Factual accuracy (15%) | Invents several figures or contradicts the source. | One or more figures invented, or several transcribed wrongly. | All key figures correct; minor rounding or labeling errors; nothing invented. | Every figure matches expected-facts.json; nothing invented; states what the source does not say (e.g. no YoY growth). |
| Export fidelity (20%) | No PPTX, or the file doesn’t open. | PPTX of flattened images; text not editable. | Text editable; charts are pictures or fonts substituted; minor layout shifts. | Text, native charts, master slides, fonts and speaker notes survive in PowerPoint; imports cleanly into Google Slides. |
| Ease of editing (10%) | Could not complete the tasks. | Completed with heavy manual work or support help. | Completed with some trial and error; revisions possible but clumsy. | All tasks completed quickly; revisions and template use are straightforward. |
| Privacy & business terms (20%) | No privacy policy or DPA; trains on uploads with no opt-out. | Policy exists; no DPA; training on uploads by default. | DPA available on a business plan; sub-processors listed; opt-out documented. | DPA on the plan tested, documented server location and sub-processors, no training on uploads by default, admin/SSO and invoices with VAT. |
Which tools this applies to
The benchmark is a deck test. It is only used for tools that produce presentations; for everything else the output criteria are not applicable and drop out of the score instead of counting as zero.
| Category | Criteria scored |
|---|---|
| AI Presentation Makers | Slide design · Structure & storytelling · Factual accuracy · Export fidelity · Ease of editing · Privacy & business terms |
| PowerPoint & Google Slides Add-ins | Slide design · Structure & storytelling · Factual accuracy · Export fidelity · Ease of editing · Privacy & business terms |
| Pitch & Sales Decks | Slide design · Structure & storytelling · Factual accuracy · Export fidelity · Ease of editing · Privacy & business terms |
| Interactive & Live Presentations | Slide design · Structure & storytelling · Factual accuracy · Ease of editing · Privacy & business terms |
| Diagrams, Charts & Infographics | Slide design · Factual accuracy · Ease of editing · Privacy & business terms |
| Images & Design for Slides | Ease of editing · Privacy & business terms – no deck test; output criteria not applicable |
| AI Assistants for Decks | Slide design · Structure & storytelling · Factual accuracy · Export fidelity · Ease of editing · Privacy & business terms |
| Video & Voice Presentations | Ease of editing · Privacy & business terms – no deck test; output criteria not applicable |
| Writing & Research | Ease of editing · Privacy & business terms – no deck test; output criteria not applicable |
| Speaker Coaching & Rehearsal | Ease of editing · Privacy & business terms – no deck test; output criteria not applicable |