Two templates ship with these archetypes. Both are storyline skeletons, and neither is a visual theme. The first, TPL-consulting-deck, carries the 25-page plan laid out in consulting deck structure. It has a title page, an executive summary with a governing-thought placeholder and three key-line placeholders, and a context page. Then come nine body pages each with an action-title placeholder and one exhibit slot, a next-steps page, and an appendix divider with a question-indexed table. The second, TPL-project-readout, carries the five-part sequence: an objective-and-scope page, four findings pages of which one is marked as the disconfirming slot, a recommendation page with a value line, a next-steps table with owner, date and value columns, and a governance page.
In what order do you fill a consulting deck template?
The mechanical part of applying either one is where things go wrong, so here it is as an order of operations.
Fill the titles first, all of them, in the template's own placeholders, before you build a single exhibit. The titles are the argument. Build an exhibit for a title you haven't written and you can spend four hours on a chart that ends up supporting nothing. When the titles are in, read them alone, in order, with the pages hidden. If the argument follows, the deck is designed and the rest is production. If it doesn't, you've found the problem for the cost of nine sentences.
Advertisement
Here's the India title stack, read alone:
- Network contribution has risen every quarter and reached ₹8.42 an order in Q4
- One cohort of four earns all of it: metro mature contributes ₹12.1094 crore
- The other three lose ₹7.4335 crore a month between them
- Contribution per order is set by density, not by city tier (r = 0.94)
- Each cohort has its own break-even density, and it rises as AOV falls
- Tier-2 is 4.1x short of its own break-even and no plan closes that
- Metro expansion is 40 orders a day short and closes within the plan
- Closing Tier-2 recovers ₹3.7207 crore a month at once
- Reinvesting that in metro density is worth more than either alternative
Nine titles for nine body pages. Read them in sequence and the argument moves without a gap: the good news, the concentration, the loss, the mechanism, the mechanism's implication, the cohort that cannot be saved, the cohort that can, the size of the prize, and where the prize should go. Page 4 is the hinge. Take it out and the argument jumps from a fact about cohorts to a conclusion about closure, with no reason attached.
How do you build the exhibits and the appendix?
Only now does exhibit production start, and you build each exhibit against the claim its title already makes. Page 4's claim is about a relationship, so its exhibit is a scatter of 24 stores, density against contribution per order, with the Pearson correlation of 0.9375 stated and the zero line drawn. Page 2's claim is a comparison across four categories, so it gets a sorted bar. Page 1's is change over ordered time, so it gets a line across four quarters. None of that mapping is a matter of preference, and the reasoning behind it belongs to chart craft.
Last comes the appendix sweep. Every exhibit you built that didn't earn a body page goes to the appendix with an index entry phrased as a question. Every body exhibit gets its source line checked for the date. In the India deck that sweep moved 8 exhibits and took about twenty minutes, the cheapest twenty minutes in the whole build.
{
"skill": "Stress-testing an archetype's title stack and appendix index before any exhibit is produced",
"use": "Once you have filled every action-title placeholder in the template and written the appendix index as questions, paste the title stack and the index into an assistant and ask it to attack the sequence: which title states no claim, which two titles make the same claim, where the argument jumps, and which of the six hardest questions a client would ask has no index entry. Fix the stack yourself; do not paste back a stack it wrote.",
"tools": ["Chat assistants (Claude, ChatGPT, Gemini)", "A plain text file holding the title stack, so no deck exists yet", "Your own computation script or spreadsheet for every figure in the titles"],
"prompt": "Below is the action-title stack of a nine-page consulting deck body, in order, followed by my appendix index written as questions. Do five things and nothing else. 1) For each title, say whether it states a claim someone could dispute, or whether it is a topic label. 2) Name any two titles that make the same claim. 3) Name the point in the sequence where the argument jumps, meaning where a reader who accepted title N would not yet accept title N+1. 4) List the six questions a sceptical client would ask about this argument, and for each say whether my index answers it. 5) Name the one claim in the stack that would most damage the argument if it turned out to be wrong. Do not rewrite my titles and do not propose new ones. TITLES: <paste>. INDEX: <paste>.",
"verify": "Every figure inside a title has to be reproduced by your own computation before the title ships: recompute it, and confirm that the rounding shown in the title matches the rounding in the exhibit built for that page. Where the assistant says a title states no claim, test it yourself by writing the opposite sentence; if the opposite is absurd rather than merely wrong, the title is a fact and not a claim. Where it names a gap in the sequence, the fix is a page you write, and the test that it worked is reading the titles aloud again with the pages hidden. Where it names an unanswered client question, either an appendix exhibit exists and the index entry is missing, or the exhibit does not exist and you now know what to build. Never accept its claim that two titles are duplicates without checking the underlying figures, because two titles about the same cohort with different figures are two claims.",
"diy": "You must be able to write nine action titles cold from a fact set, in order, with no tool open, and hear the gap yourself when you read them aloud. That is the gated skill and it is what the archetype exists to hold. An assistant can find a jump you missed at 11pm; it cannot decide which claim is the governing thought, because that decision is a judgement about what the client should do and you are the one who will defend it in the room.",
"market": "IN"
}