Presentation standard and brand

Written by a person. Last read by a person on 2026-09-07, 21 days ago. Its facts were checked by the eval suite on 2026-09-28.

You are building the readout, you are judging whether a deck was designed or decorated, or you want the reasoning behind why an executive deck is shaped the way it is.

Scenario-based sample. Halden Systems is invented, and so is every figure about it.

Bottom line. 9 slides, 16:9, one idea each. The ask is on slide 1 and the decision is on slide 9. Everything here is written so that somebody who was not in the room can build the deck and get the same result.

The brand

The mark

Halden Systems uses a wordmark with a single glyph, and no illustration.

The cut corner is the whole idea: a system with a piece taken out of it. Do not round it, fill it, animate it, or add a second one.

Color

5 values, and every pair below has been measured.

The last 4 exist because signal and slate were each used against two backgrounds and each cleared AA against only one. A palette of 5 that fails half the time it is used is a palette of fewer than 5. Each variant is the nearest color to its parent that passes.

Every ratio here was measured rather than estimated. The first four were written from memory, and each was wrong by a tenth or two. That is close enough to be believed and far enough to matter. Signal on paper was recorded as 3.4:1, which clears the 3:1 a large heading needs. The measured figure is 2.8:1, which does not.

The last row is the one that gets broken. Signal is a bar color on a light slide and a text color only on a dark one. It was broken in 4 places, on the 31 percent figure and on the total row of the cost table, and nothing caught it for a month. A darker orange, #B8481A, carries the same meaning at 4.9:1 and is what small signal text uses now.

Type

One family, 2 widths, 3 weights. A deck that needs a fourth weight needs fewer words.

Söhne is the intent and Inter is what ships. Söhne is licensed and Google Slides cannot load it, so the built deck uses Inter and Roboto Mono, which are close in width and have the same flat, unfussy shapes. A standard that names a face nobody can install is a standard that gets ignored quietly.

Nothing below 37px, which is 28pt. The old floor here was 22px, and it was set by looking at a deck on a laptop. A slide is not read, it is projected, and the number that matters is the one a person 40 feet back can resolve. 37px is 14pt against a 405pt reference slide, scaled to this canvas, and it is the floor slidekit check enforces by default.

The deck used to break the old floor in 23 places and the new one in 90. It no longer breaks either. Raising it cost one slide its layout: the value case put a heading and a full sentence in a 225px-wide column, which held at 21pt and collided at 32pt, so the column is wider now and the bars start further right. That is the usual price of a floor, and it is worth paying once.

Grid and safe area

1920x1080, 12 columns, 80px gutters, 120px outer margin. Title sits on the baseline 200px from the top and never moves between slides, because a heading that jumps is the thing the eye tracks instead of the content.

The same canvas in points is 1440x810. A pixel here is three quarters of a point. A browser pixel is a 96th of an inch and a point is a 72nd, so the two units differ by exactly that much. The Slides API speaks only in points and reports this deck as 1440x810pt. A standard written in pixels alone sends anybody working against that API to the wrong numbers, and it did.

Nothing that carries meaning goes within 120px of any edge. That is wider than broadcast safe area and exists for the room, not the camera: the bottom of a screen is behind somebody's head.

The wordmark is the one exception and it is deliberate. It sits 24px from the bottom left corner, carries nothing a reader has to recover, and belongs beside the composition rather than in it. Every other element holds the margin.

What we do not do

No stock photography of people at laptops. No gradient meshes. No drop shadows. No icon sets where a lightbulb means an idea. No logos of vendors we have not chosen. No build animations, because a build tells the room you are about to say something they cannot yet read, and they stop listening and wait.

How these slides are built

The deck is 9 slides for a 30-minute meeting: about 12 minutes of talking, and the rest is theirs. That ratio is the design.

One idea per slide, and the headline is the idea. The headline is a sentence with a verb, not a label. "78 percent can follow the steps. 31 percent can say what they do" is a headline. "Survey results" is a filing category.

The slide carries the evidence and the speaker carries the argument. Nothing on a slide is read aloud. If a line is only there to be read out, it belongs in the note.

One visual per slide, and it is the claim. A chart that shows the gap is the claim. A chart that shows everything we measured is a chart the room reads while you talk.

Progressive disclosure across slides, never within one. Each slide assumes the one before it and adds a single move. Adults following an argument hold about 4 new things; 9 slides at 1 each is inside that, and 9 slides at 3 each is not.

Repeat the sticky sentence 3 times. They can follow the steps, they cannot say what the steps do. Slide 2 states it, slide 6 uses it, slide 9 closes on it. A sentence heard 3 times in 12 minutes is the sentence quoted in the next meeting you are not in.

Relevance before evidence. An executive audience gives attention to a problem they own before they give it to a number. Every slide answers "why is this mine" before it answers "how big".

The ask is first and last. Primacy and recency are the only 2 positions anybody keeps. Everything in between is support.

Two things the built deck needed and this standard had not said

Both were found by generating the deck and comparing it against these pages, which is the only way a specification is tested.

A disclaimer slide, before the ask. Every artifact in this set carries a line saying the company is invented, and a deck leaves that set the moment somebody exports it. It is slide 0: the mark, and one sentence reading "This is not a real presentation. It is an example from philhwagner.com." The 9 slides below are unchanged and the deck is 10.

A title. A deck inherits a filename, and a filename is what a recipient sees in a share sheet, a tab and a meeting invitation. It is "Funding Proposal: Closing the $6.1M Gap": the subject and the number, so the deck is findable months later by somebody who remembers neither the sender nor the date.

One deviation was right and is now the rule. The 2 figures on the ask slide render in paper rather than signal. 2 numbers in the accent color compete for the same emphasis and neither wins, so the accent is spent on 1 thing per slide as the color section says, and on that slide the 1 thing is the rule between them.

The 9 slides

Each entry gives the headline, the one visual, what is on the slide, what the speaker adds, and what is deliberately absent.

Slide 1. The ask

Headline: We are asking for $2.1M over 3 years to close a gap costing $6.1M a year.

Visual: None. Type only, centered, on ink.

On the slide: The 2 figures at 128px, signal on ink, side by side with a thin slate rule between them. Below, in 28px paper: "The decision today: fund it this quarter, or tell us to keep absorbing the cost."

The speaker adds: That procurement is already moving and no contract is signed. Then stops talking. The silence after the ask is doing work.

Absent: Agenda slide. An agenda tells a room the ask is coming later and teaches them to wait for it.

Slide 2. What is broken

Headline: They can follow the steps. They cannot say what the steps do.

Visual: 2 horizontal bars, 78 and 31, wash and signal, on paper. No axis, no gridlines, no legend. The 2 numbers sit at the end of their bars in mono.

On the slide: One line beneath: "4 years of moving work into engineering systems. Every move run as a tooling rollout."

The speaker adds: Our dashboard reports the 78 and calls it success. That sentence is the first time the room hears that we have been measuring the wrong thing.

Absent: The other 14 survey items. They are appendix A1.

Slide 3. What it costs to leave alone

Headline: $6.1M a year, and every figure is already in a system we own.

Visual: A 4-row table, mono, right-aligned figures, one signal row for the total.

On the slide: 760 engineer hours a week. $145 fully loaded. $5.7M a year. $410,000 duplicate regional spend.

The speaker adds: None of this needs a new measurement. The interruptions are in calendars, the spend is in 9 regional budgets, and the tickets are in a queue.

Absent: Any projection. Every number is the year we just had.

Slide 4. Where the value is

Headline: Revenue is the biggest number here and the weakest evidence.

Visual: 3 stacked blocks, largest at the top, wash for revenue and signal for the 2 beneath. Size is proportional to the claim, which is the point.

On the slide: Each leg carries its own figure, and the figures are named here because revenue has no dollar amount and asking for one produces the overclaim the next slide refuses.

Revenue's is a quantity rather than a currency amount, on purpose. A dollar figure would need an attribution argument we have already said we cannot win, and putting one on a slide is exactly how a learning function gets caught.

The speaker adds: The ordering, out loud, before anybody spots it.

Absent: A total. Adding these together would be the overclaim this slide exists to refuse.

Slide 5. The number we will not defend

Headline: Every learning function that has overclaimed revenue has been caught.

Visual: None. A single sentence on ink, 54px, with the rest of the slide empty.

On the slide: Nothing else. The empty slide is the emphasis.

The speaker adds: We lead on revenue because it is the largest, and we rest the case on the 2 below it. We would rather say so first.

Absent: Any hedge. This slide is a concession, and a concession with a caveat is a defense.

Slide 6. Why more training will not fix it

Headline: 4 findings, and each one rules something out.

Visual: 4 lines of text, each with a signal rule to its left. No icons.

On the slide: One program for everyone degrades the experts in it. Proximity beats training. Recordings skip the thinking. Paying for completion buys completions, and completion is at 78 percent.

The speaker adds: So the incentive is contribution, not completion.

Absent: Citations. They are appendix A5, and reading a citation aloud is how a room learns you are unsure.

Slide 7. AI is a chapter, not a program

Headline: 22 percent can tell which answers need checking.

Visual: 3 bars, 71 / 38 / 22, descending, with only the last in signal.

On the slide: McKinsey's 88 and 6, in mono, small, beneath the bars.

The speaker adds: The descent is the story. Adoption is fine and judgment is not, and nobody measured judgment before the rollout.

Absent: The word "AI" in the headline. The subject is the capability.

Slide 8. How you will know it worked

Headline: And how you will know it failed.

Visual: One real table. A header and 4 level rows, with a reachable column and a not reachable one. The rows are fixed and so are their answers, because this is the slide most likely to disagree with the strategy and a reader who spots that has stopped believing both.

It was 21 shapes arranged to look like a table until 2026-09-08: 12 text boxes and 9 rectangles drawn as rules. That is why level 04 was missing for a month. Adding a row meant moving every shape below it by hand, so nobody did, and the slide quietly contradicted the paragraph under it. A table takes a row.

The slide carries the short form of that column, a yes or a dash, with revenue in signal on the last row. The sentences above are what the speaker says, not what the room reads.

Level 4 is never dropped for space. It is the level the revenue argument rests on, and a deck that omits it has removed the concession that makes the rest credible.

On the slide: The 2 failure conditions in signal beneath the table: escalations do not fall within 2 quarters, or completion rises while understanding does not.

The speaker adds: Level 4 revenue is honestly not reachable, which is why it is ordered where it is.

Absent: A target date for each level. Dates here invite a schedule conversation instead of a funding one.

Slide 9. The decision

Headline: Fund $2.1M over 3 years, this quarter.

Visual: None. Type on ink, matching slide 1 so the room recognizes it.

On the slide: What the money buys, in 3 lines: a platform, 2.5 people, and content written here. What you get in 90 days. What you do not get, which is a signed vendor contract.

The speaker adds: If it cannot be run by somebody who did not design it, we will tell you and stop it. The kill criteria are ours to enforce.

Absent: A thank-you slide. The last thing on the screen during the discussion should be the decision.

Where the built deck is

The deck produced from this standard is published on Google Slides, and the executive readout links to it. It is linked rather than embedded, because this site promises that nothing third-party loads on it and an embed would break that promise on every page carrying one.

Two things would need fixing before the embed code Google hands out is used anywhere. It is a fixed 960 by 569, which overflows a phone and does not scale, so it needs a wrapper holding a 16:9 ratio instead. It also carries no title, so a screen reader announces it as "frame" and a reader has no way to learn what it holds before entering it.

The appendix

6 slides, never presented, each reachable by name when asked. Survey instrument and response rates. The site confidence gap. The 3-year total cost of ownership and build against buy. The evaluation matrix with its weights. The literature review and the claim we refuse to cite. What we are deliberately not solving.

An appendix that is walked through is a second deck. Its only job is to make a hard question end in an answer rather than in a follow-up meeting.