Die Oberfläche ist lokalisiert; Skill-Beschreibungen und -Dokumente werden per KI automatisch in Ihre Sprache übersetzt. Code, Befehle, Pfade und Fachbegriffe bleiben im Original, maßgeblich ist das Dokument des Autors.
Skills entdecken
Design und Kreativität · SKILL.md

referral-program

referral-program

Design, optimize, and analyze referral and affiliate programs.

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill-Dokumentation

GitHub

name: referral-program description: "Design, optimize, and analyze referral and affiliate programs."

Referral Program

You design referral and affiliate programs. Read product-marketing-context first.

Decisions to make first

  • Who — customers (referral) or partners (affiliate)?
  • Incentive structure — single-sided (only the referrer gets paid) vs double-sided (both); cash vs credit vs feature unlock.
  • Trigger — sign-up vs paid conversion vs activation. Pay on what you actually value.
  • Attribution window — typical 30-90 days; longer for B2B.
  • Anti-fraud — block self-referrals, IP duplicates, low-value sign-ups.

Program design

  1. Offer: Give X, get X framing if double-sided. Reward must matter to the segment (cash for prosumer, credits for SaaS, swag rarely works).
  2. Surface: in-app prompt at the aha moment, dashboard widget, post-purchase confirmation, dedicated landing page.
  3. Sharing tools: one-click share to LinkedIn / X / Email / SMS, copyable link, custom share copy options.
  4. Reporting: track shares, click-throughs, signups, conversions, and reward redemptions.
  5. Lifecycle nudge: prompt happy users (post-purchase, post-NPS-9-10) to refer; never spam.

Output

  • Program spec doc (incentive, trigger, attribution, fraud rules).
  • In-app surface mockups.
  • Email/in-app prompts (hand off to email-sequence).
  • Reporting dashboard requirements. Adapted from coreyhaines31/marketingskills (MIT).
Gespeicherte Skills