Metered Paywall and Subscription Build for LatAm Publishers

Where the build starts: The meter, set from your own server logs rather than a vendor default. The number that decides this business is the article count at which a returning reader's probability of subscribing stops climbing. On the automotive and home titles I have worked on it has landed between three and seven articles in a rolling 30 days, and on banking and insurance titles lower, because the intent arrives sharper. Ship the vendor default of five and you have guessed at the one number the entire subscription line rests on. What gets installed: A metered paywall on Piano, Poool, Pelcro or Stripe Billing, chosen for the stack your engineers already maintain and the reporting your finance team has to close books against. The meter is enforced at the CDN edge, in Cloudflare Workers or Fastly VCL, not in client JavaScript. A wall that view-source defeats has already been defeated by exactly the readers worth charging. Registration comes before payment: a registered reader who has not paid is the population the propensity model learns from, and an email address survives a failed checkout. Payment, which is where launches in Mexico actually fail: Card-only checkout is why a paywall designed for a US title underperforms here. This build ships OXXO cash vouchers, SPEI transfer, Mercado Pago, and card with 3-D Secure and meses sin intereses where the issuer supports it, through Stripe or Conekta. Cash subscriptions do not auto-renew, so renewal is separate work: a reminder sequence, voucher regeneration, and a dunning model that does not classify a reader who pays four days late in cash as involuntary churn and cut off their access. Search: Paywalled articles carry isAccessibleForFree false and a hasPart block whose cssSelector matches the element actually gated, so Googlebot reads a paywall as a paywall and not as cloaking. Flexible sampling is set per section, and any section where free sampling earns more from advertising than the wall earns from subscriptions stays open. That call gets made from a quarter of real numbers, not from principle. Not included: No custom billing ledger; Stripe or your existing processor holds subscription state. No app-store in-app purchase. No acquisition pricing at one peso for three months, which teaches the reader what the product is worth and then asks them to pay eight times it. No retention flow that hides the cancel button or routes it through a phone call. I do not write, commission or restructure the journalism, and I will say plainly when the reason nobody subscribes is that nothing on the site is worth paying for. Who this is not for: Sites where fewer than roughly one session in ten comes from a reader who was also there within the last 30 days. That site does not have a subscription problem yet. It has an audience-relationship problem, and a wall placed on top converts almost nobody while costing traffic. Newsletter and registration first, paywall in the following year. How it ends: Cohort reporting in GA4 and BigQuery separating trial-to-paid, involuntary churn, voluntary churn and win-back, with the queries handed to your analyst so they run without me. A written record of every threshold set and why, so whoever touches the meter next knows what they are changing. Then a scheduled review at day 90, when the first renewal cohort is real, and a decision on the meter based on what happened rather than on what we assumed in week one.