Referral & Outreach

The referral and affiliate engine, the cold outreach system, and the per-type sourcing playbook. Two channels, one owned stack.

Two acquisition channels that both ride the existing spine: bbe-lead, D1 bbe-ledger, attribution_summary. Zero new tables for either. Referral pays people to send you customers, outreach finds customers directly.

Part 1: Referral & Affiliate Engine

Ruled 7/11, build not started

DH pays a default 30% to partners: 30% of first payment on one-time offers, 30% recurring for 12 months on subscriptions, and a flat $500 named-referral bounty on done-for-you. DFY stays out of the open program entirely, and franchise-side offers are NEVER in the open affiliate program (flat-fee only, see the NASAA note below).

The rate card

Offer	Price	Default rate	Partner earns	Founding rate (first cohort, 90-day window)
Brand Builder Club	$1,500 one-time	30% of sale	$450	40% = $600
BBB DIY	$997 one-time	30% of sale	$299	40% = $399
Paid AI tier	$49 / $99 per mo	30% recurring, 12 months	~$176 / ~$356 per referral-year	same
so bexy	$29.99 per mo	30% recurring, 12 months	~$108 per referral-year	same
Done-for-you	high-touch	EXCLUDED from open program	flat $500, paid after client’s 2nd paid month	n/a
Franchise-side	n/a	NEVER in the open program	flat-fee structures only	n/a

Advocates (partners who are also customers) can elect account credit instead of cash at the same rates, which skips the W-9/1099 machinery entirely.

Attribution: 60-day last-click, code beats click. A partner coupon code wins attribution over the click trail every time, which closes the dark-bridge blind spot: the code arrives with the money even when every click was dark. Commission accrues PENDING at sale, flips to DUE after a 30-day hold matching the refund window, pays monthly on the 15th, $50 minimum with rollover. Self-referral earns nothing.

Build vs buy: build, with a named tripwire

Click and conversion tracking is the easy third of what Rewardful-class tools sell, and the spine already covers it. What those tools actually sell is mass payout disbursement and tax-form ops, a 30-minute monthly Claude task below ~25 partners.

The mechanism

The link. Every partner gets a slug. ?via=<slug> on any DH URL. beacon.js captures it, writes referral.click; opt-in stitches session to contact and writes referral.lead.
The code. Every partner also gets a coupon code (a real discount, ~10%). Today it’s a Podia coupon; the own-account Stripe webhook maps code to partner and writes referral.sale. After the Podia exit, same code becomes a Stripe promotion code, nothing else changes.
The events, zero new tables: referral.click, referral.lead, referral.sale, referral.reversal, referral.payout. A partner’s balance is never stored, always derived: sales minus reversals minus payouts, filtered by hold_until.
The rollup: attribution_summary gains partner_id as its third dimension beside campaign_id and variant_id.
The registry: partners are contacts in D1. The human lens is the existing Notion Partnership Pipeline database, extended with outbound Status options (Prospect / Invited / Signed / Active Partner / Paused).

Payout rail

Stripe Global Payouts (~$1.50/payout, recipients need no Stripe account). W-9 collected from every US partner before the first payout, no exceptions, or 24% backup withholding applies. The 1099-NEC threshold is $2,000 for tax year 2026 (raised from $600 by the One Big Beautiful Bill Act). Non-US partners need a W-8BEN before payout.

Compliance: the two launch blockers

A commissioned partner is an endorser with a material connection. The connection must be disclosed clearly and conspicuously in EVERY piece of content, and DH is liable when a partner fails to disclose. Current penalty exposure: $53,088 per violation, and each non-compliant post can count separately.

Two extra tripwires: earnings claims are the highest-risk category for a business-growth product (no stated or implied income results without pre-approved, substantiated claims), and the Fake Reviews Rule (16 CFR Part 465) bans paying or rewarding conditioned on positive sentiment.

A full disclosure kit (social post, YouTube verbal + description, partner email, blog/review page language) is drafted and ready to ship with every partner’s swipe pack, so the compliant version is the lazy version.

The locked rule stands: no success fees on franchise deals, flat retainers only. The sharpening: the fee shape alone does not keep you outside registration, the activity does. Under the adopted Model Act (May 4, 2026), anyone who talks to specific franchise prospects, qualifies them, or steers them to a brand is doing broker activity, regardless of how the fee is structured.

The defensible lane: DH provides marketing services only, no DH contact with specific prospects about franchise merits, compensation flat and never variable with leads or deals. One clean carve-out: the Model Act exempts existing franchisees earning up to $5,000/year in referral fees, which fits a Bex+Co stylist referring a peer.

Recruiting: warm circles first

~10% of signed affiliates ever drive a sale, and ~80% of program revenue comes from the top 5%. The program is a search for 3-5 super-partners.

Wave 1, this month: the warm inner circle

Existing clients and close operators. Personal invite from Andrew or Stefi, Advocate framing, credit-or-cash choice. Target: 5-10 signed, first paid referral from this wave.

Wave 2, same month: the 50k audience

One announce email plus a permanent /partners page. Self-serve application feeds the Pipeline DB, manual approval.

Wave 3, after the kit is proven: creators

Mid-tier creators from the Audience Landscape, pitched on the founding 40% rate and Andrew’s own Copy.ai affiliate pedigree ($0 to $100K MRR run at 45%).

The minimum enablement kit, v1

A /partners recruiting page, a per-partner asset page at share.designhacker.com/partners/<slug>/, a swipe pack (3 emails, 5 social posts, 1 YouTube blurb, every asset carrying the disclosure language already baked in), a creative kit, and a monthly branded HTML earnings statement generated from the ledger. That’s the whole v1. A live partner dashboard is v2.

Part 2: Cold Outreach & Sourcing

Engine locked, DuraFleet = instance one

The outreach engine finds and messages prospects who never opted in. Front half (find, scrape, score, dossier, draft) is the lead-finder skill, unchanged. Back half (send, track, consent) is new. Data spine is D1 bbe-ledger, unchanged, with a new outreach.* event family.

The six decisions, ruled

Decision	Ruling
Sending platform	Smartlead as the sequencer (isolated client workspaces, true tenancy separation), Mailforge-class SMTP inboxes underneath, never Google Workspace
SMS	Cold SMS is dead, TCPA has no B2B exemption, damages run $500-$1,500/text with no cap. SMS stays opted-in follow-up only, after PEWC (prior express written consent) is captured
Who presses send	Human approves sequence copy and the first 50 sends per client. Machine sends the rest. AI grades replies, human approves every outbound reply draft for the first 90 days
First client after DuraFleet	Brand Builder Pros (Andrew himself). Zero sales cycle, and it proves the pack works on a different anchor source (Google Maps instead of FMCSA)
Budget ceiling	~$500/mo per 3-territory cohort. Researched stack lands at $355-$465: Smartlead, Mailforge, domains, Apify scraping, enrichment waterfall, verification, Claude API
Client dashboard	Internal for v1. Clients get a weekly branded HTML one-pager, same pattern as the DH X-ray report

The SOP loop

Find, scrape, load, send, track, optimize.

Phase 0: Intake and lock

Signed consent/authorization agreement before anything else. Fill the Pack Config (ICP, territory, offer, anchor source, scoring weights). Geo-fence: US only, Canada and EU/UK excluded. Schema lock gate: the outreach event payload is ratified by Andrew before any paid tooling purchase.

Phase 1-2: Find, enrich, verify

Anchor source ladder: government registry first (free, zero scraping risk), then licensed vendors, then Maps scrape, then compiled directories, then manual LinkedIn. Two-tool enrichment waterfall (Prospeo, Anymail Finder), 45-60% yield to verified emails. Verify every address, send-gate under 2% projected bounce.

Phase 3: Build sending infrastructure

3-4 lookalike domains per client/territory, bought in a SEPARATE Cloudflare account, 301-redirected to the real site. 2-3 Mailforge inboxes per domain. SPF/DKIM/DMARC, 14-21 day warmup before the first real send.

Phase 4-5: Campaign and ramp

4 touches over 14 days, under 80 words, plain text, no links or images in email one, no tracking pixels. Ramp 10/day per inbox, +5 every 2-3 days, cap 25-30/day.

Phase 6-7: Reply, route, optimize

Webhooks write outreach.* events to D1. AI grades every reply (interested / question / not-now / wrong-person / unsubscribe / hostile). Nightly grader auto-pauses any inbox over 1.8% bounce or any domain over 0.08% complaints.

The event payload, locked

Zero new tables. New outreach.* event types on the existing contact/identity/consent/event spine.

Event type	Payload keys
`outreach.sent`	campaign_id, variant_id, sequence_step, inbox_id, domain, territory_key, esp_message_id
`outreach.bounced`	campaign_id, inbox_id, bounce_type, esp_message_id
`outreach.reply` / `outreach.reply_graded`	campaign_id, variant_id, esp_message_id, grade, confidence
`outreach.unsubscribed`	campaign_id, method
`outreach.meeting_booked`	campaign_id, territory_key, calendar_ref
`outreach.paused`	scope, reason, threshold_value

journey.stage stays the one pipeline: cold, contacted, replied, qualified, booked, customer. attribution_summary gains campaign_id and variant_id.

The per-type sourcing playbook

The 80% reuse claim survived. What flexes per business type is three things: the anchor source, the copy pack, and the channel. Business inboxes tolerate cold email, consumer Gmail punishes it.

Type	Entry rung / anchor	Channel	Copy angle
B2B with a registry (DuraFleet)	FMCSA carrier census, free, 4.4M carriers	Cold email, lookalike domains	“You’re running 22 power units out of Suffolk”
B2C local service, referral partners	Maps scrape + PropertyRadar	Cold email to business inboxes	Partner economics, radius proof
B2C local service, homeowners	Public property/permit records	Postal only. Cold email is a skip, three independent kills	Neighborhood momentum: “three kitchens underway on your street”
Individual professionals (stylists)	State license registries (census) + IG/Booksy scrape	Instagram primary (paid + inbound-triggered DM funnels), email as a light flank only	Peer proof, show don’t tell
Franchisee buyers	FDD Item 20 exhibits (4 free state portals) + SBA 7(a) FOIA loan data	Cold email to multi-unit operators, LinkedIn manual flank for first-timers	Expansion economics without earnings claims, territory availability