B2B creative brief template for LinkedIn and Meta ads, with a worked example

What must a B2B ad brief decide before anyone designs?

A business-to-business (B2B) ad brief must decide six things: what the ad is for, which one person it must reach and what that person risks, the one message, the angles that carry it, the proof you may show with the claims you may not make, and the number that tells you to stop. A designer who gets these six can work alone. A designer who gets a product description will guess.

The risk comes early because the buyer carries it. LinkedIn's Buyability study included a survey of 750 senior B2B buyers. The Drum reports that the ability to defend the decision if things go wrong was the most important of five emotional dimensions. This is LinkedIn's own research, and we read the coverage, not the study. Gartner reports buying groups of five to 16 people across up to four functions, each with different priorities (survey of 632 buyers, August to September 2024).

So one person sees your ad, and that person must repeat it to the others. SaaS Hero says it in one line: the job of a B2B ad is to "arm a champion with a claim and proof point they can repeat". The template on this page builds that sentence into field 3.

We opened public brief templates and frameworks in October 2026 to see what they ask for. Most come from ad tool blogs and ecommerce teams. They ask for a goal, a target audience, a message, brand guidelines, platform specs, budget and timeline. Their audience and message examples are mostly consumer products.

LaunchCodex's performance creative brief (updated 3 June 2026) goes further. It asks for a psychographic trigger (the fear, desire or frustration the ad addresses), a creative angle, a hook direction, offer and call-to-action copy, platform specs, and a testing field with a kill metric. Its example stops an ad on a three-second view rate under 25 percent after 2,000 impressions, or on a cost per result 25 percent above target.

What these briefs miss for B2B is the same in nearly all of them. They do not ask what the buyer risks by saying yes. Most do not ask for claim limits or a stop rule. Some ask for the objection that stops a purchase, and some ask for words to avoid. Budget, timeline and sign-off appear in several of them. This template leaves those to your project tool. Our field 12 covers the stop rule.

What does the template look like?

Twelve fields, two pages at most. Copy the block below into your team doc and replace every [bracket]. The words in round brackets are instructions: delete them when the field is filled. A field you cannot fill is a decision nobody has made yet. Fill it before you brief anyone.

B2B CREATIVE BRIEF
Name: [short working title]
Channel: [LinkedIn / Meta / both]
Owner: [name]   Approver: [name]   Due: [date]

1. JOB OF THE AD (pick one)
Job: [introduce / be remembered / convert]
After the ad, the reader [knows / remembers / does]: [one sentence]

2. OBJECTIVE AND GOAL METRIC
Objective: [the objective set in the ad platform]
Goal metric: [the one number this objective buys, for example cost per lead]

3. READER AND RISK (one role at one kind of company)
Reader: [role, kind of company]
Risk: [what this person risks by saying yes: blame, wasted budget, a failed rollout]
They can say to colleagues afterwards: ["one sentence"]

4. AUDIENCE INSIGHT (one true thing about their week, in their words)
Insight: [one or two sentences]
Source: [sales calls / reviews / support messages]

5. KEY MESSAGE (if it needs "and", it is two ads)
This ad is about [one sentence].

6. ANGLES (2 to 4; each differs in mechanism, not in wording)
Already tried, do not repeat: [angles and hooks, or "none"]
Angle A: [name]
  Reason: [why the reader cares]
  Lever: [the mechanism: loss, proof, contrast, effort, status]
  Hook 1: [first line or first picture]
  Hook 2: [a different first line or picture]
  Hook 3: [a third one]
Angle B: [name]
  Reason: [why the reader cares]
  Lever: [the mechanism]
  Hook 1: [first line or first picture]
  Hook 2: [a different one]
  Hook 3: [a third one]
Angle C (optional) and angle D (optional): repeat the lines of angle B. Delete what you do not use.

7. PROOF WE CAN SHOW (only what you can source)
Slot 1 (angle [letter]): [named customer, role, result in their words, source and date]
Slot 2 (angle [letter]): [named customer, role, result in their words, source and date]
No proof available for angle: [letters, or "none"]

8. OFFER AND ASK
Offer: [what the reader gets]
Ask: [the next step]
Size of the ask: [small: look or read / medium: leave contact details / large: try, buy or talk]
Button text: [words]
Page or form: [web address]. It repeats this promise: [promise]

9. OBJECTION
The reader says: ["the doubt, in their words"]
Our answer: [one line, with proof from field 7]

10. THE AD MUST NOT CLAIM
- [unsourced numbers, legal promises, competitor claims, product behaviour the page does not show]
Claims to confirm before design: [hooks that name product behaviour]

11. CHANNEL NOTES
How the reader meets the ad: [LinkedIn: a professional at work / Meta: a person at leisure]
Format and size: [formats, ratios, number of concepts]
First line: [what the first line must do, and how many characters show]
Picture: [what it shows; with the words covered, it must still argue]

12. HOW WE WILL READ THE RESULT
Hypothesis: Because [observation, or "no data yet, this is a bet"], we believe [cause]. We test [concept] against [control], judged on [goal metric].
Delivery: [how the platform splits spend between the concepts, for example the LinkedIn ad rotation setting]
Kill rule: Judge once, when each concept has spent [budget]. Stop a concept if its [goal metric] is worse than [control] by [amount] at that point. Set these numbers before launch. Do not judge earlier.
If it wins, next: [the next test]
If it loses, we suspect: [the part shared by all concepts: audience, offer, form or page]

How do you fill each field?

Write the shortest line a designer can act on. A line is done when the designer can start drawing without asking you a question. The examples below use Billnod, an invented tool that collects invoice approvals in one place, so finance teams stop chasing managers by email.

Job of the ad

Pick one job: introduce, be remembered or convert. An ad that introduces tells a stranger who it is for and what the product does. An ad that is remembered may leave the product unexplained. Our reading: then the brand must be unmissable, or the reader remembers the ad and forgets who ran it. An ad that converts assumes the reader knows you and makes the next step clear. LinkedIn's B2B Institute says 95 percent of potential buyers are not ready to buy today. That is a vendor's research arm, and one review notes the percentages are not meant literally. Our reading: an ad for a cold audience should introduce or be remembered, and a convert ad belongs with people who already know you.

Objective and goal metric

Copy the objective from the ad platform and name the one number it buys. A lead objective buys leads, so the goal metric is cost per lead. A video views objective buys views, so it is cost per video view. Write one metric. Then nobody changes the judge after the results arrive.

Reader and risk

Name one role at one kind of company. Then name the risk: what happens to this person if the choice goes wrong. Blame, wasted budget, a failed rollout and a hard talk with finance or IT are typical. Last, write the one sentence the reader can say to colleagues afterwards. That sentence is what the ad must hand over.

Gartner found that content aimed at the whole buying group raised consensus, and content aimed at one person's view hurt it. Gartner surveyed buyers about the content sellers send to buying groups, not about ads, so treat it as a hint. Our reading: name one reader for the ad to find, and write the message so that this reader can repeat it to the rest of the group.

Audience insight

Write one true thing about the reader's week that your product touches. Take it from sales call notes, reviews and support messages, and keep the customer's words. A review is a source for wording. Run it as a testimonial only with the reviewer's consent.

Key message

Write one sentence that starts "This ad is about". If the sentence needs "and", the brief holds two ads. Say it the way the reader would say it to a colleague.

Angles and hooks

Write two to four angles. An angle is the reason this reader would act. Two angles differ when their mechanism differs: a different fear, a different gain or a different scene. When the words change and the reason stays, you have one angle. This test is ours. B2B ad angles explains how to find them.

Give each angle a lever, which is the mechanism that moves the reader, such as fear of loss or proof from a peer (the levers are in this guide). Then write three hooks per angle. A hook is the first line or first picture the reader meets (hook types). Three hooks give the designer a choice without a flood. This number is a choice we made, not a researched figure. The same guide shows how to write ten hooks for one angle, if you want to cut down to your best three.

A hook passes our test when the reader would paste it into the team chat with the words "this is us".

Proof

List only proof you can source: a named customer, their role, a result in their words, where it is published and the date. Leave out everything else. Proof points in B2B ads sorts what counts. An empty slot is allowed. It means the angle makes no proof claim, and field 7 says so.

Offer and ask

The offer is what the reader gets, in their terms: a result, a faster first win or less effort. The ask is the next step. Use three sizes: small (look or read), medium (leave contact details for content or an event) and large (try, buy or talk). Match the size to the job. Our reading: a reader who is not buying today meets a small or medium ask. Name the button text, so the designer does not invent it.

Objection

Write the doubt the reader voices when they hesitate. Take it from sales calls. Then write the answer in one line, with proof from field 7. Many brief templates ask for the objection. We keep the field and tie the answer to proof.

The ad must not claim

List the claims the ad cannot make: unsourced numbers, legal promises, competitor claims and product behaviour the product page does not show. The landing page sets the ceiling. If the ad promises one capability, integration or file format, the page must show it.

Channel notes

Say how the reader meets the ad. Our reading: on LinkedIn the reader is a professional at work, and on Meta the ad interrupts a person at leisure. The section on LinkedIn and Meta below shows what that changes. Add the format, the first-line length and the picture direction. The limits are on LinkedIn's help page for single image ads (intro text up to 150 characters and headline up to 70 characters to avoid cuts; vertical 4:5 images serve on mobile only) and on Meta's ads guide for Facebook Feed images (50 to 150 characters of primary text, a 27-character headline, ratio 4:5).

How we will read the result

Write the hypothesis and the stop rule before launch. Use the goal metric and a stated cause: "Because [observation], we believe [cause]. We test [concept] against [control], judged on [goal metric]." If you have no observation, write "no data yet, this is a bet". The learning sentence takes the same form after the test.

For the stop rule, fix the spend per concept in advance, and look once. Evan Miller's advice on split tests (A/B tests) is to "decide on a sample size in advance and wait until the experiment is over", because repeated peeking raises the rate of false positives. Our reading: in an ad test, the part you can fix is the spend. A rule that waits for a set number of leads per concept never ends for the concept that gets the fewest.

Write the delivery setting too, because the platform decides how much each concept gets. LinkedIn's recommended ad rotation setting serves creatives evenly at first, then delivers more impressions to the ones that perform best, to optimize the ad set's click-through rate. Its other option, rotate ads evenly, enters each creative into the auction evenly, but the impressions each ad gets still depend on ad performance and auction competition. LinkedIn sets rotate ads evenly by default for brand awareness ad sets, and you cannot change it there. So expect uneven exposure, and judge each concept at the same spend, not at the same date.

Then set the numbers, budget first. Two concepts with about 25 leads each can only be told apart when one costs about twice as much per lead, so a small budget catches only a large gap (when to pause an ad has the table). Set the amount in the kill rule to a gap that your budget can show.

What does a finished brief look like?

Here is the template filled in for one channel. Billnod is an invented product. Every proof field is a placeholder, and the reader insight is constructed for this example. In your brief, the insight comes from your sales calls and reviews.

EXAMPLE ONLY. Billnod is an invented product: invoice approvals in one place, so finance teams stop chasing managers by email. Every proof field is a placeholder.

B2B CREATIVE BRIEF
Name: Who approved this invoice?
Channel: LinkedIn
Owner: paid social lead   Approver: head of marketing   Due: [date]

1. JOB OF THE AD
Job: introduce
After the ad, the reader knows: every invoice approval can carry an approver's name and a date, without an email chase.

2. OBJECTIVE AND GOAL METRIC
Objective: lead generation with a LinkedIn lead form
Goal metric: cost per lead

3. READER AND RISK
Reader: head of accounts payable at a mid-sized company
Risk: pays an invoice nobody can show was approved, then has no answer when an auditor asks. Second risk: buys a tool that managers ignore.
They can say to the chief financial officer afterwards: "Every approval has a name and a date."

4. AUDIENCE INSIGHT
Insight: The reader chases approvals by email every week and keeps a private spreadsheet. When someone asks who approved an invoice, the reader searches an inbox.
Source: constructed for this example

5. KEY MESSAGE
This ad is about: every invoice gets an approver's name and a date, with no email chase.

6. ANGLES
Already tried, do not repeat: none
Angle A: The audit question
  Reason: the reader fears having no answer when an auditor asks
  Lever: loss aversion, made concrete by the auditor's question
  Hook 1: Who approved this invoice? Nobody remembers.
  Hook 2: The auditor asked. The inbox did not answer.
  Hook 3: Every approval, with a name and a date.
Angle B: The chase
  Reason: the weekly email chase annoys everyone in finance
  Lever: enemy naming. The enemy is the email thread, never a colleague.
  Hook 1: Reminder number four. Same invoice.
  Hook 2: The invoice is fine. The inbox is the problem.
  Hook 3: Managers approve where they already work.
Angle C: The easy start
  Reason: the reader fears a rollout that managers ignore
  Lever: risk reversal. The first step is small and can be undone.
  Hook 1: Start with one team. Keep your finance system.
  Hook 2: Managers approve without a new login.
  Hook 3: Run it next to your current process first.

7. PROOF WE CAN SHOW
Slot 1 (angle A): [named customer, role, result in their words, source and date]
Slot 2 (angle C): [named customer, role, result in their words, source and date]
No proof available for angle: B. It makes no proof claim.

8. OFFER AND ASK
Offer: the approval audit checklist, the questions an auditor asks about invoice approvals
Ask: leave a work email in the lead form
Size of the ask: medium
Button text: Download
Page or form: the lead form. It repeats this promise: the checklist arrives by email at once.

9. OBJECTION
The reader says: "Our finance system already has approvals."
Our answer: it runs next to the finance system, so nothing is replaced. Proof: slot 2, once filled. (Claim to confirm.)

10. THE AD MUST NOT CLAIM
- "Audit-proof", "compliant" or any other legal promise
- Any saving in hours or money, unless a proof slot holds a sourced figure
- That it works with every finance system. Name the systems it supports, with the setup page as source.
- Anything about a competitor
Claims to confirm before design: hooks A3, B3, C1, C2 and C3, and the objection answer

11. CHANNEL NOTES
How the reader meets the ad: a professional at work. Name the role in the first line.
Format and size: single image, 1:1 and 4:5 (4:5 serves on mobile only). One concept per angle.
First line: carries the hook. LinkedIn recommends up to 150 characters of intro text and 70 for the headline.
Picture: a plain invoice list with an approver name and a date on every row. No stock photo of an office.

12. HOW WE WILL READ THE RESULT
Hypothesis: Because we have no account data yet (this is a bet), we believe the audit fear pulls more leads than the chase. We test A, B and C as three concepts with the same audience and offer, judged on cost per lead.
Delivery: ad rotation set to rotate ads evenly, so no concept is starved early. Exposure will still differ.
Kill rule: Judge once, when each concept has spent [budget]. Stop a concept if its cost per lead is worse than the best concept's by [amount] at that point. Set these numbers before launch. Do not judge earlier.
If A wins, next: test the audit question with a proof slot filled.
If all three lose, we suspect: the offer or the form, because all three share them.

Run these five checks before the brief goes out.

1. Cover everything except fields 3 and 5. Can a stranger say who the ad is for and what it says?
2. Read the three angles. Do they differ in mechanism, or is it one idea in three phrasings?
3. Find every claim in the hooks. Does each one sit in field 7 or on the product page?
4. Read each hook as the reader. Would they paste it into the team chat?
5. Look at field 12. Are the numbers of the stop rule written down?

How do you brief the same concept for LinkedIn and for Meta?

Read every angle twice: once as a professional at work, and once as a person at leisure who gets interrupted. Our reading: a work fear fits the first reader. A shared annoyance or a small first step travels better to the second, in situation words and with no job title. Fields that the table does not list carry over word for word.

A pooled cost per lead is an average of two different things. The channel with more leads dominates it, so it describes neither channel. LinkedIn ads against Facebook ads for B2B goes deeper on the channel rewrite.

How do you brief from a competitor ad or one of your own winners?

Add a source block to the brief. Say what the source ad does, what you keep, what you change and what you do not copy. Our rule: keep the core move of the source ad, and keep its style of execution. A meme source gets a meme brief. A bare screenshot stays bare. Your data can sharpen the hooks, but it does not overturn the direction you picked. When the two disagree, note it in field 12 and brief your direction anyway.

None of the pages we read splits a reference ad into what to keep, what to change and what not to copy. LaunchCodex asks for two to four reference ads, each as a link or file name, to show tone, format or style. Other templates ask for examples to emulate or avoid, or for top performers. The split is ours.

SOURCE AD (add under field 1 of the brief)
Source: [ad library link or ad ID, or the name of your own ad]
Whose ad: [competitor / our own winner / our own loser]
What it does: [job, reader, angle, hook, proof, ask: one line each]
Why it might work: [our reading, one sentence]
Keep: [its core move, and its style of execution]
Change: [the product, the proof, the claims, the offer]
Do not copy: [their words, their claims, their logo, their layout]
The one variable under test: [what differs from your own control ad]

Three rules apply to the block.

• A competitor ad shows what that company runs. It does not show what works. LinkedIn's Ad Library help page lists the preview, format, advertiser and payer, and for ads targeted at the European Union it adds impressions, targeting and run dates. It lists no spend and no clicks. Meta's page describes a transparency hub for active ads and names no results. So "why it might work" is a reading, and you label it as one. Their claims are theirs. Do not reuse them as yours.
• For your own winner, change one thing. A win on one ask shows that the mechanism worked for that ask. It shows nothing yet about a new ask. If you change the ask, keep the angle and the hook.
• For your own loser, separate the layout from the idea. Name the layout you drop and say what it was for. Then find another execution of the same argument.

The LinkedIn Ad Library guide shows how to code competitor ads in rows before you pick a source.

How do you brief an AI model with the same template?

An artificial intelligence (AI) model needs the template, a context pack and rules that stop it inventing. It then writes a first brief in one run. You then decide the risk, the claims and the stop rule. Anthropic's prompting guide says to think of Claude as "a brilliant but new employee who lacks context on your norms and workflows". Our reading: the same holds for ChatGPT. The context pack is the missing context.

Paste what a strategist who is new to the account would ask for on day one:

• the product, the company and the brand rules, with the words to avoid
• your best and worst ads on the goal metric, with their copy
• the text of the landing page or the lead form
• customer words from sales calls, reviews and support messages
• the proof you may use, each item with its source and date, and the claims you must not make
• earlier briefs or drafts your team approved or turned down, with the reason
• the angles you already ran, and a few competitor ads as copy

Leave out anything you would not hand a freelancer on their first day.

Prompt 1 writes the brief. Fill the brackets in the first line, the Task line and the Context lines. Paste the template from this page into the template block as it is. The model fills its brackets, except the budget and the amount in the kill rule, which you set.

You write creative briefs for a B2B ad team. Write one brief for ads on [LinkedIn / Meta / LinkedIn and Meta].

Task: [one sentence: what the brief is for, who it is for and what the ads should get]

Context
- Product and company: [paste]
- Brand rules (voice, words to avoid): [paste]
- Reader: [role and kind of company]
- Campaign objective and goal metric: [paste]
- Our best and worst ads on that goal metric, with their copy: [paste, with one line on what each did]
- Text of the landing page or lead form: [paste]
- Customer words (sales call notes, reviews, support messages): [paste]
- Competitor ads, copy only: [paste, or "none"]
- Angles we already ran: [list, or "none"]
- Proof we can show, with source and date: [paste, or "none"]
- Claims we must not make: [paste, or "none"]

<template>
[paste the twelve-field template from this page]
</template>

Fill every field of the template, in order, and keep each one short enough for a designer to act on. Work to these rules.
1. Stay inside the context. A fact you do not have is "unknown", with a note on what I should supply. If a context line still shows [brackets], list it first.
2. Every customer, quote, number and proof point comes from the context, with its source. Where an angle has no proof, say so in field 7.
3. Give each angle its own reason for the reader to act. If you can swap the hooks of two angles and nothing changes, merge them. Make at least one angle new against the angles we already ran.
4. In field 11, use only these limits and write "unknown" for other formats: LinkedIn single image ads take intro text up to 150 characters and a headline up to 70. Meta Facebook Feed images take 50 to 150 characters of primary text, a 27-character headline and ratio 4:5.
5. In field 12, write "no data yet, this is a bet" if the context holds no observation. Leave the budget and the amount in the kill rule for me.
6. Write short, plain sentences, with no dashes, no exclamation marks and no stock phrases.

After the brief, add two lists. List one: the weakest angle and why. List two: every claim in the hooks that needs a source before design starts.

Read the output with four steps. First, check that the risk in field 3 is true for your reader. Second, delete every claim that has no source in field 7 or on the product page. Third, set the numbers of the stop rule yourself. Fourth, run prompt 2.

Prompt 2 reviews a brief, from a designer's view. It works on briefs from people and from models.

You are a designer who has never heard of our product. Read the creative brief below. Do not rewrite it. Answer each check in one line, and quote the words that support your answer.
1. Reader: which single role is the ad for? Does the brief say what that person risks?
2. Message: say the key message back in your own words. If you cannot say it in one sentence, say so.
3. Angles: does each angle give the reader a different reason to act, or do some repeat one reason in new words?
4. Hooks: which hooks could a competitor run unchanged?
5. Proof: which claims in the hooks have no proof in the proof field?
6. Limits: does any hook or the message break the "must not claim" field?
7. Ask: does the size of the ask fit the job of the ad?
8. Result: can you tell from the brief which number would make us stop the ad, and when?
9. First question: what is the first question you would ask the brief owner before you start?
Then give a verdict: ready to design, or not ready, with the single field to fix first.

BRIEF
[paste the brief]

Nimblet builds this pack from your connected ad accounts, your brand, your team documents and your tracked competitors. In the app, it then writes the brief and drafts the ads. Over the Model Context Protocol (MCP), Claude or ChatGPT reads the same pack, and either writes the brief itself or asks Nimblet to write it. Write B2B ads with Claude or ChatGPT covers the full batch workflow.