Lucas Lim
Kai Yi

Branded search converts well but only captures people who already know the brand. Growth comes from non-brand, occasion-led search (anniversary dinners, gifting, corporate hosting, location terms) served by dedicated landing pages, indexable HTML menus and structured schema, measured on non-brand share of organic traffic and first-time customer acquisition.

When aggregators outrank a brand's own pages, the booking journey and customer data leave its ecosystem. Claimed and verified Google Business Profiles, consistent listings, weekly posts and SOP-driven review responses improve map-pack visibility, protect direct bookings and turn reviews into a weekly sentiment readout for operations.

I completed my marketing degree in July 2025 while working full-time; when the MMD role ended that October I took four months out, building a local LLM command centre, this site and Gmail/Calendar automations on Google Cloud and Azure, while interviewing. Then I chose scope over title: Burnt Ends is group-level, multi-brand, and a far larger estate than two monitor brands in one market.

Answer and generative engine optimisation means feeding AI systems accurate, structured data: Restaurant, Menu, FAQ and Review schema, consistent entity data across listings, and content that answers real customer questions directly. Test by asking the engines what customers ask, fix source data where answers go wrong, and track AI-referral traffic in GA4.

It means the reporting, sentiment analysis and first drafts are AI-assisted, built with Claude and Windsor.ai, but every output is checked against source data before it reaches anyone. It took my monthly leadership report from at least two weeks to two to four days, by my estimate; it doesn't replace the judgement of someone who has run the channels it's reporting on.

An in-house operator who runs paid, organic, SEO and reporting themselves catches problems agencies bill by the hour to investigate, and can react same-day. Running SEM, paid social and SEO internally rather than outsourcing has measurably reduced reliance on external agencies while keeping full visibility on what's actually working.

Both, but the building is the differentiator. A local-first marketing OS assembling every group brand's performance data and running 25 playbooks on a schedule, an engine that audits and generates sites against a versioned 51-rule SEO/GEO standard, an audit that scores whether a brand should run paid media at all, a WooCommerce scheduling plugin running in a live restaurant-group checkout, an on-device document-retrieval tool, and this site itself were all built directly, not briefed out. Each one has a write-up under Builds explaining how it works.

AI writes a lot of the code; the specification, the architecture and the rules are mine, and the rules are the part that matters. Each build has a written spec with locked decisions, a deviations log recording every place reality differed from that spec, and constraints the code enforces on itself: never display a figure that can't be sourced, never present cached data as live, never let an automated step grant itself permission. Anyone can generate code. Knowing what to demand of it, and being able to say why each decision was made, is the actual skill.

Because the reporting layer is where marketing claims get made, and buying that layer means trusting someone else's definitions. Building it means knowing exactly what a number is made of: which sources it came from, what window it covers, and when it was last refreshed. It also removes the recurring assembly work: the same exports, the same joins, every week. That time goes back into campaigns.