Build your brand-brain
Create a folder on your laptop called brand-brain/ and fill it with your brand's raw material. On Day 1 you'll load it into your CLAUDE.md in the Your D2C AI Command Center session, and every session after that runs on it.
Create the brand-brain/ folder
Anywhere you can find it easily — Desktop, Documents, or a synced Google Drive folder. Inside it, create one more folder called voice-dna/. That's the whole structure:
brand-brain/
positioning.md
catalog.csv
pdp-targets.md
competitors.md
reviews.md
unit-economics.csv
ad-creatives/
voice-dna/
The more complete your inputs, the more functional the systems you walk out with. A half-filled brand-brain still works — it just means the AI teammates you build on Saturday know half your brand.
Must-have files
Eight inputs. Each one feeds a specific session, so nothing here is busywork.
| File | Format | What it feeds |
|---|---|---|
| Brand positioning brief | .md or .txt, 1 paragraph | Voice, audience, your wedge, top 2 competitors. Baked into CLAUDE.md and used by every session. |
| Product catalog | CSV or .md | 10–20 SKUs with name, category, price, key claim/USP. |
| 5 SKUs + 1 landing page URL | .md file with URLs | The Landing Page & PDP Optimizer session on Day 2. |
| Competitor notes | .md, one line each | Market Intelligence. One line on why each competitor matters is enough. |
| Customer reviews | Links, screenshots, copy-paste | Customer Voice Mining. 20+ minimum. Amazon and Flipkart have no native export, so copy-paste is normal. |
| Last 90 days unit economics | CSV or a table in .md | Growth Dashboard & Decision Engine. Monthly revenue by channel, CAC, AOV, contribution margin, RTO%. |
| Recent ad creatives | Screenshots | Ad Copy & Creative Strategy Engine. 3–5 Meta or Google ads you've actually run. |
| Voice DNA samples | Copy-pasted text or screenshots | Content Factory. 5–10 real samples across formats — see below. |
Swipe the table sideways →
Voice DNA — where to pull samples from
Drop these into brand-brain/voice-dna/, one file per sample. This is what stops the Content Factory sounding like everyone else's Content Factory.
- 2–3 product detail pages — full PDP copy
- 3–5 ad copies that worked — headline + body + CTA from Meta or Google
- 2–3 email broadcasts — promo or newsletter
- 2–3 WhatsApp or SMS broadcasts — from your BSP (AiSensy, Wati, Interakt)
- 5–10 Instagram captions or LinkedIn posts — founder-led ones especially
- Packaging copy or inserts
- Your brand voice guide, if you have one
Screenshots are fine. So is a single messy .txt file with everything pasted in. The model reads the words, not the formatting.
Nice to have (skip if it slows you down)
- CRM exports — last 90 days of support tickets (Freshdesk, Zoho Desk, Gorgias)
- WhatsApp broadcast exports
- Existing SOPs — PDFs, Docs, Notion links
- Marketplace listings — Amazon Seller Central CSVs or A+ screenshots
- Brand assets — logo, colour codes, packaging photos
How to approach this
- File names don't matter. Formatting doesn't matter. Raw copy-paste is fine.
- Delegate it. Your CMO, ops lead, growth lead or founder's office can pull most of this in an hour. Send them this page.
- Don't polish. Incomplete beats absent. A rough catalog you actually bring is worth more than a perfect one you don't.
- Not on marketplaces yet? The PDP Optimizer session runs off your D2C catalog, and we'll hand you a sample marketplace schema to work against.
Showing up Saturday with nothing and planning to "use a sample brand". You'll learn the workflow and lose the output — the systems you build won't be yours to run on Monday.
brand-brain/ folder createdbrand-brain/voice-dna/Confirm your stack access
Log into every platform your brand runs on, today — not Saturday morning. The live builds work far better on your real data than on a sample dataset, and a forgotten password at 10 AM costs you a session.
What to confirm
- Shopify admin — or whatever your storefront runs on
- Google Drive — where your
brand-brainfolder will live - Meta Ads Manager — logged in as an admin on the brand account, not a personal one
- Amazon / Flipkart Seller Central — if you sell there
- CRM or support tool — Freshdesk, Zoho, Gorgias, if you use one
If access is blocked
Work laptops often block personal Gmail or third-party logins. Sort it now:
- Ping the co-founder or colleague who holds the credentials today
- Have a personal laptop ready as a backup, and decide Saturday morning which one you're using for the live builds
- Still stuck? Tell us in the WhatsApp group — we'll run your sessions on sample data so you're never sitting idle
Test the actual login, don't assume. "I'm sure I have access" is the single most common Saturday-morning surprise.
Connect your Meta ad account
Day 2 runs on your real Meta data — actual campaigns, ad sets, targeting and creative — in the Ad Copy & Creative Strategy Engine and Growth Dashboard & Decision Engine sessions. Two ways to connect. Pick one.
Meta Ads MCP
Connect once inside claude.ai and your ad data is always live. Needs a paid Claude plan (Pro, Max, Team or Enterprise).
System User token
Generate a read-only token once and keep it. More steps up front, more reliable in the room.
Don't run ads at all? Skip this task and tell us in the WhatsApp group — we'll run your Day 2 sessions on a sample ad account.
Path A — Meta Ads MCP
- Go to claude.ai → Settings → Connectors (claude.ai/customize/connectors)
- Click Add custom connector — Meta Ads isn't in the built-in directory
- Enter Name:
Meta Adsand URL:https://mcp.facebook.com/ads, then click Add - Click Connect on the new connector, log in with your Meta Business Manager account, and grant access to the ad accounts you want
- Open a new claude.ai chat and ask: "list my Meta ad accounts". You should see your accounts pulled live.
Connect this from claude.ai, not from the Claude Code CLI — the CLI currently fails the OAuth handshake for this connector.
Path B — System User token
You'll need admin access to a Meta Business Manager, and at least one ad account that has actually run ads (an empty account returns empty reports).
Step 1 — Create the System User and generate a token
- Go to business.facebook.com → Business Settings → Users → System Users
- Click Add and create a System User named
bootcamp-sync. Employee access is enough. - Click the new System User → Add Assets → Ad Accounts → select your account → grant "View performance"
- Back on the System User page, click Generate New Token. Pick any app, set expiration to Never, and tick
ads_readonly. - Copy the token immediately — Meta shows it exactly once. It starts with
EAA…and runs about 200 characters.
Step 2 — Get your ad account ID
- Open adsmanager.facebook.com
- Look at the URL bar for
…?act=1234567890&… - Copy just the number — no
act_prefix, the script adds it
Generating the token before assigning the ad account in step 3 gives you a token that can see no ad accounts. If your sync comes back empty, this is almost always why. Assign the asset first, then generate.
Save both values somewhere safe
Put these in your password manager or a local note you'll have open on Saturday:
META_ACCESS_TOKEN: EAA... your ~200-character token
META_AD_ACCOUNT_ID: 1234567890 (just the number, no act_ prefix)
This token is a password. Never paste it into WhatsApp, Slack, a screenshot or a shared doc — including the bootcamp group. If it ever leaks, regenerate it in Business Settings → System Users and the old one dies instantly.
ads_read token generated (starts with EAA…) and saved securelyact_ prefixCommon questions
I can't finish all of this before Saturday. What matters most?
Task 1, and inside it: the positioning brief, the catalog, and the reviews. Those three feed the most sessions. Everything else can be pasted in live if you have it open in a tab.
Do I have to use Google Drive for the brand-brain folder?
No. It just needs to be a folder on the laptop you'll be working on, so Claude Code can read it. Drive is convenient if your team is helping you fill it.
My data is messy and incomplete. Should I clean it first?
No. Cleaning is the part AI is good at. Bring it raw — exports, screenshots, half-finished sheets. Spend your two hours gathering, not formatting.
I don't sell on Amazon or Flipkart yet.
Fine. The Landing Page & PDP Optimizer session runs off your D2C catalog, and we hand you a sample marketplace schema so you still build the reusable pipeline.
Is it safe to share my brand data in the sessions?
You keep your data on your own machine — the brand-brain folder never leaves your laptop, and you choose what appears on screen. Share only what you're comfortable showing the room.
Can my team member do the pre-work instead of me?
Yes, and they should. Gathering is delegable. Just make sure the folder ends up on the laptop you will be building on, and that you can log into the tools yourself.
Stuck on any of this?
Message the bootcamp WhatsApp group — that's the fastest route, and someone else has usually hit the same thing. Or email hello@d2cinsider.com.
Haven't installed Claude Code yet? Start with the setup guide. Not registered yet? Grab a seat — the cohort is capped at 25.