Publisher and scope
Eddie Cheng writes about profit-first ecommerce growth from both sides of the work: Penang Media, the performance agency he founded, and VIBAe, the footwear brand he co-owns.
The blog covers profitable ecommerce growth, paid acquisition, brand operations, measurement and practical AI. Some articles link to Penang Media, the agency I founded. A commercial link never changes the evidence or conclusion.
Topic selection and research
Launch topics come from live search and community evidence, not a list of opinions supplied by the business. Candidate questions are scored for the weakness of current results, demand, business relevance and citation potential. The editor can change the order or reject a topic, but a topic without evidence cannot enter the calendar.
Each article starts with a brief that records the sources, claim notes, search intent and one distinct angle. Non-obvious factual claims must map to a source or to documented first-party input.
First-party experience
Every article needs something public sources cannot supply: an operator insight, an anonymised example or approved benchmark data. The review record shows where that contribution appears and how it was used.
We may clean a spoken answer for grammar and filler. We do not add claims, remove meaningful hesitation or turn a qualified statement into a certainty.
How we use AI
AI can assist with research retrieval, drafting and language checks. It does not choose the final topic, supply undocumented facts or approve publication. Every article passes automated checks and a human review.
Benchmarks and client privacy
Benchmark reports use aggregated results. A figure cannot be published from fewer than eight accounts, and the approval screen blocks combinations that could identify a client. Reports prefer medians and percentile bands, round figures and use data from a fully closed month.
Corrections and updates
Material edits create a revision and update the visible modified date. A refresh goes through the same review path as a new article. If an article has become redundant or risks competing with a stronger page, the editor can consolidate it instead of adding another version.