Recovering an inherited publishing-commerce platform
Reconstructing an undocumented operating path, taking the product to production, and turning the stable system into measurable commercial delivery.
Outcome: A production operating model, 22× faster image delivery, 83% smaller generated files, and annualized infrastructure costs reduced from about $540 to $72.
What arrived
The handover was a code snapshot without usable development history or deployable infrastructure configuration. Before expanding the product, I reconstructed its runtime, environments, deployment path, integrations, operating documentation, and production controls.
I worked part-time as the hands-on fractional product and technology owner, directly owning and implementing the agreed work while coordinating founder and partner decisions.
Rebuilding the operating path
I separated recovery work from product work. The first sequence established repeatable local, staging, and production environments; automated deployment; health checks; monitoring; rollback procedures; access hardening; and operating documentation.
I preserved useful service boundaries instead of rewriting the application for preference. The priority was to make a release understandable: what would change, how it would be tested, how the dependent services would move together, and how to recover if the release failed.
Owning delivery end to end
Once the operating baseline was stable, I owned the product and technology path from prioritized improvements through production follow-up.
Measured changes
- The image-heavy loading path improved from about four minutes to about 11 seconds—22 times faster—while image file sizes fell 92%.
- Generated files became 83% smaller, reducing the measured download time from 1 minute 40 seconds to 17 seconds.
- The annualized AWS bill fell approximately 87%, from about $540 to $72 per year, saving about $468 annually while the required services remained available.
- Database growth in the measured workflow fell 88%.
These were different interventions with different baselines. I kept them separate rather than combining them into one general performance claim.
From recovery to evidence-led growth
After production and delivery became reliable enough to measure, I implemented consent-aware product analytics, first-party attribution, technical SEO, referral continuity, and lifecycle foundations.
A subsequent 28-day comparison recorded:
- Google Search clicks increasing by about 113%.
- Consent-gated PostHog visitors increasing by 442%.
These figures describe observed, non-causal movement within consent- and instrumentation-dependent coverage. They do not establish conversion lift or prove that one intervention caused the change.
What remains private
The client and product remain intentionally unnamed. Private URLs, repositories, provider identifiers, customer and order data, infrastructure topology, screenshots, credentials, and security findings are excluded. The measurements describe their recorded scopes; they are not guarantees for another product.