BRIGHTWATER ANALYTICS CAMPAIGN BRIEF (2026-01-14, v3)

EVERYTHING IN THIS FILE IS INVENTED. Brightwater Analytics does not exist, neither do the
people in the approved examples, and the 1.4m figure is made up. It is here to show the
target structure and tone for a real brief, not to be copied as content. Replace every word.

Covers the "Stockout audit - retail ops" campaign. Replaces v2 "inventory intelligence"
brief dated 2025-11-02, which led with the product and got 2 replies in 400 sends.

-----

ANGLE: Audit-led. Open on one specific, publicly visible thing about how they run stock,
then offer a read of their own numbers rather than a demo. Deliberately drops the previous
product-led framing. Acknowledged risk: "free audit" is a familiar shape to this ICP and
several competitors use it. The hedge is that the audit is genuinely 20 minutes of work on
data they already have, and the message says exactly what comes back.

ANGLE TYPE: Research-anchored.
OFFER TYPE: Free Audit - a 20 minute read of one month of their stockout data, findings sent back.

CONNECTION REQUEST: Blank. Higher acceptance rate, and there is nothing worth saying in 200
characters that is not better said after they accept.

PERFORMANCE GATES: 50 accepted connections before judging anything. Target 5 percent reply rate.

-----

FORMATTING RULES (HARD)

- No em dashes. No en dashes. Hyphens, commas and full stops only.
- Blank line between each section of M1. Three visually distinct blocks, never a wall of text.
- First name in every message. Company name once in M1, then "your team".
- No emojis.
- Digits, not words. "3 sites", not "three sites".

-----

M1 STRUCTURE - LOCKED. Three blocks, blank lines between. Variable text is {first_name},
{observation}, {category}.

{first_name} - {observation}.

We do a free stockout read for retail ops teams. One month of your {category} data, back
inside a week, showing where the lost sales actually sat. Worth a look?

LOCKED: The offer sentence is locked. "One month of your {category} data, back inside a week,
showing where the lost sales actually sat" must appear verbatim. Do not paraphrase the
timeframe or the deliverable.

M1 WORD COUNT: 50 to 85 words.

-----

M2 STRUCTURE - LOCKED. Adds the number M1 did not have. Variable text is {first_name},
{category}.

{first_name} - the last three of these we ran found between 4 and 9 percent of {category}
revenue sitting in avoidable stockouts. One of them was 1.4m a year.

Happy to run yours the same way. No call needed, we just need the export.

LOCKED: The second sentence of the first block is locked. "One of them was 1.4m a year"
appears verbatim or not at all. Never round it, never inflate it.

M2 WORD COUNT: 30 to 75 words.

-----

M3 STRUCTURE - LOCKED. Two sentences. Variable text is {first_name}.

{first_name} - we have run this read for 40 retail ops teams in the last year and it has never
taken them more than the export. Want me to send the one page brief on what we need?

LOCKED: Entire message. M3 WORD COUNT: 20 to 40 words.

-----

SLOT GUIDE

### {first_name}

The lead's first name, one word, no title. If the profile shows only an initial, or the name
is ambiguous enough that you might get it wrong, skip the lead rather than guessing.

### {observation}

One specific, publicly discoverable thing about THIS person or company. 10 to 25 words.

**Fill rules:** One observation, never two. Must include the company name. Must pass the swap
test: if you swap in another lead from this list and it still reads fine, it is not an
observation, rewrite it. Open naturally, the way you would type it. Never fabricate. Never
restate their headline back at them. No verdict on it afterwards, no "that is impressive",
no "smart move".

**Examples:**

a) "saw Tilbury opened the two Midlands sites in October and kept the same central
   replenishment setup" - datable, specific, and it implies the problem without asserting it
b) "noticed Nordvik moved to weekly ordering across the whole range last quarter" - a real
   operational change, and only true of them
c) "saw you have gone from 40 to 70 lines in the pharmacy range at Cavendish since spring" -
   a number, from their own announcements, that makes the offer obviously relevant

**Never:** "I see you work in retail operations". "Loved your recent post". Anything true of
every company on the list. Anything you inferred rather than read.

### {category}

The specific product category or range the lead is responsible for. 2 to 5 words. Must be
their words, taken from their site or their profile.

**Fill rules:** Concrete and narrow. If you cannot name a specific category, the lead is not
researched enough to message. Skip them.

**Examples:**

a) "chilled and fresh" - how a grocer describes it themselves
b) "pharmacy and healthcare lines" - specific, and it is the range they actually own
c) "seasonal homeware" - narrow enough that it could not be anyone else on the list

**Never:** "your products", "your inventory", "your SKUs", "your categories". Placeholder text
of any kind reaching a real person is a hard fail.

-----

NICHE NOTES - TONAL GUARDRAILS

- This ICP is pitched by inventory software vendors constantly and has learned to ignore
  anything that opens with a capability. Lead with their situation, never with the product.
- Write it as the sender would actually type it. Conversational, direct, a bit clipped.
- Never assert what they are struggling with. Use "I would imagine" if you go near it at all.
- If the observation reads like a database row, rewrite it.

LANGUAGE TO USE: "stockout read", "the export", "lost sales", "worth a look"

LANGUAGE TO AVOID: "solutions", "leverage", "visibility", "best in class", "supply chain
excellence", "hop on a quick call", "circle back", "touch base". Never let {first_name},
{category} or any placeholder reach a prospect.

-----

SOCIAL PROOF RULES

- The only approved claim for this angle: "between 4 and 9 percent of category revenue sitting
  in avoidable stockouts", and "one of them was 1.4m a year".
- Verbatim, or not at all. Do not round the 1.4m up.
- Client names stay out. If they ask which one, the answer is the vertical and the size, not
  the name, unless permission exists in writing.
- One credential per sequence. Never stack proof points.

-----

PRE-SCRIPT EXCLUSIONS

- No inventory or supply chain software vendors. Competitor territory.
- No companies under 50 or over 2,000 employees.
- No pure e-commerce with a single warehouse. The angle does not apply and they know it.
- No lead where a specific {category} cannot be named. A generic fill kills the message.

-----

HANDLING REPLIES

IF THEY ASK WHAT YOU NEED: One screen, not a deck. One month of stockout or lost-sales export
at SKU and site level, in any format. Findings back inside a week. One qualifying question at
the end: "who owns replenishment for {category} at the moment?"

IF THEY ASK WHAT IT COSTS: The read is free and there is no obligation attached. Say what
happens after only if they ask, and say it plainly.

IF THEY ASK WHY IT IS FREE: Because it is 20 minutes of work and it is the only honest way to
show whether there is anything worth talking about. Do not dress this up.

IF THEY SAY NO: Grace out. "No problem, door is open if that changes." Move them to closed.
Never a second attempt after an explicit no.

-----

QA CHECKS (MECHANICAL)

- M1 word count: FAIL if under 50 or over 85
- M2 word count: FAIL if under 30 or over 75
- M3 word count: FAIL if under 20 or over 40
- Connection request blank: FAIL if populated
- No em dashes: FAIL if found anywhere
- No en dashes: FAIL if found anywhere
- Company name in M1: FAIL if absent
- Company name used more than once after the first mention: FAIL
- First name present in all three: FAIL if any slot is unfilled
- Locked phrasing verbatim: FAIL if the offer sentence or the 1.4m line is altered
- Placeholder leakage: FAIL if a brace, bracket or the word "category" as a literal appears
- Kill-list words: FAIL if "solutions", "leverage", "visibility", "best in class" appear

QA CHECKS (CONTENT)

- SWAP TEST: remove {first_name}, {observation} and {category} from M1. If what is left still
  applies to 5 or more other leads on this list, FLAG the observation for rewrite.
- CATEGORY SPECIFICITY: {category} must be a named range, never a generic noun.
- OBSERVATION SOURCE: every observation must be traceable to something publicly visible. Flag
  anything that reads inferred.
- CROSS-MESSAGE REPETITION: M2 must introduce the 4 to 9 percent figure as new information,
  not restate M1's offer. M3 must not repeat M2's number.
- READ-ALOUD TEST: read M1 out loud. If it sounds like a cue card, rewrite it.

-----

APPROVED EXAMPLES

Dana Whitfield, Head of Retail Operations, Tilbury Retail Group. 340 stores, opened 2 Midlands
sites in October, chilled and fresh is her range.

M1:
"Dana - saw Tilbury opened the two Midlands sites in October and kept the same central
replenishment setup.

We do a free stockout read for retail ops teams. One month of your chilled and fresh data,
back inside a week, showing where the lost sales actually sat. Worth a look?"

M2:
"Dana - the last three of these we ran found between 4 and 9 percent of chilled and fresh
revenue sitting in avoidable stockouts. One of them was 1.4m a year.

Happy to run yours the same way. No call needed, we just need the export."

M3:
"Dana - we have run this read for 40 retail ops teams in the last year and it has never taken
them more than the export. Want me to send the one page brief on what we need?"

-----

Marcus Ilori, Supply Chain Director, Nordvik Home. Moved the whole range to weekly ordering
last quarter, seasonal homeware is the volatile part.

M1:
"Marcus - noticed Nordvik moved to weekly ordering across the whole range last quarter.

We do a free stockout read for retail ops teams. One month of your seasonal homeware data,
back inside a week, showing where the lost sales actually sat. Worth a look?"

M2:
"Marcus - the last three of these we ran found between 4 and 9 percent of seasonal homeware
revenue sitting in avoidable stockouts. One of them was 1.4m a year.

Happy to run yours the same way. No call needed, we just need the export."

M3:
"Marcus - we have run this read for 40 retail ops teams in the last year and it has never
taken them more than the export. Want me to send the one page brief on what we need?"

-----

Ruth Kavanagh, Category Director, Cavendish Pharmacy Group. The tricky one: the range grew
fast, which is publicly visible, but nothing about stockouts is. The observation carries the
relevance without asserting a problem she has not mentioned.

M1:
"Ruth - saw you have gone from 40 to 70 lines in the pharmacy range at Cavendish since spring.

We do a free stockout read for retail ops teams. One month of your pharmacy and healthcare
lines data, back inside a week, showing where the lost sales actually sat. Worth a look?"

M2:
"Ruth - the last three of these we ran found between 4 and 9 percent of pharmacy and
healthcare lines revenue sitting in avoidable stockouts. One of them was 1.4m a year.

Happy to run yours the same way. No call needed, we just need the export."

M3:
"Ruth - we have run this read for 40 retail ops teams in the last year and it has never taken
them more than the export. Want me to send the one page brief on what we need?"

-----

CAMPAIGN STATUS

Status: active
Scripts: needs review, calibration pending on the first 5 leads
Batch size: 50

-----

OPEN FLAGS FOR CALIBRATION

1. CATEGORY FILL QUALITY. {category} is doing more work than {observation} in M2 and M3. If
   the fills come back generic across the first 50, M2 and M3 both go flat and the whole
   sequence rests on M1. Check the first 5 fills specifically for this.

2. THE 1.4m FIGURE. It is a real number from one engagement and it is the largest one. If
   replies come back asking "on what revenue?", the number is doing more harm than good.
   Fallback: drop the absolute and keep the 4 to 9 percent range only.

3. AUDIT FATIGUE. Two competitors on this list run a free-audit offer. If reply sentiment
   skews sceptical rather than curious, the differentiator is not landing and the
   "no call needed" line should move up into M1.

4. THE SWAP TEST ON SMALL SEGMENTS. On a tight list, several leads genuinely share the same
   publicly visible change. The swap test will flag observations that are correct. Judge those
   by hand rather than rewriting them into something vaguer.
