Three things make a slide usable by someone using a screen reader, and all three make it better for everyone else.
What is reading order, and how do you check it?
Reading order is the sequence in which assistive technology announces the objects on a slide. It's the pane's list read from the bottom entry upward, and that's the order the objects were created in. Where they appear on screen has nothing to do with it. A slide whose five steps were pasted in the order 3, 1, 5, 2, 4 announces them in that order. All five steps are announced in a position that does not match the visual layout, and it takes 4 adjacent swaps to restore the visual order. Nothing about the slide looks wrong. That's how 17 of the 42 slides in the audited deck came to have the defect without anybody noticing. The check takes ten seconds. Press Tab repeatedly on the slide and watch where the focus goes. If it doesn't follow the order a reader should meet things in, fix the pane. Moving things on the slide won't help.
Advertisement
What is alt text, and how much work is it?
Alt text is the sentence a screen reader announces in place of an object. Two facts make this cheaper than it looks. First, most objects don't need it: of 486 objects on the audited deck, 173 carried content and 313 were decorative and could be flagged as such in one action each. Second, alt text for a chart carries the chart's message, and you've already written that message in the title. "Bar chart showing revenue by service line" tells a listener nothing. "Radiology contributed 55.5% of this year's growth of ₹155 crore" tells them what the sighted reader gets in one glance.
The pass costs about what you'd expect, and less than people fear. On that deck, 154 alt-text entries at about 25 seconds each is 64.17 minutes, and 313 decorative flags at about 3 seconds each is 15.65 minutes, for a total of 79.82 minutes, or 1.9 minutes a slide. Doing it as you build costs almost nothing. Doing it at the end costs an afternoon. Not doing it costs a reader the deck.
What contrast does a slide need?
Contrast is the one part that is pure arithmetic, and the thresholds are published. Under WCAG 2.1 at level AA, body text needs a contrast ratio of at least 4.5:1 against its background. Large text, meaning 18 pt or 14 pt bold, needs 3:1, and so do graphical objects and interface components. Contrast is computed from relative luminance. You can't estimate it by eye, and two colours that look clearly different can measure 1.4:1.
The gradient ladder from the opening is the case worth working. Five bands cut from a ramp between #1F3B57 and #7FA8C9 have adjacent-pair contrasts of 1.51, 1.51, 1.44 and 1.40 to one. Zero of the four pairs reaches 3:1, so the encoding is not distinguishable to a reader with reduced colour discrimination, and it is marginal for everyone else. The end-to-end contrast between band 1 and band 5 is 4.58:1. So the ladder looks fine at a glance and fails the moment a reader has to tell band 3 from band 4. A white label placed on band 1 measures 11.53:1 and passes comfortably. The same label on band 5 measures 2.52:1. It fails both thresholds and would have to be 1.79 times better to reach the body-text line.
How do you fix a low-contrast ramp?
The repair follows from the arithmetic, and taste doesn't come into it. Adjacent ratios multiply along a ramp, so a ramp measuring 4.58:1 end to end cannot hold two adjacent gaps of 3:1, which would need 9:1 of range. Five categories are hopeless on it and three are out of reach. Two is what it carries. The ceiling is low even at the extreme, because white against black is 21:1 while four levels would need three gaps of 3:1, which is 27:1, so no lightness ramp separates more than three levels. Either the number of encoded levels comes down to what the ramp in use can actually hold, or the encoding changes from lightness to position, length or an explicit label. That's the general form of the accessibility fix. It also explains why the accessible version is usually the clearer one. The threshold forces you to encode the distinction in something stronger than a shade.
Verify (checked 2026-09-08): the thresholds quoted are WCAG 2.1 level AA. WCAG 2.2 is published and adds criteria without changing these three ratios, and jurisdictional requirements differ. Check the standard your organisation is held to, and check the built-in accessibility checker's coverage in your release, since it finds missing alt text and some contrast problems but does not find every reading-order defect.