Data I collected myself.
Not a roundup of other people’s studies. Each of these comes from a dataset I pulled, cleaned, and had to make a decision with. Method is stated on every one so you can reproduce it or argue with it.
What the numbers said.
- I Analyzed 866 Ads From 30 Custom Software Companies Screened from 585 advertisers running paid ads to US businesses. Three tiers, real prices, and the offers that have survived longest. MARKET STUDY
- Only 3 of 30 Software Companies Publish a Price Pricing opacity is near-universal in this market, which makes publishing one the cheapest differentiator available. PRICING STUDY
- Only 4 of 30 Offer a Guarantee And three of those four run the longest-lived ads in the set. The exact wording of each. OFFER STUDY
- Ad Longevity Is the Closest Thing to a Public Conversion Signal One ad ran 564 days on a single creative. Nobody keeps paying for a losing ad. METHOD
- The Vibe-Coding Wall One company launched 80 ad variants in two weeks around a single idea: you already tried building it with AI, and hit a wall. MARKET INSIGHT
- 10 Hook Patterns Behind 33.1M Plays 19 Instagram posts over 1M plays, transcribed and analyzed. The patterns, ranked by plays behind them. CONTENT STUDY
- TikTok and Instagram Are Not the Same Game Shares are the lever on one, not the other. Duration is bimodal. The middle is dead. PLATFORM STUDY
- Carousels Are Depth, Not Reach A tenth of the likes of a reel, and the same number of comments. What that means for what you post. FORMAT STUDY
- 44% of AI-Enriched B2B Contacts Were the Wrong Employer Measured across 254 contacts at 100 companies. Two failure modes, and the check that catches both. DATA QUALITY
- The False-Zero Machine Ad-library search ranks the wrong page first and returns a confident zero. How to detect it before it poisons your analysis. METHOD
How I work.
The research feeds the guides. If you want the how-to rather than the findings, start at resources, or see the comparisons for head-to-head calls on specific tools.