The hook is the unit of test. Not the ad creative, not the landing page, not the offer, the hook. So this is the hook log. Every variant we've tested across every channel, with the actual copy, the actual impressions, the actual click-through rate, and a one-line takeaway. 9 winners kept. 4 still running.14 total. Updated rolling.
"GitHub commit velocity preceded 219 startup-period observations by 21-47 days."
Takeaway: Numerically specific + research-anchored. Doubles as the SSRN paper claim. This is the hook the entire funnel is built on. Holds across every channel except HN (where research-paper framing reads like marketing).
"Stop playing warm-intro roulette. Source on engineering data the partners already see."
Takeaway: Polarising enemy framing. Names the visible villain, the warm-intro game, without naming a competitor. Doubled the home scroll-to-pricing rate vs the prior 'Find startups before they raise' hook.
"Five startups every Sunday whose engineering just accelerated."
Takeaway: Right-shaped bait hook. Specific number, specific cadence, specific shape of result. The reason this beats 'weekly intelligence digest' is that it's countable, readers can mentally complete the offer before they click.
"Imagine a Tuesday in August: you reach the founder before the deck circulates."
Takeaway: Future-pace mental movie. Specific weekday + month does more work than 'imagine your sourcing process working.' Lifted Dashboard click-through ~30% over the prior generic close.
"By the time the deck reaches you, the round is forming."
Takeaway: Pain-side framing of the same insight. Pairs with the 47-days hook on the result side. Reads better in B2B social copy where 'fundraises' as a noun fatigues.
"The Data Nerd reads commit logs for fun, so you don't have to."
Takeaway: Founder-character hook. Not measured per-click, measured by how often readers reply to emails by name. Reply rate roughly tripled when From-name shifted from 'GitDealFlow' to 'The Data Nerd'.
"47 days before the deck."
Takeaway: Ultra-compressed variant of the load-bearing hook. CTR is lower than the long form but qualifies harder, 1-in-3 clicks become a signup vs 1-in-5 for the long form.
"Read the engineering of a startup in five minutes the way an analyst reads it in three weeks."
Takeaway: Decade-in-a-Day frame. Compresses the time-saved value to a number the buyer recognises, 5 min vs 3 weeks. Best CTR per impression of any non-pricing hook on the site.
"The Harmonic dataset, at the price of a Spotify subscription."
Takeaway: Comparison hook. The Spotify reference is doing the heavy lift, readers translate €49 instantly. Edge case: works on the comparison page, doesn't transplant to home (too specific without setup).
"You're not Insight Partners. Stop paying like one."
Takeaway: Disqualifier hook. Repels wrong-buyer harder than it attracts right-buyer. Works as a shape-of-traffic filter, not a click engine. Kept as control in the FOR/NOT-FOR block.
"Engineering acceleration is the new opportunity. Network is the old vehicle."
Takeaway: New-opportunity hook. Frames the product as a category replacement, not an upgrade. Stays the spine of the welcome-sequence D2 email, every other variant lost to it.
"Your agent reads commit logs for fun too. Wire it up via MCP."
Takeaway: Agent-audience hook, publishing where the buyer already is. Highest-CTR hook on the site, but the audience size is small. Quality: it qualifies the buyer harder than any pricing copy could.
"Free 104-page book: 7 GitHub Signals That Predicted 219 Fundraises."
Takeaway: Free+Shipping hook. Long-form anchor. Conversion data still maturing. Hypothesis: book-readers convert to Dashboard at 3× the rate of digest-only readers.
"Paste a tweet. €1. We'll grade the founder's GitHub against the 219-fundraise pattern in 4 hours."
Takeaway: Micro-tripwire hook. The €1 number is doing the qualifier, anyone who thinks €1 is too much is not the buyer. Watching whether it cannibalises €7 First Look or amplifies it.
We don't run isolated A/Bs. We ship a hook into a channel where it has the right home (Twitter for short hooks, email for character hooks, /pricing for comparison hooks) and watch it for 14 days against a baseline. If it beats baseline by >15% on CTR and the sample is > 1,000 impressions, it gets promoted. If it's close, it stays in rotation. If it underperforms, it's killed and the takeaway gets logged.
Every hook above is in production today, was in production at some point, or is mid-test. Numbers are last-30-day rolling unless the row says otherwise. We don't round to make winners look better; CTR is to one decimal place even when the cleaner number would read stronger.
🚀 Explore Our Network