# Discovery To Proposal Deck

> Discovery to Proposal Deck

- Skill: `ingridleiria/discovery-to-proposal-deck` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ingridleiria/discovery-to-proposal-deck`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ingridleiria/discovery-to-proposal-deck/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- Author: ingridleiria (https://skillmd.com/u/ingridleiria)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/ingridleiria/discovery-to-proposal-deck

---


# Discovery to Proposal Deck

The deck that follows discovery has one job before it has any other: prove the problem was heard correctly. Most of them skip it. They open with the firm, move to a methodology, and present a solution to a problem the client recognises only in outline. The client then spends the meeting correcting the situation rather than reacting to the approach, which costs a cycle. Or, more often and more expensively, they do not correct it. They thank you, say they will discuss it internally, and buy from whoever played their situation back to them in their own words.

That is the failure this skill prevents, and the cost is not the deck. It is the second discovery round that has to be arranged without admitting the first one was wasted, the three weeks that adds to a decision, and the credibility spent asking a busy operations director to explain their problem twice. A client who reads the first section and thinks "yes, that is our situation" will argue about the approach instead, which is the only argument worth having in that meeting, because it is the argument that ends in a scope.

## When to use this, and when not to

Use it after discovery with a named client: one call, a set of interviews, a workshop, or a diagnostic phase that has finished. Use it when what is needed is agreement on the shape of the work rather than a price and a signature. Use it when the client has a problem, a rough budget expectation, and no agreed approach.

Do not use it where there is no named buyer. A concept you want to propose to the market, a new service line, an idea to pitch internally: that is `ideation-deck` in the chief-of-staff track, which starts from an idea nobody has approved and has to manufacture the "why now" from evidence you assembled. This skill starts from one named client's discovery notes and the "why now" is something they told you. The two decks look similar and are built from opposite ends: `ideation-deck` argues a concept into existence, this one plays a situation back.

Do not use it for the document that asks for a yes at a price. That is `proposal-writer`, which carries the investment section, the commercial terms and the signature path. Do not use it for the contractual boundary of the work, the deliverable acceptance criteria, the exclusions and the change control: that is `sow-and-scope`, also in the chief-of-staff track, and it comes after approval, not before. Do not use it to rebuild a standard capability deck around one account, which is `tailored-client-deck`. This deck exists in the gap between the discovery conversation and the priced proposal, and where the client is ready to buy without it, skip it and go straight to `proposal-writer`.

## What you need before starting

**Discovery notes or transcripts, with attribution.** The raw material. Every point in the understanding section will be traced to one of these, so notes without speakers are half useful. Missing: reconstruct within twenty-four hours of the call while recall is intact, and mark the reconstruction as such in your own file so you never quote it as verbatim.

**The client's own phrasing for the problem.** Their nouns, their internal names for teams, systems and processes, their unit of measure. Missing: this is the one input worth going back for. Send two sentences and ask "is this how you would put it", which takes the client ninety seconds and is not a burden.

**Who was in the room, and what each of them is measured on.** The deck is read by people with different scorecards, and an objective that serves one of them and threatens another will stall without anyone saying why. Missing: ask, or infer from title and state the inference in your own notes rather than in the deck.

**Any figures the client stated.** Volumes, headcount, cycle times, cost, error rates, targets. These are the only numbers permitted in the value section. Missing: the value proposition describes the mechanism rather than the magnitude, and says so.

**The constraints they named.** Budget expectation, timing, internal capacity, a system freeze, a committed programme, a person who has to approve. Missing: ask before building a plan that will collide with one. A phase plan that ignores a stated freeze reads as inattention.

**What they have already tried.** Nothing damages a recommendation faster than proposing something the client attempted last year. Missing: ask directly in these words, "what have you already tried here", and expect the most useful answer in the whole of discovery.

**Who approves, and what their decision process is.** Committee, single owner, budget cycle, procurement threshold. This decides the next step slide, and a next step that asks for a decision nobody in the room can make wastes the meeting. Missing: name the assumed approver on the next steps slide so the room corrects you.

**Open questions from discovery.** The things you did not get to. Missing: there is no such thing as discovery with no open questions; if the list is empty, it was not written down. Put them in the deck as open questions rather than filling them with plausible invention.

## The method

1. **Build an evidence ledger before writing a slide.** One line per usable statement: the quote or close paraphrase, who said it, which session, and what it evidences. Twenty to forty lines from a good discovery round. This is the source of truth for the whole deck, and the rule that makes the deck defensible is that anything in the understanding section must cite a ledger line. Judgement call: whether a paraphrase counts. Rule: a paraphrase is usable if the speaker would accept it read back verbatim; if you are not sure, mark it and check it in the meeting rather than asserting it.

2. **Separate what they said from what you concluded.** Two columns in the ledger: client statements feed the understanding section, your conclusions feed the hypothesis. Mixing them is the most common structural error, invisible to the writer and obvious to the client, who reads a sentence about their own business and does not recognise it.

3. **Name the constraint, as a hypothesis.** One sentence: given what we heard, we believe the binding constraint is X, and relieving it would produce Y. Constraint, not problem list. A discovery round produces eight problems and usually one thing that holds the rest in place. Judgement call: which one. Rule: pick the constraint that, if relieved, changes the most other symptoms in the ledger, and where two compete, pick the one the client can act on without a decision from someone outside the room.

4. **Write three to five objectives, each measurable and each traceable.** Measurable means there is a number and a way of observing it. "Improve enablement" is not an objective. "Cut ramp time for a new sales hire from five months to three, measured by date of first closed deal" is. Traceable means each one points at a ledger line. Judgement call: how ambitious. Rule: set the target at the level the client themselves named, or one the evidence supports; a target you invented to look impressive becomes the standard you are held to.

5. **Choose the shape of the engagement, and let the client's risk position choose it.** Where budget authority sits in the room and the problem is bounded, propose the whole engagement in phases. Where budget is uncertain, trust is unproven or the constraint is genuinely unclear, propose a short paid diagnostic first with a decision point at its end, and say plainly that the second phase should not be committed until the first has run. A firm that offers the small version first wins more work in total, because the client's real objection is rarely price, it is risk.

6. **Write the design principles before the phases.** Two or three sentences governing choices inside the work: nothing is built that the client's team cannot maintain, say, or every recommendation is tested at one site before rollout. Principles matter more than the phase diagram, because they are what the client holds you to when month two goes unexpectedly, and they are what separates an approach from a schedule. Two rules make them real rather than decorative. Each one has to be traceable to something the client said in discovery, named in the note where it appears, because a principle nobody asked for is a preference. And each one has to be capable of ruling something out: write down the specific thing this principle forbids you from doing, and if nothing comes to mind, the sentence is a value statement and not a principle. Two principles that would rule out the same choice are one principle, so cut to the sharper wording.
7. **Name deliverables as artefacts, with a format and an owner.** "Workshops" is an activity. "A one-page operating rhythm, in the client's template, owned by the operations lead" is an artefact. The test: could an unfamiliar person walking past the desk in month three point at it. Where a deliverable genuinely is a capability rather than a document, say what evidence will show it exists.

8. **Build the phase plan around the client's dependencies, not yours.** Each phase gets a duration, a milestone that is a verifiable event rather than a percentage, and the client-side inputs it needs: data, access, people's time in hours, decisions with dates. Client dependencies are what actually slip a plan, and showing them is not defensive, it is the part experienced buyers look for. Judgement call: how granular. Rule: name any client input that, if it arrived two weeks late, would move the end date.

9. **Size the value only with the client's own numbers.** Where the ledger has volumes and rates, do the arithmetic in front of them and show the working on the slide. Where it does not, describe the mechanism, name the figure you would need to size it, and ask for it. A benchmark borrowed from another industry undoes the credibility the understanding section built, and an experienced buyer recognises a borrowed number immediately.

10. **Write the understanding section last, then run the read-back test.** It goes first in the deck and is written after the approach is settled, because only then do you know which parts of the situation matter. Then read the section aloud as though you were the client's operations director. Any sentence you would not say about your own business is rewritten or cut.

11. **Set one next step, with an owner and a date.** One primary ask. Where the approver was not in the room, the next step is the meeting that gets them there, not the approval itself.

## Slide by slide

Fourteen to eighteen slides for a full engagement, eight to ten for a diagnostic. Every title is a sentence that carries the message, so a slide titled "Approach" says less than "Fix the qualification gap first, then automate". One idea per slide.

| # | Slide | What is on it | Source |
| --- | --- | --- | --- |
| 1 | Title and purpose | Client name, date, and one line on what this meeting is for | Your framing |
| 2 to 4 | Our understanding | Three or four points, each a client statement | Evidence ledger |
| 5 | What we heard that surprised us | One point, optional, that shows you listened past the brief | Evidence ledger |
| 6 | Solution hypothesis | The constraint, stated as a hypothesis inviting correction | Your conclusion |
| 7 | Engagement objectives | Three to five, measurable, each tied to the understanding | Ledger plus targets |
| 8 | Approach | The shape of the work in one diagram | Your design |
| 9 | Design principles | Two or three, in plain sentences | Your design |
| 10 | Deliverables | Named artefacts, format, owner | Your design |
| 11 to 12 | Phases and milestones | Duration, verifiable milestone, client inputs | Your design |
| 13 | What we need from you | Data, access, hours by role, decisions with dates | Your design |
| 14 | Value | Sized with their numbers, or the mechanism plus the figure needed | Ledger |
| 15 | Open questions | What discovery did not settle | Ledger gaps |
| 16 | Next step | One ask, one owner, one date | Your framing |

Nothing about the firm appears before slide six. Credentials, if they appear at all, sit at the back and are chosen for this client's sector.

## Worked example

**Situation.** A twelve-person operations advisory firm ran discovery with Ardent Vale Foods, a chilled-goods manufacturer with three sites and about 600 staff. Discovery was three interviews across two weeks: the chief operating officer, a site manager, and the head of planning. All figures in this example are in US dollars. The presenting problem, as briefed, was "we need better production planning". The firm had ten days to come back with something the chief operating officer could take to a budget conversation on the 26th.

**Task.** A deck that would get agreement on the shape of the work, not a price. Good meant the chief operating officer forwarded it to the finance director without needing to explain it, and that the meeting was spent on the approach rather than on the situation.

**Action.** The first build was wrong and took two days. It followed the firm's standard five-phase operating model, ran to 34 slides, opened with the firm's credentials and a slide on the model itself, and reached the client's situation on slide nine. A colleague reviewed it and asked one question that killed it: which of these sentences did they actually say. The answer, checked against the notes, was two of eleven. The rest were sector-standard statements about planning maturity that would have been true of any chilled-goods manufacturer, and the client would have read them as a template with their logo on it. Two days were discarded.

The rebuild started with the evidence ledger: 31 lines from the three interviews, with speaker and session on each. Two things came out of it that had not survived the first version. The head of planning had said, in passing, that the weekly plan was rebuilt by hand every Monday because the forecast arrived after the plan was due, which was a sequencing problem rather than a planning capability problem. And the site manager had said his team ignored the central plan in the last week of every month because of a month-end volume push, which meant the planning system was being overridden by an incentive nobody had mentioned.

The hypothesis changed accordingly: the binding constraint was not planning capability but the order of two inputs, the forecast and the plan, and a month-end incentive that made the plan unfollowable in week four. That was one sentence on slide six and it was the sentence the meeting turned on.

Objectives were cut from six to four, all measurable, all traceable to ledger lines: move the forecast to Thursday of the preceding week measured by delivery date; cut Monday replanning time from a stated six hours to under one; hold plan adherence in the final week of the month above 80 percent, from a stated baseline the head of planning put at "about half"; and give the planning team one owned artefact they maintain themselves.

The shape chosen was a five-week paid diagnostic at 24,000 US dollars with a decision point at the end, rather than the full engagement at roughly 145,000. Budget authority was not in the room, the month-end incentive touched sales rather than operations, and the honest position was that the second phase should not be committed until the incentive question had an owner. The deck said that in a sentence.

Value was sized with the client's own numbers only: six hours of replanning weekly across three sites at a stated loaded rate, which came to a figure the client had supplied every component of. The forecast benefit of plan adherence was left unsized, with the sentence naming the one figure needed to size it, the cost of a short-notice line changeover, which nobody in discovery could give.

The deck came to sixteen slides. Nothing about the firm appeared before slide six.

**Result.** The chief operating officer forwarded it unedited to the finance director the same afternoon. The meeting on the 26th spent most of its time on the month-end incentive, which was the intended argument, and produced an outcome the deck had not proposed: the diagnostic was approved at 24,000 with the head of planning named as the internal owner, and a sales director added to the second-phase decision.

The full engagement was signed eleven weeks later at 132,000, below the original estimate, because the diagnostic showed two of the five workstreams were unnecessary. The firm treated that as the correct outcome rather than a lost 13,000, on the reasoning that a scope the client trusts is worth more than one they tolerate.

### A second scenario, where it goes differently

The same firm ran discovery with Bellamy Rail Services across five interviews and found the two senior stakeholders held incompatible views. The engineering director described a maintenance scheduling problem. The finance director described a contract profitability problem and called the scheduling work "a distraction from the commercial issue". Both statements were in the ledger, attributed, and neither could be omitted.

The method changed at step three. There was no single constraint the evidence supported, and picking one would have made the deck an argument for one executive against the other, which is a way to lose both. So the deck presented both readings side by side on one slide, each with its supporting quotes, and stated plainly that the evidence supported two different diagnoses and that the difference mattered because it changed what should be measured.

Two objectives sections were drafted, one per reading, and the approach slide showed the two-week piece of work that would distinguish them: a margin analysis across twelve contracts set against the maintenance overrun data. The value slide was left unsized, with the reason given. The next step was not approval but a forty-five minute session with both directors and that analysis in front of them, with a proposed date and the person who would book it named.

What changed, and why: with a single aligned buyer the deck's job is to prove you heard them and propose a shape. With two buyers who disagree, proposing a shape means taking a side before the client has, and the deck's job becomes making the disagreement visible and cheap to settle. The engagement that eventually signed was scoped from that session, not from the deck.

## Output

A deck of fourteen to eighteen slides following the table above, delivered as a file the client can forward without explanation, plus two working artefacts that stay with the firm: the evidence ledger and the open questions list.

The evidence ledger, which is never sent to the client:

| Ledger # | Quote or close paraphrase | Speaker and role | Session and date | What it evidences | Used on slide |
| --- | --- | --- | --- | --- | --- |

The objectives slide, in the deck:

| Objective | Measured by | Baseline, from discovery | Target | Ledger reference |
| --- | --- | --- | --- | --- |

The phase table, in the deck:

| Phase | Duration | Milestone, as a verifiable event | We provide | You provide, with dates |
| --- | --- | --- | --- | --- |

Next steps, one line: the ask, the owner on each side, the date, and what happens if the date passes.

## Failure modes

**Inference presented as understanding.** The version that loses deals quietly. Recognise it with the colleague's question: which of these sentences did they actually say. Anything without a ledger line goes to the hypothesis section or goes out.

**The methodology deck with a client name on it.** Recognise it when the same fourteen slides were sent to another client last month with three words changed, or when the client's situation first appears after slide six. Rebuild from the ledger.

**Objectives that cannot be measured.** Recognise them by the verbs: improve, enhance, optimise, strengthen. Each one needs a number, a baseline and an observation method, and where the baseline is unknown, establishing it is the first deliverable.

**Deliverables written as activities.** Workshops, sessions, support, guidance. Recognise them because nobody could point at one in month three. Convert each into a named artefact with a format and an owner.

**A phase plan with no client dependencies.** Recognise it because every box belongs to the firm. It will slip, and the slip will be a surprise, and the surprise will be attributed to you. Name the client inputs with dates.

**Value sized with a borrowed benchmark.** Recognise it by a percentage with no source, or a figure round enough to have come from a brochure. Use the client's numbers or describe the mechanism and name the figure you need.

**Three next steps.** Recognise it in a final slide with a bulleted list. Three asks produce none. One ask, one owner, one date.

**Proposing the full engagement to a room without budget authority.** Recognise it by asking who present can approve the number on the slide. Where the answer is nobody, the ask becomes the meeting that includes them, or a first phase small enough to approve today.

## Edge cases

**Discovery was one thirty-minute call.** Say so on the understanding slide, name the ledger's thinness as an open question, and propose a short diagnostic rather than a full engagement. A confident deck built on one call is a guess with formatting.

**The client's stated problem is not the real one.** Frequent, and it is handled by sequence rather than by contradiction. Play back their statement of the problem first, accurately and without editorialising, then introduce what the evidence also showed, then state the hypothesis. Correcting a client on slide two, before they have seen that you listened, does not work even when you are right.

**The client asks for the price in the meeting.** Give a range with its basis, or the first phase price alone, and say what would move it. Refusing after they asked reads as evasion; a precise number invented on the spot anchors everything after it.

**The deck will be read without you.** Assume it will. Every slide title carries its message, the understanding section works without narration, and the next step names its owner. Where a slide only makes sense when spoken, either rewrite it or move it to an appendix.

**Discovery surfaced something they did not ask about.** Two teams running incompatible processes, say. Put it in the understanding section, attributed to what was observed, and recommend nothing about it. Naming an unasked-for finding accurately is the strongest credibility available here; recommending work on it in the same breath makes it look manufactured.

**A procurement template is mandatory.** Fill it, and keep the ordering rule inside it: their situation before your firm, even where the template opens with a company overview. Where it forbids that, put a one-page understanding summary in the covering note.

## Quality bar

- Every point in the understanding section traces to a ledger line with a named speaker, and the section uses the client's own vocabulary.
- The hypothesis is stated as a hypothesis, in one sentence, and is separated from what the client said.
- Every objective has a number, a baseline and an observation method, and points at a ledger line.
- Every deliverable is an artefact with a format and an owner, not an activity.
- Every milestone is a verifiable event, and every client-side input that could move the end date is named with a date.
- Every figure in the value section is the client's own, with the arithmetic visible, or the section says what it cannot size and what it needs.
- Nothing about the firm appears before the understanding section.
- Exactly one next step, with an owner on each side and a date.

## Adapting this to your context

This assumes a services firm of ten to a hundred people selling a scoped engagement to a named client after discovery, to a buyer with a rough budget and no agreed approach. The evidence discipline holds everywhere; the shape of the deck does not.

- **The deck itself.** Fourteen to eighteen slides suits a client who will forward it. A public sector or grant buyer often cannot accept a deck: the output is a structured response against published criteria, and the understanding section becomes their specification played back with paragraph numbers cited.
- **Three to five measurable objectives.** These assume the client already measures something. Where they do not, the first objective is to establish the measurement, and the target is set at the next review rather than invented now.
- **The paid diagnostic first.** Offering a small version depends on being able to sell one. Where everything must be tendered whole, that option disappears and the deck carries the full shape with stage gates instead.
- **Sizing the value.** With no client figures, describe the mechanism, name the one figure that would size it, and ask for it. That usually produces the figure.

- **What not to change.** Nothing about the firm appears before the understanding section, and every point in it traces to the evidence ledger.

## Related skills

`customer-interview-synthesis` builds the evidence base where discovery ran to many interviews rather than a few, and its coded corpus feeds this ledger directly. `ideation-deck` in the chief-of-staff track is the sibling for a concept with no named buyer: it argues an idea into existence, where this plays a named client's situation back. `proposal-writer` takes this deck's agreed shape and produces the priced document that asks for a yes. `pricing-and-resourcing-model` sets the number that goes into it. `sow-and-scope` in the chief-of-staff track draws the contractual boundary after approval, with acceptance criteria and exclusions this deck deliberately leaves out. `tailored-client-deck` rebuilds an existing standard deck around one account, which is a different job from building one from discovery. `content-quality-gate` runs over the narrative slides before the deck is sent.