Recommendations

CRM: HubSpot, already in place. The task is configuration for a considered, relationship-led sale. Three things earn their place. Enquiry-source capture on every deal, through UTM capture and first-touch source properties, which closes the gap that left the earlier spend unreadable (Section 13). A referral source category that records who referred each enquiry, so referral contribution becomes visible and the programme in Section 12.11 can be measured. And custom properties for the charter specifics, grounds, vessel, dates, party size and budget, that carry the long nurture and brief the brokers who handle the phone-first enquiry. The pipeline stages map the long consideration journey to a booking and on to repeat charter, and lead scoring reads for fit with the high-value audience.

Website: Claude Code on Sanity, built and run in-house. The recommended stack is Sanity as the headless CMS, a coded front end built with v0 and Claude Code, the repository held in GitHub, and Vercel for deployment, where every push goes live automatically. v0 generates the interface components fast, on the shadcn component library, and Claude Code assembles them, wires them to the Sanity schema, and carries the ongoing development. The point of this stack is independence. The team develops and extends the site itself against the Sanity schema, which removes the reliance on an external web agency and the cost and delay that come with it. The design direction already maps its sections to Sanity schema types, so the build path is short. The alternative, a visual builder such as Framer, is easier to use and holds no repository for the team to build on, so it is set aside to keep the in-house capability. A booking and channel-management system sits alongside this as a scoping task, since the owned vessels listed on competitor sites need rate parity held between a direct booking and the same boat sold through a co-broker.

The AI foundation: Claude, on an enterprise plan. The strategy leans on AI at several points. The site is built and extended with Claude Code, the content estate is produced at volume against the brand’s voice, and the answer engines are a channel in their own right. A single AI platform underneath these gives the team one place to build its internal processes: the brand voice and copy standards held as shared projects, content drafted and checked against them, charter briefs and CRM notes summarised for the brokers, and the site developed in the same environment. Claude Enterprise is the recommended base, since the build capability in this stack already runs on Claude Code and consolidation keeps the processes in one system. The lighter Team plan is the sensible entry, moving to Enterprise when security review, wider rollout or higher usage warrants it. The equivalent platforms from OpenAI and Google are credible alternatives, and the deciding factor is the build stack already chosen.

AEO tooling.One capability comes with a tool already in place: HubSpot carries a native AEO tool, so measurement of the AI-citation proxy can start there. Ahrefs with Brand Radar is the recommended addition, tracking AI visibility and citations alongside the search estate it already baselines. For deeper, prompt-level tracking of which questions surface the brand against competitors, a dedicated tool earns a pilot, with Eightlab, Profound and Otterly the current options. The category’s AI estate is near-empty (Part One), so this tracking is cheap leverage taken early.

PR tracking.The PR programme (Sections 12.4 and 12.13) is recommended to run through an agency, and the agency brings its own database and monitoring. What stays in-house is visibility of the output. Coverage lands in a CoverageBook report or an equivalent, links and referring domains are read in Ahrefs, and the operational measures in Section 13.5 count placements and links from those two sources, so the programme is measured on tools the team holds rather than on the agency’s word. If the programme later comes in-house, Telum Media, strong across Australian and Asia-Pacific newsrooms, or Muck Rack is the media database to pitch from, and SourceBottle carries inbound journalist requests at low cost.

Buying the high-attention media. The high-attention channels are bought, not installed, and the recommendation is one demand-side platform seat rather than several. StackAdapt is the recommended seat. It buys online video, connected TV, programmatic audio across streaming services and podcast inventory, and the curated display placements the Selective programmatic role calls for, in one platform. It opens self-serve, and it becomes a managed service once a minimum spend is met, so at the spend this strategy reaches the platform’s own team carries campaign execution while the in-house team keeps the seat, the data and the ownership (Section 14.2). Quantcast is the credible alternative, and its channel set is narrower, so it is held as the alternative. Three buys sit outside the seat by necessity. YouTube is bought through Google Ads, since its inventory does not trade on third-party platforms. Host-read podcast sponsorships (Section 12.2) are negotiated directly with the networks that hold the shows, LiSTNR, ARN’s iHeart and Acast among them, since a host’s voice cannot be bought programmatically. And cinema (Section 12.3) is bought through Val Morgan, which represents the major and independent exhibitors, placed by catchment and by film. The attention metrics that Section 13 reads as a quality check are asked of these partners at the buy, read against the research of Amplified Intelligence and Lumen.

Measurement: a staged stack for brand lift and incrementality. The stack grows in three stages, matched to the data available at each (Section 13).

Now, the leading indicators. Share of search read through Ahrefs Brand Radar and Google Trends as the brand-health proxy. Enquiry-source and referral attribution through HubSpot. Server-side tagging and UTM discipline feeding GA4, BigQuery and Looker Studio. And the attention metrics supplied by the media partners as a quality check on the high-attention buys.

Before the media, the creative. System1’s Test Your Ad reads the emotional response to a film before the media commits, scored for long-term brand-building potential against category norms. The strategy concentrates its creative risk into a small set of films that feed video, cinema and social at once (Section 12), so the films and their cinema and connected TV cuts are tested ahead of the phase-two launch, and the media is committed against creative that has earned it (Section 14.4). The panel is general population, so the scores are read as directional, and the testing is held to the films rather than every cut-down.

As the campaigns run, brand lift. Happydemics measures lift in flight through survey-based studies that run across the channels the strategy buys, video, audio and cinema among them, where the platforms’ own lift tools do not reach. Results are benchmarked against category norms, and they are read as directional given the bounded audience a survey panel can reach.

As history builds, incrementality. Geo and time-based holdout tests through Google Meridian GeoX, an open-source tool that holds media out in some regions and measures the lift it caused, or through Haus, run where the volume supports them. Contribution modelling across the mix is set aside at this scale, since a model run before the booking history exists returns noise, and the combination of lift studies and holdouts carries the question until the volume argues otherwise.

One consolidated platform for brand lift and incrementality is the natural question, and the honest answer is that no single tool covers both credibly at this data scale. Happydemics is the nearest consolidation on the brand side, one platform reading lift across every channel the strategy buys. Incrementality stays with the holdout tests, which measure behaviour rather than recall. The staged stack stands: creative tested before the buy, share of search and enquiry data now, lift read in flight as the campaigns run, and geo holdouts as volume allows.