Short posts that go one level deeper than the guides, with sources. A new one most days.

Monthly logo churn of 3 to 5 percent is typical below 1 million dollars in ARR, by CRV's 2026 benchmarks. What changes after that, and when we worry.
Read more →
Start from the phrases customers repeat when they describe the problem, check each one against other customers, and keep listening after the page ships.
Read more →
Fourteen days is the default B2B SaaS free trial, but length barely moves conversion. What ChartMogul and Userpilot report, and why the card question is first.
Read more →
A ghostwriter can work for a B2B founder when the ideas are the founder's own. What LinkedIn and Edelman found, and what it costs you in interview time.
Read more →
Buyers ask AI tools to compare vendors, and a brand's own pages are rarely the source. What G2 and Trendos found, and how we write a fair comparison page.
Read more →
AI drafts with no real input fall back on the same LinkedIn structures. What Originality.ai and LinkedIn found, and what gives a post a real point of view.
Read more →
Ahrefs tested 1,885 pages that added JSON-LD against matched controls. What it found for ChatGPT and Google, and what earns AI citations instead.
Read more →
Fathom, Granola, and Fireflies turn sales calls into recaps and searchable notes. We look at what each one outputs and what a call still needs to become a post.
Read more →
When TikTok is worth a funded test for a B2B company, which products suit a short video feed, how to budget it next to Meta, and how to judge it on pipeline.
Read more →
How a small B2B team can run voice of customer research from the sales calls and support tickets it already has, and what to do with the words customers use.
Read more →
Motion's data on 578,750 Meta creatives ranks text-only ads first for hit rate. When video earns its cost, and what Facebook's full-screen test changes.
Read more →
How to find the objections that keep coming up on your sales calls, how many calls you need, how to tell a stall from a real no, and who keeps the list.
Read more →
Meta's default attribution window since March 2026, when a 1-day click window fits better, and why changing the setting moves your reports but not your results.
Read more →
Turning a sales call into a LinkedIn post: which moment to pull, how specific the quote has to be, and when to ask the customer before you share their words.
Read more →
Founder intelligence turns what a founder and their customers say on calls into content, positioning, and a running list of what buyers raise. How it differs.
Read more →
Most B2B software hides its price. Only 4 percent of G2 product profiles list a number, and 58 percent of B2B SaaS respondents publish none at all.
Read more →
No single number counts as good. Bessemer targets under 12 months for SMB sellers and under 24 for enterprise, and the 2026 industry median is 16 months.
Read more →
Exposure count beats the calendar. Published Meta lifespans run 2 to 3 weeks, and an analysis of 1.29 billion dollars in spend puts half off by day 28.
Read more →
Cassandra's 2026 marketing mix study puts Performance Max at 4.64x median incremental ROI, between Search Non-Brand and Search Brand, with steadier results.
Read more →
Almost every tool ChatGPT names has a G2 and Capterra profile, yet review volume tracks slightly negative against ranking and the pages earn no citations.
Read more →
ChatGPT cites Wikipedia often, but studies count it in different ways and disagree, and Wikipedia's own rules mean no company can buy its way onto a page.
Read more →
Established companies spend near 7.7 percent of revenue on marketing in Gartner's 2026 survey, while startups budget from funding, a different measure.
Read more →
Ask active users how they would feel without your product. Under 40 percent very disappointed, fix the product before you scale go-to-market spend.
Read more →
Google and Microsoft search ads average 6.64 percent CTR and Meta traffic campaigns 1.71 percent. Neither is a target, so judge an ad by its own history first.
Read more →
Advantage+ needs conversion data before it works, and manual still wins under 100 dollars a day. Most accounts should run both, with a job for each.
Read more →
ChatGPT and Google AI Overviews name similar brands about a third of the time and cite almost entirely different sources. Plan for each engine on its own.
Read more →
An llms.txt file is a plain text index of your key pages for AI systems. The one large study of it found no link to citations, so keep it small.
Read more →
Reddit is the most cited domain in AI search, and its ChatGPT share fell 86% in under a month. What a brand that does not own Reddit can do about it.
Read more →
Domain authority barely predicts who ChatGPT names. Semrush's 50,000-brand study points to depth on buyer questions, and Seer's data adds freshness.
Read more →