# The two documents, and evidence discipline

## Why two

A findings pack that contains everything is unreadable before a meeting. A brief that contains only conclusions is unusable six weeks later when someone asks *"where did that come from?"*

| | **BRIEF** | **FINDINGS** |
|---|---|---|
| Read by | Account team, before the next meeting | You, and whoever inherits this |
| Length | As short as possible | As long as necessary |
| Contains | Decisions, risks, posture, what to say | Every quote, number, source |
| Optimised for | Action | Traceability |

Cross-reference them by section number so a claim in the BRIEF can be traced to evidence in FINDINGS.

---

## BRIEF structure

Sections that have earned their place:

1. **Where we are** — status, date, what happens next
2. **The finding that explains everything** — the one insight the engagement turns on
3. **The few numbers that carry the story** — with source and confidence on each
4. 🔴 **The constraint that must shape the demo** — usually compensation coupling
5. 🟢 **The unblocker** — the thing that makes this winnable
6. **Architecture tension** — where internal factions disagree; explicitly *do not take a side*
7. **Personas**
8. **The customer's own words as demo spec** — their phrasing, not your translation
9. **Demo environment status**
10. **Proposed narrative**
11. **Workstreams**
12. **Immediate actions**
13. **Lines worth using** — phrases that landed, verbatim
14+. **Append, don't rewrite** — a new numbered section per new source

### Append, don't rewrite

When a new source arrives (internal debrief, official transcript), **add a section** rather than editing earlier ones. This preserves what you believed and when, which matters when a later source contradicts an earlier one.

Then add an explicit **corrections** subsection listing what changed. Being visibly willing to correct yourself is what makes the rest of the document trustworthy.

---

## Evidence discipline

### Confidence marking

| Mark | Meaning | External use |
|---|---|---|
| ✅ Verbatim | Quoted in an official transcript | Yes |
| 🟡 Caption / recollection | Heard but unconfirmed | **No** |
| 🔴 Inferred | Your reasoning or arithmetic | Only if labelled as ours |

### The trap

The most quotable number in your findings is often the least verified. A striking figure spreads fast — into a deck, into an email to the customer — and by then nobody remembers it came from a caption.

Concrete example of the failure mode: a large backlog figure captured from live captions turned out to be **absent from the official transcript entirely**, and the customer had actually described a *quality* problem rather than a *volume* one. The recommended fix was the same either way — but the framing, and the credibility of presenting it, was not.

**Rule:** before any number goes in front of a customer, confirm it appears in a source you can cite.

### Arithmetic you perform is yours, not theirs

If you divide one customer number by another and get something alarming, that ratio is **your inference**. Two figures stated an hour apart in the same meeting were never compared by the customer. Present it as a question, not a finding:

> *"Is that 400 accounts in total, or per rep? We're trying to understand the ratio against the open lead volume."*

Let them do the multiplication. If you're right, it lands far harder coming from them.

### Sanity-check figures against each other

Numbers that survive individually can be jointly impossible. If the customer says ~400 accounts and separately says a maximum of 60 accounts per portfolio, that implies about seven portfolios — implausible for a large retailer. **The contradiction is the finding.** Ask the question.

---

## Recording disagreement

When internal factions disagree in front of you — business wanting one thing, IT another — record it **verbatim and attributed**, and resist resolving it.

Capture:
- who said what, in their own words
- what each is optimising for
- what they each already have in flight
- where the genuine bridge is

Then write the team position **before** anyone says it live. A defensible stance ("we integrate, we don't replicate") written down in advance beats an improvised one under pressure.

---

## Vocabulary capture

Record the customer's terms and use them in every artefact:

| They say | Standard product calls it |
|---|---|
| *(their word)* | Opportunity |
| *(their word)* | Lead |
| *(their word)* | Job site |
| *(their word)* | Book of business |

Renamed entities tell you what the business actually manages. Using their vocabulary in a demo signals you listened; using yours signals a template.

---

## Quotes worth keeping

Keep a running list of lines that captured something precisely. They are the most reusable output of a discovery session — for the readout, the persona doc, and the demo narrative.

Prefer quotes that describe **behaviour** over ones that describe **opinion**:

- weak: *"the UI is not great"*
- strong: *"they use it because they have to, not because they want to"*
- strongest: *"we believe we have tried, but nothing had worked"*

The last kind tells you the emotional state of the account, which determines how you should show up.