To reverse engineer a company's marketing, build its history in date order, not a snapshot. Pull one monthly traffic curve from a single source for the whole window, then log every dated move: pages added, pricing changes, the first month each ad pixel appeared, homepage headline changes, launches, funding, creator waves, and new job posts. Mark the month growth stepped up and stayed up. Group moves that happened within six weeks of each other and compare growth in the three months starting with each group with the three months before it. Treat timing as a lead, not proof, because funding, press, seasonality, and a rising category can cause the same jump. Then compare what the company does with what you do, and only copy the moves your stage, budget, and channels can support.
How the Reverse Engineer Marketing skill works
- 1
Set the target, mode, and budget
The agent reads your product marketing context, then confirms the target domain, the window (default: the last 24 months), your own company and stage, which of your data sources are connected, and a cap for paid data calls (default $5). It picks a mode: a full teardown, a quick first look, content patterns only, Funnel Hack (map every upsell and email in their funnel), Landscape (compare three to five competitors), or Watch (a monthly diff of a company already torn down).
- 2
Build the spine from free data
A bundled zero-dependency script reads the target's sitemap for page families, the Wayback Machine for the first month each pixel or script appeared, pricing and homepage headline changes by month, DNS records for email platforms and tool verifications, and certificate logs for new subdomains. Each finding becomes a dated row in moves.csv next to one traffic curve from a single source. The script detects the industry, and one of 31 industry packs adds the signals that matter there, from Shopify catalogs and CFPB complaints to npm downloads and app store reviews.
- 3
Read the funnel and go inside
The agent walks the funnel stage by stage: ranked keywords by page family, AI answer visibility, active and long-running ads, creator and founder posts, launches and press, sign-up and onboarding, pricing, referral loops, and what open roles say they'll scale next. With your approval for each action, it signs up, books a demo, or logs the email sequence for 30-60 days under your real name and screenshots every step. It never pays without your OK and checks their terms first.
- 4
Listen to the team and find hidden signals
It finds podcasts, talks, threads, and AMAs from founders and staff, and logs every dated growth claim, labeled as reported. It reads full job posts, current and archived, because "scale programmatic SEO to 10,000 pages" is the plan written down. It also sweeps paid promotion that ad libraries miss: newsletter and podcast sponsorships, affiliates, and coupon sites.
- 5
Find what moved the curve
Code does the arithmetic. Moves within six weeks of each other form one cluster, and lift is average month-over-month growth in the three months starting with the cluster, minus the three months before it. The script ranks which channel's activity leads growth by zero, one, or two months, and declines to call it with fewer than 18 months of data or 6 moves on a channel. Every lift is reported with the confounders that could explain it.
- 6
Compare with you, then turn it into plays
The agent compares the target with your company stage by stage, using the same vendor's estimate for both and your real Search Console and analytics data for depth. It mines 1-2 star reviews, switching posts, and dropped markets for where the target is weak. Each gap becomes a ranked play with evidence, effort, the metric to watch, the skill that runs it, and a paste-ready prompt, plus plays to skip and the one play to start this week.
When to use it
Your agent reaches for this skill when you ask for things like:
What's inside
The playbook covers
Reference library — 11 deep-dives
See it in action
Reverse engineer how Lovable grew. We're an early-revenue AI app builder and want to copy what worked.
Cluster 2025-03 to 2025-04 (4 moves) - 2025-03-04 seo /templates/* family passes 100 pages (Observed: sitemap-growth) - 2025-03-18 referral "Made with" badge on free outputs (Observed: Wayback) - 2025-04-02 paid Meta Pixel first seen (Inferred: pixel shows the channel was possible, not spend) - 2025-04-10 pricing Free plan limit raised from 3 to 10 projects (Observed: history) Lift: +11.4 pts average month-over-month visits growth (3 months from the cluster's start vs the 3 before) Brand searches: +8.2 pts Confounders: seed round announced 2025-03-27; category search interest rising since 2025-01. Read: credit the cluster, not the badge alone.
1. Build a templates page family (Observed: 480 /templates/* pages, ~41% of their non-brand traffic, DataForSEO estimate pulled 2026-10-30; you have 12 pages and 900 impressions/month in Search Console) Skill: programmatic-seo · Loop: weekly page-family refresh Effort: a week for the first 50 · Watch: non-brand impressions on /templates/* Skip: their podcast tour. It ran the same months as their seed round and a founder with 200K followers; you have neither. Start this week: play 1.
More things to ask
Real requests the skill is tested against:
What content works for Notion on YouTube and TikTok? I just want their thumbnail and hook patterns, nothing about their overall growth.
My competitor's traffic tripled in March right after their founder tweeted their revenue. So the tweet caused it, right? Build me a plan around posting revenue tweets.
Do a growth teardown of acme-analytics.com and sign up for their trial so we can see their onboarding and the emails they send.
Quick scan: how does Granola get customers? I have a call with a prospect in their space in an hour and $1 of budget.
We did a teardown of Acme last month. Check what they've done since and tell me if anything matters.
FAQ
How do I find out how a competitor grew?
Build a dated history. Get one monthly traffic curve from a single source, then log every move you can date from public evidence: new page families in the sitemap, pricing changes in the Wayback Machine, the first month each ad pixel appeared, homepage headline changes, launches, funding, and job posts. Mark the month growth stepped up and stayed up, then look at what happened in the weeks before it. Treat those moves as candidate causes and name the confounders, such as funding, press, or a rising category.
How is a growth teardown different from a competitor profile?
A competitor profile describes what a company is today: positioning, features, pricing, and SEO strength. A teardown explains how it got there, in date order, by lining up its moves with its traffic curve and comparing them with your own data. Use the competitor-profiling skill for a snapshot, and reverse-engineer-marketing when you want to know what worked and what to copy.
What can you learn about a company's marketing from public data?
More than most teardowns check. The sitemap shows which page families they built and when. Archived pages show when ad pixels, referral tools, and pricing changed. DNS records reveal their email platforms and tool verifications, and certificate logs date new subdomains for campaigns and programs. Public job boards show which channels they're hiring for next, app store reviews show what customers complain about, and ad libraries show their longest-running ads.
Is it legal to reverse engineer a competitor's marketing?
Studying public evidence is standard competitive research: public pages, archives, DNS, ad libraries, and public posts. The skill respects blocks and terms, using an archived copy when a site blocks scripted access instead of working around it. Signing up or booking a demo happens only with your approval, under your real identity, after checking their terms of service. It studies the company, not individuals, and doesn't build lists of their customers, staff, or followers.
How do I know which marketing move actually caused a company's growth?
You usually can't prove it from outside, so the skill reports timing as a lead, not a verdict. Moves within six weeks of each other are grouped and credited together, never one move alone. Lift compares growth in the three months starting with a group with the three months before it. A channel is only called a leading channel with at least 18 months of data and 6 moves on it, and every result names the confounders that could explain the same jump.
Can an AI agent monitor competitors' marketing over time?
Yes. Once a teardown exists, Watch mode reruns it monthly and reports only what's new: pricing and headline changes, new pixels and subdomains, sitemap growth, new job posts, review trends, and new ads and posts. Each change is appended as a dated move. When two or more launch signals land within a few weeks, such as a new subdomain, a trademark filing, new job titles in one area, or a sitemap jump, it flags a likely launch and says what it points to. Most months the honest answer is "no new moves."
Related skills
Use it in your agent
Reverse Engineer Marketing is a plain SKILL.md folder, so the same skill loads in any agent that reads the skills format. Setup guides:
Build a tool for this workflow?
Marketers run this skill mid-task — the exact moment your tool matters. Partner tiers put your tool inside the skill their agent runs.
See partner tiers