— 18 pieces · sole marketing leader
The automation layer behind one marketer shipping at the volume of a team. Roughly 5 foundational artifacts spanning translation, production, measurement, and reasoning, plus a longer tail of scripts, flows, and frameworks that returned hours per week to strategic work.
Per-batch localization tool. Ran in production on 2 locales across 46 English pages, totaling 92 pages. Replaced 3 CS-team part-timers running ~21 hrs/week × 6 weeks of manual translation work per batch. Saved 96 hours of labor.
End-to-end blog production agent. LLM APIs + Semrush keyword research + auto image generation + alt text + slugs + page titles + meta descriptions, fed by the internal-link analyzer and audited by the schema auditor.
CAC payback methodology authored from first principles. Three-flavor framework (naive / simple / empirical), piecewise empirical retention curves with front-cliff modeling, gross-margin haircut, segregated monthly-vs-annual cohort economics. Pairs with a closed-form formula relating ARR/CAC ratio, gross margin, payback target, and monthly churn rate.
HubSpot → Microsoft Ads offline-conversion importer. SHA-256 email and E.164 phone hashing per Microsoft's API spec, automated weekly from the HubSpot contact-to-payment-data joins. Replaced manual exports and fed Microsoft's Smart Bidding equivalent the CRM signal it otherwise lacked.
Industry-positioning automation derived from the Fletch framework. A prompt chain that ingests a real ICP brief, runs the positioning research workflow per industry, and outputs a populated positioning table per cell. Compressed multi-day workshops (4-8 hours) to ~30 minutes per industry; 85% first-output buy-in when fed a real ICP brief, degrades to generic on thin inputs.
Sitemap crawler that maps a page corpus and surfaces topically relevant internal-linking targets for any new draft. Fed the agentic-blog system the link graph each new article should join. The SEO move most teams skip because it's too tedious to do manually.
Structured-data compliance auditor. Validates JSON-LD output against the schema.org spec relevant to each page type (Article, Product, Service) and flags missing or malformed fields before publish. Ran as a gate in the agentic-blog pipeline.
Headless-browser CMS builder. Playwright walks a page brief through the CMS editor UI (fields, slugs, meta, image uploads), replacing manual entry. Scoped to free ~12 hours per week of intern CMS time during the 235-page rollout.
Figma plugin that fans out ad-creative variants from a single source frame. Locale, copy, and offer swap via JSON config; output is publish-ready PNGs and MP4s for Meta, Google, and Microsoft. Removed the per-locale design step from the production loop.
Google Ads Customer Match list generator. Pulls customer segments from HubSpot (active subscribers, churned, free-trial-stalled), SHA-256-hashes per Google's spec, and pushes to Customer Match audiences on a weekly cadence. Powers audience-targeting layers on branded and competitor campaigns.
Custom HubSpot reports joining contact-email to Braintree credit-card and bank-transfer payment data. Built Q4 2024 because HubSpot's lifecycle-stage automation was unreliable; measurement worked at the contact-email to payment-data layer instead, insulated from the 62.6% customer-stage undercount that surfaced in the April 2025 cleanup. Remained as redundant verification post-cleanup.
5-tier country-efficiency framework. Top-10 markets ranked on ARR/CAC ratio × empirical CAC payback × trial-to-paid conversion × annual subscriber mix. Each tier prescribes the action: A Compounder (scale), B Workhorse (maintain), C Bifurcated (annual-only), D Marginal (pull paid), E Broken. Drove a five-figure DKK monthly reallocation from underperforming geographies.
Campaign architecture spine. <MARKET>-<TYPE>-<SUBTYPE> taxonomy applied across Google, Microsoft, and Meta Ads; 36 of 89 tenure-wide campaign renames. The artifact that survived three paid-agency transitions: each new agency inherits names already conformant, so reporting filters and naming-driven attribution rules survive handover.
PLG onboarding framework. Trigger logic, email tone, and activation-definition rules abstracted into a portable template that drops into any PLG SaaS. Authored from pattern-matching across prior tenures, Paperturn-validated. Not Paperturn-deployed (PLG onboarding wasn't the bottleneck the work prioritized); capability is the documentation and the workflow, not a shipped instance.
Custom ARR report generator. Pulls Chargebee's recurring-revenue export, joins it against active-customer cohorts from HubSpot, and outputs a country × plan-type ARR breakdown matching the LTV formula's grain. Replaced ad-hoc SQL pulls that historically blocked the client-lifetime reporting module.
Pandas utility for cross-system CSV joins. Built during HubSpot Cleanup v2 to merge contact exports against Chargebee subscribers, Braintree payment data, and Google Ads click-ID tables on a per-contact basis. Reused on any reporting question that needed data from more than one system.
Operational-glue automations across Zapier, n8n, and Make.com. ~25 flows across tenure: notification routing for paid-account churn, lead-form pass-throughs from the website to HubSpot, CRM enrichment from Chargebee, sync triggers for ARR dashboards. The boring infrastructure that kept the rest of the stack readable.
Reusable LLM prompt chains, version-controlled. ~6 chains underpinning the agent tools above: positioning research, copy generation against a brief, translation-glossary lookup, page-template population, ad-creative ideation. Each chain hand-tuned against a worked example before going into production.