Make a Company Deck Readable Without Flattening the Design
Make a Company Deck Readable Without Flattening the Design
A slide is readable when three things survive the trip from your laptop to wherever someone actually reads it: the main point, the evidence behind it, and the condition that limits it. Most company decks keep the first two and quietly lose the third. The qualification goes small and gray and lands in a corner, and nobody objects, because the corner was empty and the slide honestly does look better without it.
The second failure doesn't show up in a screenshot at all. The file keeps its own order — a sequence of objects that something other than a pair of eyes will follow — and that sequence is often nothing like the one you see. You can fix the visual problem completely and still leave the structural one untouched. Making the headline bigger does not do both jobs, because size is appearance and order is structure, and a readable slide needs both to agree.
So the work runs in three places: the layout, the stored sequence, and the delivered file. Below, that runs through one fictional pilot deck — built here on paper so the moving parts stay visible. No presentation was opened, exported, or read by a screen reader for this article. The deck is a construction. The checks are the ones you'd run on your own file.
Keep the qualification beside the claim it changes
The deck belongs to Harborline, a fictional freight brokerage running a three-week intake-flow pilot at four depots. Quote-to-booking time, before and after: Ashcroft 38 minutes to 27, Bellview 39 to 35, Cedar Park 42 to 36, Dunmore 41 to 40. Pooled across the 210 quotes in the pilot, the average fell from about 39 minutes to about 32 — roughly a 19% drop.
The headline captures that pooled number, and the pooled number is honest. What it needs beside it is the scope. Ashcroft carried 105 of those 210 quotes — half the pilot — so the pooled figure is Ashcroft's mean averaged with the other three depots' mean, 27 against 37. That is how a 19% company-wide drop turns out to be one depot's 29% blended with a 9% rest. Three weeks and 210 quotes is a small sample, and two of the four depots — Ashcroft and Bellview — had someone whose job was partly to run the pilot at all.
On the built slide, that caveat sits in eight-point gray under the footer rule, overlapping the chart's baseline. It reads as a caption for the footer. Detached from the number it limits, "down about 19%" becomes a claim about the company. Attached, it becomes a claim about Ashcroft's half of the traffic with company-wide ambitions — which is a different, defensible, and considerably more useful statement. Moving it didn't tidy the page. It changed the argument.
The fix is a grouping decision, not a deletion. Give the pooled number and its weighting one visual unit, because they're one thought:
Pooled average: 39 → 32 minutes Weighted by Ashcroft, which carried 105 of the 210 pilot quotes
Then the sample-size and champion notes sit in the same block, at a readable size, separated from the footer by more space than the footer itself occupies. Proximity does most of this work. A caveat can be small and still be attached, as long as it's aligned into the claim's column and closer to the number than anything else on the slide.
Give different kinds of information different visual jobs
That slide is carrying four kinds of information, and they shouldn't share a treatment. There's the claim. There's the evidence — the chart, its axis, its direct labels. There's the qualification, now attached. There's the source line for the data. Four jobs, four treatments, built from spacing, scale, contrast, and alignment rather than from font weight alone.
Contrast is where most of these decks get thin. If the four depots are distinguished only by four hues of the brand palette, a reader who can't separate the hues has a chart with no names in it. Direct labels on the bars solve it without changing the palette's character. This also happens to be how the accessibility guidance treats the problem: a US federal authoring-and-testing guide for Microsoft PowerPoint 365, checked on 2026-09-08, checks reading order, meaningful slide titles, and visual contrast as three separate items across its printed pages 1–2 and page 13. That's official guidance for one authoring context. It isn't a compliance verdict on your deck, and none of it was tested against one here. What it establishes is narrower and more useful: contrast is its own check, not a byproduct of picking nicer colors.
The projection problem needs a different answer than a rule of thumb. The same slide gets projected in a room and read alone in a PDF, and those are different conditions. Rather than adopting a universal minimum size, test the conditions you actually have: project on the wall you'll use, stand where the back row sits, and find the smallest text that's carrying meaning. If the room isn't available, take the smallest text that carries meaning and ask what it contributes at that distance. A direct label you can't read has become decoration. A source line that isn't legible is a formality nobody will check.
None of this requires flattening the design. Keep the poster-scale number, the full-bleed photograph, the palette. The test is whether a flourish is sitting where a required reading task needs to happen. If the brand mark occupies the only uncluttered surface near the claim, the mark moves — not because decoration is bad, but because that particular space is doing a job.
Check the sequence stored inside the slide
A slide title is not just the biggest text on the canvas. It's a field attached to the slide that says what the slide is about, and it can be read before the objects on the slide no matter where those objects sit. Typing a 28-point sentence into a text box at the top is not the same as putting that sentence in the title. The federal guide checks reading order and meaningful titles separately, which is the operational version of that distinction, and Microsoft's support documentation for the Reading Order pane — checked on 2026-09-18, and covering listed PowerPoint versions rather than every build — describes the pane as holding the meaningful sequence of objects, with decorative objects taken out of it and reordering capable of moving overlapping artwork. Check the exact controls in your version; labels and screenshots do move between releases.
Here's what the stored order looked like on the constructed comparison slide. Visually: headline top left, chart right, caveat bottom. As the file held it: twelve axis tick labels one at a time, then legend swatch art, then four bar shapes, then the chart's internal title, then the source footnote, then the headline text box, then the caveat, and finally the title placeholder, which had never been changed and still said "Update."
That order isn't wrong in a way anyone can see. It's wrong in the way a paragraph read out of order is wrong — the reader hears twelve numbers before learning what they measure, meets a data source before meeting a claim, and arrives at "Update" as the last thing the slide has to say.
The revision is structural, and it doesn't touch a single color. Title placeholder becomes "Pilot results: pooled quote-to-booking time down about 19%, weighted by Ashcroft." It is read first. The claim reads second. The chart arrives third, as one grouped unit whose internal order runs title, axis, direct labels, bars. The legend art comes out of the sequence, since the direct labels already name everything the swatches named. The qualification and source follow, in that order. Now the announced sequence is the argument.
Then verify it by walking the order and reading what each stop actually says — not by watching a warning disappear. A cleared flag and a usable order are two different events, and only one of them is the thing you wanted.
The adoption diagram on the second slide has its own version of this problem. If the arrows between boxes are marked decorative, the reader loses the direction of flow; if they stay in the order, a reader hears "arrow" between two box names and still has to infer what it meant. Write the relationship into the box text — "after approval, the ticket routes to Depot Ops" — and the meaning survives whether or not the arrow is announced. A decorative mark shouldn't be the only carrier of information.
Inspect the delivered format, not an attractive thumbnail
The third slide is the ask: a 12-week extension at all four depots, a named owner at each, and the same before-and-after measure. It's reasonable, and it happens to test the caveat directly, since only two of the four depots had an owner during the pilot.
But the ask is delivered as something, and that something is rarely the authoring file. Usually it's an export — a PDF, a projected copy, sometimes both. Open the export, not the thumbnail. Check four things. Clipping: text that auto-fit inside a text box in the authoring canvas can overflow where the export renders it differently. Lost distinctions: the thin rule under the header, the subtle shade that separated the qualification block from the footer, both of which can vanish at export resolution. Detached qualifications: the caveat landing on a page break away from its number, which is the original problem restored by a different mechanism. Navigation: pages arriving in order, and whatever readable structure the export preserves.
The test plan needs the reading route someone actually uses — the assistive technology on the machine that will open the file. If that route isn't available to you, say so in the record rather than assuming your visual pass covered it. And keep two claims apart while you're doing this. A judgment that a deck reads clearly and an accessibility conformance assessment are different statements with different evidence. Microsoft's pane documentation doesn't establish that an export preserves what you built. The federal guide is authoring guidance for one context, not a finding about your file. Neither conclusion follows from a clean thumbnail, and neither follows from a broken one.
The revised slide, and the record
Visually, the comparison slide changed in three ways: the caveat moved from the footer margin into the claim block, promoted from eight-point gray to a size that reads at the back of the room; the chart gained direct labels on all four bars; and the legend art came off the slide entirely, since it was naming things the labels had already named.
In stored order, it changed in four: the title placeholder received a meaningful sentence; the headline moved ahead of the chart; the chart became one group with a deliberate internal sequence; and the source and qualification were placed after the claim rather than wherever they'd been dropped. Same four depots. Same numbers. Same palette. The slide looks like the one somebody designed.
The record is short, and it needs to be written down, because none of it is memorable later:
- Environment: the application version, the export format, the reader or assistive technology used, and the conditions — projected, laptop, printed.
- What the visual pass found: where text clipped, which distinctions survived, whether the qualification stayed with its number.
- What the stored pass found: the announced order, start to finish, in the delivered file rather than the authoring file.
- Unresolved: what you couldn't check, and why.
That's the part that stays empty until someone runs it. The construction above predicts what such a record would contain; it doesn't replace the run.
What the example is meant to show is that the two requirements don't pull against each other. The worst reading problem on the comparison slide — text stacked on the chart's baseline, a qualification exiled to the footer margin — was fixed by a design decision, not a simplification. The slide got more legible because the layout finally agreed with the argument. That's the whole trade. There isn't one.
Frequently asked questions
What three things should survive for a slide to be readable?
The main point, the evidence behind it, and the condition that limits it. Most company decks keep the first two and quietly lose the qualification. The article also stresses that a file keeps its own stored order, which may differ from the visual order; a readable slide needs both appearance and structure to agree.
How should a qualification be attached to the claim it changes?
Treat it as a grouping decision, not a deletion. Give the pooled number and its weighting one visual unit, then put sample-size and champion notes in the same block at a readable size, separated from the footer by more space than the footer itself occupies. Proximity does most of the work: a caveat can be small and still be attached if it is aligned into the claim's column and closer to the number than anything else on the slide.
What is the difference between a slide title and a sentence typed at the top of the slide?
A slide title is a field attached to the slide that says what the slide is about, and it can be read before the objects on the slide no matter where those objects sit. Typing a sentence into a text box at the top is not the same as putting that sentence in the title. The federal guide checks reading order and meaningful titles separately, and Microsoft's Reading Order pane describes the meaningful sequence of objects, with decorative objects taken out.
What should be checked in the delivered export rather than in an attractive thumbnail?
Check clipping, because text that auto-fit inside a text box can overflow where the export renders it differently. Check lost distinctions, such as a thin rule or a subtle shade that can vanish at export resolution. Check detached qualifications, where a caveat lands on a page break away from its number. Check navigation, meaning pages arrive in order and whatever readable structure the export preserves. The test plan also needs the reading route someone actually uses, on the machine that will open the file.
What limits apply to the Harborline example and the guidance cited?
Harborline is a fictional freight brokerage, and the deck was built on paper; no presentation was opened, exported, or read by a screen reader for the article. The checks are the ones you would run on your own file. The federal guide is authoring guidance for one authoring context, not a compliance verdict on your deck. Microsoft's pane documentation does not establish that an export preserves what you built, and a clean or broken thumbnail establishes neither a readability judgment nor an accessibility conformance finding.