Hide the bodies, and the whole test takes four minutes on a twelve-slide deck. In most software this is the outline view, and on paper it's a sheet with the titles typed out in order and nothing else. Read them in sequence. Then write down, in your own words, the argument they make.
Three outcomes are possible and each has a different repair.
The stack makes the argument you intended. That's rare on a first draft, and when it happens the deck is close to finished.
The stack makes no argument at all, because the titles are labels. The repair is the subject-verb-number rewrite applied twelve times. It's mechanical work, about ninety seconds a slide once you have the numbers to hand. You'll see this outcome most often in a deck that was built by starting with the exhibits.
The stack makes an argument, but not the one you intended, or one with a hole in it. This is the interesting case, and it's the one that pays for the test. A hole in the argument is nearly always a missing slide, and only rarely a badly written title. So the test has just found a defect in the analysis itself, and no amount of rewording the presentation will fix it.
Advertisement
How do you run the read-only-the-titles test fairly?
The test is hard to run fairly because you already know the argument. You'll read "Gross Margin" and hear the whole story, because the story's in your head. Two tricks get around that. Give the title list to somebody who hasn't seen the deck and ask them to tell you the argument. It takes them four minutes and costs you nothing. Or leave it for a day and read the list cold. That's worse than the first option and much better than reading it warm.
Run the test at three points. On the ghost deck, before any exhibit is built, where a repair costs minutes. After the first full draft, where a repair costs an hour. And before it ships, where a repair costs an evening and a relationship with whoever is waiting for the file. Finding the same defect costs roughly ten times more at each stage, and that's the case the sketching work makes for a first version on paper.
Horizontal logic: gaps, leaps and repeats
Horizontal logic is the property that the titles, read in order, make one continuous argument. Three things break it, and naming which one you have is most of the repair.
What breaks horizontal logic in a deck?
A gap is a missing step. The stack asserts A and then asserts C, and you need B to get from one to the other, but nobody wrote it. Gaps are almost always causal. The deck says a thing changed, then says a consequence followed, and never establishes the link. The board pack in worked example 1 did this. It said channel mix moved, then said gross margin fell, and never said the first caused the second. A director asked to approve a capital deferral on the strength of a margin story can't verify that story from the pack she was given. The repair is a slide, and the slide already exists as a computation: mix took 1.27 points of the 1.85-point fall and rate took 0.58.
A leap is a step the evidence doesn't support. The stack asserts A and then asserts C, and C is bigger than A can carry. Leaps hide in the last third of a deck, where a set of findings suddenly becomes a recommendation of a size nobody has justified. The repair is either a slide that closes the distance or a smaller claim.
A repeat is the same claim twice in different words. It's the cheapest defect to fix, and the most common in decks assembled from other decks. Two slides that make one claim are one slide. Merging them usually improves both, because the two bodies together are better evidence than either alone.
One more condition looks like a defect and isn't. A deck may legitimately hold two independent chains, most often when a pack covers performance and then a separate approval item. The test is whether you announce the second chain. Give the divider page a title that states the switch, something like "The second item is the Nashik line, on which a decision is needed today", and horizontal logic holds across the seam. Switch without a word and it doesn't. Your reader spends the first two slides of the second chain trying to fit them into the first.
How should you order the titles?
All of this gives you one ordering rule, and it's worth stating plainly. Order titles so that each one either answers the question its predecessor provokes or begins a named new chain. If a title does neither, it's in the wrong place, and moving it costs nothing.
There's a cheap test for whether an order is doing any work. Shuffle the titles into a different sequence and read them again. If the argument survives the shuffle, you had a list, and anybody in the room can reorder a list. Your sequence wasn't carrying any meaning for you to defend. A real chain breaks audibly when two pages swap, because the second one starts answering a question nobody has asked yet.
You have two orders to choose from, and the choice matters. An inductive order presents the findings and lets the recommendation follow. It suits a reader who has to be convinced and will resist a conclusion offered before its evidence. A deductive order leads with the recommendation and uses the findings as support. That suits a reader who has already decided to act and needs to know what they're approving. Board packs and consulting read-outs usually run deductive at the top and inductive inside each section. So an executive summary can lead with an ask while the body pages behind it build to one.
Vertical logic: the body proves its own title, and nothing else
Vertical logic is a claim about one page. Cover the title and read the body. Then write down what the body proves. Compare that with the title, and you'll get one of four outcomes:
The body proves the title. The page is sound.
The body proves less than the title. This is the commonest defect. The title claims a cause and the body shows a correlation. Or the title claims a trend and the body shows two points, or a company-wide result over a body that shows one segment. Either shrink the title to what the evidence carries or add the evidence the title needs.
The body proves more than the title. Don't mistake this one for harmless. Put three findings under a title that names one and the other two will be missed, and they were the expensive part. Either promote the extra finding to its own page or cut it. A body that says more than its title is usually two slides.
The body proves something else. This is the dangerous one, because a reader who notices it stops trusting every other number in the deck. It gets the next section to itself.
Run vertical logic before horizontal logic on a first draft and after it on a final one. On a first draft the bodies are all you've got, and shrinking each title to what its body proves tells you what argument you can actually make. On a final draft the argument is settled, and the check is that nothing drifted while you were editing.
One practical caution about the middle case. Analysts under-claim far more often than they over-claim, because under-claiming feels safe. A body that supports a causal statement, titled with a descriptive one, has thrown away the analyst's contribution and left the reader to draw the conclusion. They'll do it less well, and they may not do it at all. The decomposition that shows mix caused 68.8% of the margin fall is a hard piece of work. Titling that slide "Gross Margin Analysis" throws the work away, however modest it feels.
When the body proves a different claim
A title-body contradiction is a different order of defect from an under-claim. An under-claim wastes credit, and a contradiction destroys it. A reader who finds one stops taking the rest of the deck on trust and starts auditing it, which is both slower and worse for you.
Worked example: a cost slide its own table contradicts
Usually it's a causal claim in the title that the author's own table refutes. Kadamba's cost slide is a clean instance. The analyst wrote "Operating costs grew faster than revenue". That's what the EBITDA fall feels like from the inside, and the table beneath it says otherwise: total operating cost grew 21.4% against revenue's 21.9%. Two of the four cost lines did grow faster, advertising at 24.7% and distribution at 26.1%, and that's presumably what the analyst was looking at when the title got written. The aggregate didn't.
Now look at what the correct version buys you. Operating cost as a share of revenue fell 0.14 points, so the entire 1.70-point fall in EBITDA margin came from the 1.85-point fall in gross margin. The two figures reconcile with nothing left over. A title that says so turns a defensive slide into the hinge of the argument. It rules out cost control as an explanation and forces the meeting back to mix, which is where the decision is.
How do you catch a title the body contradicts?
Three habits catch these before a reader does.
Read every title against its own body last, in one pass, with the deck finished. Defects appear during editing, when a chart gets rebuilt and its title doesn't.
Check aggregates whenever a title makes a claim about a total. A statement about "costs" or "the business" or "customers" is a claim about a weighted whole, and the weighted whole often moves the other way from its visible components. It's the same arithmetic that makes a mix effect possible, and of the three habits this one catches the most.
Say the title out loud with the word "because" appended, and see whether the body finishes the sentence. If the body can't finish it, the title is claiming a cause it hasn't shown.