Three stats, three sources — and the one I had to change

Day 70 shipped the first real article in the Content Library: a 1,964-word piece on what a fake review actually costs a small business, and three ways to fight one — flag it through Google, reply publicly, escalate legally. It also shipped a RelatedReading trailer that now sits at the bottom of every article page. But the part of today worth writing about isn't the word count. It's that the article carries three externally-sourceable stats, all three got checked against a primary source before a word was written, and one of them isn't the number I originally asked for.

Verification as a feature

The article leans on three numbers: Michael Luca's finding that each additional star is worth roughly 5-9% in revenue (Harvard Business School working paper, 2011, revised 2016), Womply's finding that businesses in the 4.0-4.5 star bracket see a 28% revenue premium over average (a 200,000-business study from 2019), and the FTC's $51,744-per-violation penalty from the 2024 Rule on the Use of Consumer Reviews and Testimonials.

My instruction going in was strict: verify or hedge, never invent. If a number can't be confirmed, say "roughly" or drop the sentence — inventing a stat to fill a sentence is worse than just not having the sentence. That instruction got tested immediately. I'd originally asked for a specific comparison: 4.0-4.5 star businesses against 3.0-3.5 star businesses, with some specific percentage gap between them. That exact comparison doesn't exist in Womply's published data. Rather than round off toward a number that sounded plausible, the substitution was a real, checkable figure that does exist — the 28% premium over average for the 4.0-4.5 bracket — flagged explicitly as a swap before the article shipped.

The honest read: most small-business content marketing runs on invented stats. The number is round, the source is a vague "studies show," and nobody checks. Skeptical readers can smell that from a distance. Verification isn't a competitive edge here — it's the floor. It's just that most floors in this category are currently on fire. Day 67 was about TSC passing clean without proving the feature actually worked. This is the same lesson one level up: the compiler doesn't check your facts either.

The FTC catch

Way 3 of the article covers when a fake review is worth escalating legally. My original instruction put defamation claims and the FTC's $51,744 penalty in the same section without drawing a line between them — an easy way to imply something false. The pre-ship read caught it. That $51,744 figure is leverage the FTC holds against parties who create, buy, or sell fake reviews — reputation-fraud operations, a competitor paying for fake reviews about you. It is not a payout an individual business owner collects by suing the person who posted a fake review; that's a separate mechanism, capped by state defamation law and actual provable damages.

The article now says it straight: the penalty is federal leverage, not a personal cause of action, but it's real context — regulators aren't shrugging at fake reviews as a gray area anymore. Small catch. Would have been an embarrassing one to ship, because getting the mechanism wrong isn't a typo, it's a wrong legal claim in a piece meant to build trust.

The RelatedReading principle

Every article page now ends with a RelatedReading block — up to three same-category articles, topped up from the featured set if the category's thin, silent if there's nothing worth surfacing. No orphan heading over an empty grid.

The design decision worth naming: card density should match the surroundings. The homepage FeaturedResources block uses the same heavy card weight as the Pricing tiers below it, because it's competing for attention against something dense (that's the Day 69 polish fix). RelatedReading uses lighter cards, because the reader just finished 1,964 words and a second heavy block would read as a shout. Lighter says "here's more if you want it" instead of competing for the same attention twice. Same principle, opposite output — that's the part worth keeping.

Closing the Day 69 loop

Yesterday's post was a bet: ship the structure — the hub, the routes, the sitemap, the homepage teaser — before writing a word of content, and let the articles land into slots that already work. Today's article was written directly into a URL that already existed, was already linked from the homepage, already had four sibling stubs, and now has the RelatedReading trailer surfacing those siblings automatically. One article in isn't a pattern yet — four more stubs to go — but the bet is holding so far.

The Content Library isn't purely an SEO play. It's the surface where the discipline behind the product becomes visible to a skeptical reader. If the article is honest and specific, the software behind it is a little more likely to be too. The library fills in slowly, one verified sentence at a time.