What was needed
A measurable multi-platform publishing programme with clear automation boundaries and reliable reporting.
Back to portfolio
Technical project management · AI systems · Programme design
Content syndication programme
I built a multi-platform publishing programme from zero in one month, then audited the numbers before calling it a success.
My work covered programme design, platform classification, content-engine handoff, scheduler architecture, publishing reliability, reporting, and measurement audit.
1 monthfrom plan to working programme
352articles produced
62.6% → 25.1%measured duplication
132s → 5sweekly report runtime
What I did
Designed the platform matrix, publishing routes, retries, deduplication, live checks, and measurement controls.
What was achieved
352 articles; duplication reduced from 62.6% to 25.1%; weekly report runtime reduced from 132 seconds to 5 seconds.
Public documentation
The case is documented as a delivery story.
Use the sections below to review the brief, delivery plan, risk controls, status logic, and close-out. Sensitive implementation details remain outside the public boundary.
Programme briefThe goal, guardrails, scope, and decision to keep the pilot reversible.Delivery planA phased plan from strategy validation to live measurement.Risk & measurement registerReliability controls and the distinction between output and impact.Status reportWhat shipped, what the audit found, and what could not yet be claimed.Close-out & lessonsA narrow conclusion, follow-up work, and the privacy boundary.
Programme brief
The goal was measured distribution, not volume for its own sake.
The programme needed to turn one source video into distinct, branded versions for multiple publishing platforms. It also needed a scheduler, connectors, review paths, and reporting that could answer whether the distribution was useful.
- Goal: validate a working publishing programme before scaling an unproven model to clients.
- In scope: content-engine handoff, platform matrix, scheduler, API/RSS/manual publishing paths, live checks, and reporting.
- Guardrail: platform credentials, client names, source content, account details, and publishing data remain private.
- Core decision: validate cheaply, separate the content engine from publishing, and keep the pilot reversible.
Delivery plan
The programme separated content production from distribution reliability.
That separation made it possible to improve the publishing layer without pretending that every platform supported the same level of automation.
| Phase / risk | Action / impact | Result / control |
|---|---|---|
| 1. Validate | Test the strategy with a small set of branded platforms. | A reversible pilot with explicit measurement questions. |
| 2. Prepare content | Move from source video and transcript to a master and platform-specific variants. | Distinct content ready for review and scheduling. |
| 3. Distribute | Use direct APIs, RSS paths, or a manual queue based on platform capability. | Automation where reliable; honest boundaries where not. |
| 4. Harden | Add retries, deduplication, concurrency control, render checks, and live URL checks. | A safer publishing pipeline with visible failure states. |
| 5. Audit | Reconcile database totals with public feeds and test crawlability and claims. | A report that separates evidence from assumptions. |
Risk & measurement register
The strongest project decision was to make measurement limits visible.
The system could produce output quickly, but output was not proof of reach, backlinks, or AI visibility. The reporting model was designed around separate questions and separate evidence.
| Phase / risk | Action / impact | Result / control |
|---|---|---|
| Truncated content | A generator may save an article that ends mid-sentence. | Reject incomplete output and add a repair path. |
| Published ≠ live | An API may return an ID while the post remains a draft. | Check returned publication state and the live URL. |
| Unsupported automation | Some platforms do not offer a safe publishing API. | Classify API, RSS, manual queue, or lower-priority path. |
| Misleading metrics | Clicks, backlinks, and AI citations require different evidence. | Separate scheduler, analytics, crawler, and AI-readability checks. |
| Unsupported claims | A reachable source URL does not prove a claim is true. | Editorial audit and correction pass before reporting confidence. |
Status report
The report showed both delivery progress and evidence gaps.
A useful status report did not hide the limits of the measurement model. It reported what shipped, what improved, and what remained unproven.
- Delivered: working publishing system across roughly 12 Web2 platforms, with API, RSS, and manual paths.
- Measured: 352 articles, 527,369 words, duplication reduced from 62.6% to 25.1%, and report runtime reduced from 132 to 5 seconds.
- Issue identified: 79 articles were blocked by robots rules or noindex directives.
- Decision required: treat human traffic, backlinks, and AI citations as unproven until better instrumentation exists.
Close-out & lessons learned
The programme succeeded as a working distribution layer, with a deliberately narrow conclusion.
The system was built from zero to production volume in one month. The audit made the final conclusion more credible by refusing to turn incomplete evidence into a marketing claim.
- Success: the programme produced 352 articles and a working multi-platform publishing layer.
- Success: distinct platform versions, reliability controls, live checks, and reconciliation improved operational confidence.
- Partial success: the available evidence did not prove human traffic, backlinks, or AI citations.
- Lesson: database counts are not enough; reconcile internal state with public feeds and live checks.
- Follow-up: improve instrumentation, continue monthly AI-readability checks, and keep source coverage visible in every report.
Sanitized for public review
The public version removes client names, private records, credentials, internal URLs, production IDs, and sensitive business rules. It preserves the project thinking, delivery decisions, controls, and measurable outcomes a recruiter or hiring manager needs to evaluate.