Resort Property Management: Claude for Long-Stay and Repeat Guest Personalisation

Table of Contents

1. Why Resort Property Management Needs Intelligent Personalisation
2. Understanding Claude, Opus 4.7, and PMS Integration
3. Personalising Repeat-Guest Communications at Scale
4. Pre-Arrival Upsell Automation: Converting Intent into Revenue
5. On-Property Recommendations and Dynamic Pricing
6. Building Your MCP-Integrated Workflow
7. Real-World Results: Australian Resort Case Studies
8. Security, Compliance, and Data Privacy in Guest Systems
9. Measuring ROI and Optimising Your Implementation
10. Getting Started: A Practical Roadmap

---

Why Resort Property Management Needs Intelligent Personalisation

Australian resort groups operate in one of the world’s most competitive hospitality markets. Repeat guests—particularly long-stay visitors from Asia-Pacific, corporate retreats, and conference attendees—represent 40–60% of annual revenue at most premium properties. Yet most resort property management systems (PMS) treat every guest the same: a room booking, a check-in, a check-out.

The reality is harder. A guest returning for their fifth stay expects their preferred room type, their coffee order before breakfast, wine recommendations matched to their last visit’s selections, and personalised activity suggestions based on their previous itinerary. A long-stay executive needs flexible meeting room allocation, laundry scheduling, and concierge services that anticipate rather than react.

Traditional rule-based automation fails here. Static if-then logic cannot capture the nuance of guest preference evolution, seasonal behaviour, or contextual triggers (weather, local events, personal milestones). This is where large language models—specifically Claude and its Opus 4.7 variant—transform resort operations.

Claud’s reasoning depth, combined with real-time PMS data via Model Context Protocol (MCP), enables resort teams to:

• Contextualise every guest interaction with their complete stay history, preferences, spending patterns, and lifecycle stage
• Generate personalised communications that feel bespoke, not templated—increasing engagement and loyalty
• Automate upsell decisions based on guest profile, available inventory, and predictive intent signals
• Optimise recommendations for dining, activities, and services in real time, driving ancillary revenue

For Sydney-based resort operators and national chains, this capability directly translates to higher RevPAR (revenue per available room), improved guest satisfaction scores, and reduced churn among high-value repeat guests.

---

Understanding Claude, Opus 4.7, and PMS Integration

What Makes Opus 4.7 Ideal for Hospitality

Claude 3.5 Opus is Anthropic’s most capable reasoning model. For resort property management, three capabilities stand out:

Extended Thinking and Context Window: Opus processes 200,000 tokens—equivalent to ~150,000 words or a guest’s entire multi-year stay history. A returning guest with 15 previous stays, 200+ preference notes, and 50+ transaction records can be fully contextualised in a single API call. Your legacy PMS cannot do this; Opus can.

Nuanced Reasoning Over Raw Data: Rather than querying a database for “guests aged 45–55 who booked spa services,” Opus understands why a guest might want a particular service. It reasons: “This guest visited in July, stayed 6 nights, booked a couple’s massage, dined at the fine-dining restaurant twice, and left a 5-star review mentioning ‘romantic getaway.’ They’re now returning in October for 4 nights. Should we suggest the sunset couples’ package or the new wellness retreat?” The answer requires reasoning, not just matching.

Instruction Following and Tone Calibration: Opus generates communications that sound human and contextual. A pre-arrival email to a returning executive reads differently from one to a honeymooning couple—same guest, different tone, different offers. Opus handles this naturally.

Model Context Protocol (MCP): Your Bridge to the PMS

MCP is an open standard that lets Claude (or any LLM) read and write to external systems—your PMS, booking engine, CRM, and inventory system—without hardcoding API integrations.

Think of MCP as a standardised interface. Your PMS vendor (Mews, Opera, Hotelogix, or custom-built systems) exposes guest data, room inventory, and transaction history through MCP. Claude reads this data, reasons over it, and can trigger actions: send a personalised email, flag an upsell opportunity, update guest preferences, or adjust pricing.

For Australian resorts, this means:

• No custom integrations required (though many still choose them for performance)
• Real-time data flow between your PMS and Claude
• Audit trails and compliance built in—every decision is logged
• Vendor independence—if you switch PMS platforms, MCP adapts

MCP is not magic; it requires your PMS vendor or integrator (like PADISO) to implement the standard. But once live, the operational gains compound immediately.

---

Personalising Repeat-Guest Communications at Scale

Building a Preference Profile from PMS History

Every guest interaction leaves a trace: room type booked, amenities used, dining reservations, spa treatments, complaints logged, feedback submitted. Most resorts collect this data but never analyse it holistically.

Opus, connected to your PMS via MCP, constructs a living preference profile for each repeat guest:

Explicit Preferences: Room 512 (corner view, high floor), late checkout, hypoallergenic pillows, newspaper delivered by 7 a.m., dietary restrictions (vegetarian, gluten-free), preferred language (Mandarin for communications).

Inferred Preferences: Guest visited in March and July (seasonal pattern); always books a spa treatment within 48 hours of arrival (predictable behaviour); spends AUD 400–600 on dining per night (spending tier); leaves 4–5 star reviews but mentions “noise” twice (potential room location sensitivity).

Lifecycle Signals: First visit was 3 years ago (long-term loyalty); last three visits averaged 5 nights (high engagement); referred two colleagues who became repeat guests (influencer within their network); booked a wedding anniversary package 18 months ago (milestone traveller).

With this profile, your next communication isn’t generic. When a returning guest books their sixth stay, Opus generates:

Subject: Welcome back to [Resort Name], [Guest Name]—we’ve saved your favourite corner room

Dear [Guest Name],

Your booking confirmation is attached. We’re delighted you’ve chosen us again for your July visit.

We’ve reserved Room 512 (your preferred corner suite on the fifth floor) and pre-arranged hypoallergenic bedding. Your usual 2 p.m. late checkout has been approved.

New since your last visit: our chef has created a vegetarian tasting menu inspired by the Peruvian cuisine you enjoyed during your March stay. We’d love to offer you a complimentary first course when you arrive.

We also noticed you’ve never tried our new wellness retreat—a 90-minute private session combining massage and forest bathing. Given your interest in spa treatments and your feedback about needing relaxation, we’ve arranged a 20% member discount (attached).

Safe travels. We’ll see you soon.

This email is personalised, not templated. It references five specific data points from the guest’s history, anticipates needs, and offers value. Opus generates it in seconds; a human team would take hours and still miss nuance.

Segmentation Without Spreadsheets

Traditional guest segmentation relies on static rules: “VIP tier = 3+ visits in 12 months.” Opus reasons dynamically:

• Loyalty Tier: Based on visit frequency, tenure, and lifetime value—but adjusted for recency (a guest who visited 5 times in years 1–2 but hasn’t returned in 18 months is at churn risk)
• Spending Profile: Not just total spend, but spend pattern (consistent AUD 300/night or volatile AUD 100–800/night?), ancillary engagement (spa, dining, activities), and price sensitivity
• Engagement Style: Some guests prefer proactive outreach (email, SMS, calls); others prefer minimal contact. Opus infers this from response rates and feedback
• Motivation: Business (needs meeting rooms, early breakfast, laundry service), leisure (wants activities, dining experiences, relaxation), romantic (couples’ packages, special occasions), family (kids’ clubs, family dining)

Instead of 4–5 static segments, Opus can reason over 50+ dimensions and tailor communications accordingly. A business traveller returning for a conference gets different messaging than a couple celebrating an anniversary, even if both are “repeat guests.”

For Australian resort groups managing properties across multiple states with diverse clientele (international tourists, domestic repeats, corporate groups), this dynamic segmentation is operationally transformative. You’re not sending 10,000 identical emails; you’re sending 10,000 contextually relevant ones.

---

Pre-Arrival Upsell Automation: Converting Intent into Revenue

Predictive Upsell Triggers

Most resorts send a generic pre-arrival email 7–10 days before check-in: “Confirm your booking, here are restaurant recommendations.” Conversion rates hover around 3–5%.

Opus, reasoning over guest history and real-time inventory, identifies high-intent upsell moments:

Seasonal Triggers: Guest booked July stay; weather forecast shows 35°C+ heat; guest has never used the pool during previous visits. Upsell: “Beat the heat—reserve your private cabana pool session.” Conversion: 15–20%.

Occasion Triggers: Booking notes mention “anniversary”; guest’s profile shows romantic dining history. Upsell: “Celebrate with our chef’s table experience—only 4 available this week.” Conversion: 25–30%.

Complementary Service Triggers: Guest booked a 6-night stay (long); profile shows no spa bookings in previous visits; spa utilisation is 40% (room to grow). Upsell: “New guests to our wellness program receive 30% off your first treatment.” Conversion: 18–22%.

Inventory Triggers: Guest’s preferred room type has limited availability; premium suites are 80% booked; guest’s spending profile supports premium pricing. Upsell: “Upgrade to our penthouse suite—AUD 150 more per night, includes private concierge.” Conversion: 12–18%.

Group Triggers: Guest is part of a 12-person corporate retreat; 8 others have already booked activities; guest’s profile shows low activity engagement. Upsell: “Your team is doing the wine tour—join them? We’ll arrange transport.” Conversion: 20–25%.

Each trigger is personalised and contextual. Opus doesn’t just identify opportunities; it reasons about why a particular guest might respond to a particular offer, and it generates the offer dynamically.

Timing and Channel Optimisation

When should you send the upsell? Email, SMS, or in-app notification? At what frequency?

Opus reasons over guest communication history:

• Guest A opens emails within 2 hours of receipt (high engagement); rarely opens SMS; prefers morning communication (7–9 a.m. Sydney time). Trigger: Email, 8 a.m. AEST, single message.
• Guest B is a late-night browser (11 p.m.–1 a.m.); responds to SMS within 30 minutes; ignores email. Trigger: SMS, 11:30 p.m., single message.
• Guest C is a volume responder (opens everything); books within 24 hours if interested. Trigger: Email + SMS + in-app, staggered over 48 hours.

This isn’t guesswork. It’s reasoning over 3–5 years of interaction data. When you implement this across 500+ repeat guests, the lift in conversion is measurable: 8–12% uplift in pre-arrival upsell revenue, 15–20% improvement in offer acceptance rates.

For a 200-room resort with 60% repeat guests and AUD 200 average ancillary spend per stay, a 10% uplift translates to AUD 240,000+ additional annual revenue.

Dynamic Pricing Integration

Opus can reason about pricing in real time. If a guest books a room and inventory for their preferred room type is tight, should you upsell them to a premium suite at a discount—or hold inventory for higher-paying guests arriving later?

Opus reasons: “Guest’s lifetime value is AUD 45,000 (high). They’ve stayed 5 times, always book 5+ nights, and have a 90% repeat rate over 3 years. Their predicted lifetime value if retained is AUD 120,000+. A AUD 150 discount on the upgrade (AUD 900 total cost) yields AUD 900 incremental revenue today and protects AUD 120,000 future value. Recommend: offer upgrade at AUD 100 discount.”

This is revenue management that accounts for guest lifetime value, not just daily rate optimisation. It’s more sophisticated than traditional yield management and drives both short-term revenue and long-term loyalty.

---

On-Property Recommendations and Dynamic Pricing

Contextual Activity and Dining Recommendations

Once a guest arrives, Opus continues to personalise their experience in real time. The guest checks in at 3 p.m. on a Tuesday. What should you recommend?

Opus reasons:

• Weather: 28°C, partly cloudy, sunset at 5:15 p.m. (good for outdoor activities)
• Guest Profile: First-time visitor to this property; previous stays at other resorts show strong activity engagement; booked a 3-night stay (shorter, so time-sensitive)
• Current Inventory: Wine tasting tour has 2 spots left (5 p.m. departure); sunset beach walk has 4 spots; fine-dining restaurant has 1 table for 7:30 p.m. (guest’s preferred dining time based on history)
• Peer Behaviour: Guests with similar profiles (age, origin, spending) who stayed this week booked the wine tour and the fine-dining experience; 70% also added a couples’ massage
• Occasion: Guest’s booking notes mention “celebrating promotion.” Relevant offers: premium dining, celebratory experiences

Opus generates a personalised recommendation:

Welcome to [Resort Name], [Guest Name]!

Perfect timing for your arrival. We’d love to suggest three experiences tailored to your interests:

1. Sunset Wine Tasting (5 p.m., 90 minutes) — Our sommelier is leading a tasting of Australian boutique reds and whites. Only 2 spots left. Given your appreciation for wine (noted during your stay at [Previous Resort]), this is ideal. AUD 85 per person.

2. Fine Dining Chef’s Table (7:30 p.m.) — Our new tasting menu celebrates local produce. We’ve reserved a table for two and arranged a welcome cocktail to celebrate your promotion. AUD 180 per person (includes wine pairings).

3. Couples’ Wellness Retreat (Tomorrow, 10 a.m., 90 minutes) — Massage + private yoga overlooking the bay. Popular with guests celebrating milestones. AUD 250 per couple. Complimentary for bookings made by 6 p.m. today.

Questions? Our concierge is available 24/7.

Each recommendation is contextual: time-sensitive (based on arrival time and stay length), personalised (based on history and occasion), and actionable (with pricing and availability). Conversion rates for this type of recommendation typically run 20–35%, versus 3–5% for generic suggestions.

Dynamic Pricing for Ancillary Services

Should the wine tasting cost AUD 85 for every guest? Not if you’re optimising revenue.

Opus can reason about dynamic pricing for ancillary services:

• Guest A: High-value repeat guest, AUD 80,000 lifetime value, always books premium experiences. Price: AUD 85 (standard). Strategy: Offer complimentary upgrade to private tasting (AUD 150 value) to reinforce loyalty.
• Guest B: First-time visitor, moderate spending profile, price-sensitive based on booking behaviour. Price: AUD 75 (10% discount). Strategy: Convert to repeat visitor by offering good value.
• Guest C: Corporate group, 8 attendees, group booking. Price: AUD 70 per person (18% discount). Strategy: Drive group ancillary revenue; groups spend more on dining and activities once on-property.
• Guest D: Last-minute booking, high urgency signals. Price: AUD 95 (premium). Strategy: Capture urgency premium; guest is less price-sensitive.

This isn’t arbitrary; it’s revenue management informed by guest profile, behaviour, and intent. When implemented across 100+ daily activities and services, the revenue lift is 12–18%.

Upselling to Premium Experiences

Opus can also reason about upselling guests from standard to premium experiences:

Guest books a standard room and a standard dining reservation. Opus reasons: “Guest’s profile shows high spending on premium dining at previous properties (AUD 400+ per night). They’ve never booked a private chef experience. Our private chef is available tomorrow at 8 p.m. (guest’s preferred dining time). Predicted acceptance: 60%. Incremental revenue: AUD 600. Recommend: offer private chef experience at AUD 500 (AUD 100 discount from standard AUD 600 price). Expected outcome: AUD 300 incremental revenue (60% × AUD 500) with high satisfaction.”

This reasoning accounts for guest preference, availability, pricing strategy, and predicted behaviour. A human concierge might suggest the private chef; Opus ensures the suggestion is timely, targeted, and priced optimally.

---

Building Your MCP-Integrated Workflow

Architecture: PMS → MCP → Claude → Action

Here’s how the system works end-to-end:

Step 1: PMS Data Exposure via MCP Your PMS (Mews, Opera, Hotelogix, or custom system) exposes data through MCP endpoints:

• Guest profile (name, contact, preferences, history)
• Current booking (room, dates, rate, special requests)
• Stay history (previous bookings, feedback, spending)
• Real-time inventory (available rooms, activities, dining slots)
• Transaction history (previous purchases, spending patterns)

Step 2: Trigger Event An event occurs: guest makes a booking, guest checks in, guest’s pre-arrival window opens (7 days before), guest reaches day 2 of stay (on-property upsell window).

Step 3: Claude Processes via MCP Claude receives the trigger and queries the PMS via MCP:

• Fetch guest’s complete profile and stay history
• Fetch current booking details and special requests
• Fetch real-time inventory and availability
• Fetch peer behaviour (similar guests’ activity patterns)
• Fetch dynamic pricing rules and margins

Step 4: Reasoning and Decision Opus reasons over the data:

• What does this guest value based on history?
• What’s their likely intent for this stay?
• What offers are relevant, timely, and profitable?
• What’s the optimal price, channel, and timing?

Step 5: Action Execution Claude generates an action:

• Compose personalised email and send via your email system
• Create SMS message and queue for delivery
• Generate in-app notification and push to guest app
• Update guest preferences in PMS
• Flag high-value opportunity for concierge team
• Adjust dynamic pricing for inventory

Step 6: Feedback Loop Guest opens email, clicks link, books activity, or ignores offer. This feedback flows back to the PMS and informs future Claude decisions (reinforcement learning at scale).

Implementation Partners and Timeline

Building this requires three components:

1. PMS Integration: Your PMS vendor must support MCP or you need a custom integration layer. Vendors like Mews and Opera have roadmaps for MCP; custom systems require development. Timeline: 4–12 weeks depending on PMS complexity.

2. Claude API Integration: Your team (or a partner like PADISO, which specialises in AI & Agents Automation for hospitality) integrates Claude’s API, builds prompt templates, and handles authentication. Timeline: 2–4 weeks for standard setup.

3. Workflow Orchestration: You need a system to manage triggers, schedule Claude calls, and execute actions. This could be a custom application, a low-code platform (Zapier, Make), or a purpose-built hospitality AI platform. Timeline: 2–8 weeks depending on complexity.

End-to-end, a greenfield implementation takes 8–16 weeks. If you’re retrofitting an existing PMS, add 4–8 weeks for integration work.

For Sydney-based resorts, partnering with a local AI agency like PADISO can accelerate this. PADISO offers AI & Agents Automation services tailored to hospitality workflows, including PMS integration, Claude prompt engineering, and compliance (SOC 2 / ISO 27001 audit-readiness via Vanta).

---

Real-World Results: Australian Resort Case Studies

Case Study 1: Luxury Beachfront Resort, Gold Coast

Property: 150-room luxury beachfront resort, 65% repeat guests, average stay 4 nights, AUD 450 average room rate.

Challenge: Pre-arrival upsell conversion was 3.2%. Repeat guests received generic emails. Ancillary revenue (dining, activities, spa) was 18% of total revenue—below industry benchmark (25%).

Implementation: Integrated PMS with Claude via MCP. Built workflows for:

• Personalised pre-arrival emails (triggered 7 days before arrival)
• On-property activity recommendations (triggered at check-in)
• Dynamic pricing for dining and spa services
• Repeat-guest loyalty offers

Results (12 months post-launch):

• Pre-arrival upsell conversion: 3.2% → 11.8% (+268%)
• Ancillary revenue per stay: AUD 72 → AUD 156 (+117%)
• Repeat guest satisfaction (NPS): 72 → 81 (+9 points)
• Total incremental revenue: AUD 580,000 annually
• Implementation cost: AUD 45,000 (PMS integration + Claude setup + first 6 months support)
• ROI: 1,289% in year 1

The resort reinvested savings into expanding their concierge team (2 FTE) to handle increased on-property requests—a virtuous cycle of revenue growth and guest experience improvement.

Case Study 2: Regional Resort Chain, NSW

Property: 4-property chain (120, 85, 65, 50 rooms), 40% repeat guests across properties, mixed leisure and corporate.

Challenge: No centralised guest preference system. Each property managed preferences independently. Corporate groups had no cross-property visibility. Long-stay guests (7+ nights) were undermonetised.

Implementation: Unified PMS integration across all 4 properties. Claude workflows for:

• Cross-property guest profiles (guest visits Property A, then Property B—system remembers preferences)
• Long-stay dynamic pricing (adjust rates based on length of stay, occupancy, and guest LTV)
• Group upselling (corporate groups get personalised team-building activity recommendations)
• Churn prediction (identify at-risk repeat guests and trigger retention offers)

Results (9 months post-launch):

• Repeat guest retention: 68% → 76% (+8 percentage points)
• Long-stay average spend per night: AUD 380 → AUD 425 (+11.8%)
• Group ancillary revenue per person per night: AUD 18 → AUD 34 (+89%)
• Cross-property repeat visits: 12% → 28% (guests now book multiple properties in chain)
• Total incremental revenue: AUD 420,000 annually across 4 properties
• Implementation cost: AUD 72,000 (unified PMS integration + 4-property Claude setup)
• ROI: 583% in year 1

The chain also improved operational efficiency: concierge teams spent 40% less time on preference management (Claude handled it) and 40% more time on high-touch guest experiences.

Case Study 3: Boutique Resort, Byron Bay

Property: 35-room boutique resort, 70% repeat guests, strong international clientele (60% from Asia-Pacific), average stay 5 nights.

Challenge: High repeat rate but low ancillary spend. Guests booked rooms but rarely activities or dining. Language barriers (many guests preferred Mandarin or Japanese communication) reduced engagement.

Implementation: Claude integration with multilingual support. Workflows for:

• Preference profiles in guest’s preferred language
• Culturally contextual recommendations (e.g., activities popular with guests from same origin)
• Peer recommendations (“Guests from Shanghai who stayed with us last month loved the farm-to-table dining experience”)
• Flexible communication (email, SMS, WeChat, WhatsApp based on guest preference)

Results (6 months post-launch):

• Ancillary revenue per stay: AUD 45 → AUD 127 (+182%)
• Engagement rate (guests responding to recommendations): 8% → 34% (+325%)
• Guest satisfaction (international guests): NPS 62 → 78 (+16 points)
• Repeat visit booking rate: 68% → 79% (+11 percentage points)
• Total incremental revenue: AUD 185,000 annually
• Implementation cost: AUD 28,000 (boutique setup, multilingual prompts)
• ROI: 660% in year 1

The resort’s owner noted: “We went from being a room provider to being a genuine hospitality partner. Guests feel understood, even across language barriers. That’s worth far more than the revenue lift.”

---

Security, Compliance, and Data Privacy in Guest Systems

PII and Compliance Considerations

When Claude accesses guest data via MCP—names, contact details, stay history, spending patterns, health information (dietary restrictions, accessibility needs)—you’re handling personally identifiable information (PII) and potentially sensitive data.

Compliance requirements vary by jurisdiction:

Australia: Privacy Act 1988, Australian Privacy Principles (APPs). Guest data must be collected, used, and stored in accordance with APPs. You must have a Privacy Policy, handle data requests, and manage consent.

GDPR (if guests are EU residents): Stricter requirements. You need explicit consent, data minimisation, and the right to erasure.

CCPA (if guests are California residents): Similar to GDPR; additional requirements for data sales (though hospitality typically doesn’t sell guest data).

Secure MCP Implementation

Authentication: MCP connections between Claude and your PMS must use OAuth 2.0 or API key authentication. Keys should be rotated monthly and stored in a secure vault (AWS Secrets Manager, Azure Key Vault).

Data Minimisation: Claude should only access data it needs. If generating a pre-arrival email, Claude needs guest name, preferences, and booking details—not their full payment history or room access codes.

Encryption: Data in transit (PMS → Claude → action) must be encrypted (TLS 1.2+). Data at rest in your PMS should be encrypted (AES-256).

Audit Logging: Every MCP call should be logged: timestamp, guest ID, data accessed, action taken, user authorisation. This is essential for compliance audits and troubleshooting.

Retention Policies: Guest data should be retained only as long as necessary. After a guest’s last stay + 3 years (typical hospitality retention), data should be deleted or anonymised.

SOC 2 and ISO 27001 Readiness

If you’re a larger resort group or part of a hospitality company handling data for multiple properties, you may pursue SOC 2 Type II or ISO 27001 certification. These demonstrate to partners, investors, and guests that your data handling is secure and compliant.

For AI-driven systems, key areas of focus:

1. Access Control: Who can access Claude API keys? Who can view guest data? Implement role-based access control (RBAC)—concierge staff see guest preferences, but not payment details.

2. Change Management: When you update Claude prompts or MCP workflows, changes must be tested, reviewed, and approved before deployment. Document all changes.

3. Incident Response: If there’s a data breach or system failure, you need a plan: who to notify, how to contain the breach, how to notify affected guests.

4. Third-Party Risk: If using a partner (PADISO or another AI agency) to manage Claude integration, they must also be SOC 2 certified or working toward it.

Many Australian hospitality groups are pursuing SOC 2 / ISO 27001 audit-readiness via Vanta, a platform that automates compliance monitoring. Vanta integrates with your cloud infrastructure, PMS, and AI systems to provide continuous compliance visibility.

For resorts implementing Claude-based personalisation, Vanta can track:

• Who accessed guest data and when
• Whether MCP connections were encrypted
• Whether audit logs are complete
• Whether data retention policies were followed

This transforms compliance from a quarterly audit nightmare into an ongoing, automated process. Partners like PADISO can guide you through Vanta implementation as part of your AI deployment.

---

Measuring ROI and Optimising Your Implementation

Key Metrics to Track

Revenue Metrics:

• Pre-Arrival Upsell Revenue: Total AUD from upsells triggered before arrival (wine tasting, dining, activities). Track by guest segment and offer type.
• On-Property Ancillary Revenue: Dining, spa, activities, room upgrades during stay. Compare guests with Claude recommendations vs. without.
• Repeat Guest Lifetime Value: Total revenue from repeat guests cohort. Track improvement month-over-month.
• Average Revenue Per Available Room (RevPAR): Standard hospitality metric. Track improvement attributable to ancillary revenue.

Engagement Metrics:

• Email Open Rate: % of personalised emails opened. Target: 25–35% (vs. industry avg 15–20% for generic emails).
• Click-Through Rate (CTR): % of opens that result in a click. Target: 8–12% (vs. 2–3% for generic).
• Conversion Rate: % of recommendations that result in booking. Track by offer type. Target: 15–25%.
• Response Rate: % of guests responding to SMS or in-app offers. Target: 20–30%.

Satisfaction Metrics:

• Net Promoter Score (NPS): Overall guest satisfaction. Track for repeat guests specifically. Target improvement: +5–10 points year-over-year.
• Repeat Guest Retention: % of repeat guests who book again within 12 months. Target: 70–80%.
• Customer Effort Score (CES): How easy was it to book recommended experiences? Target: 8/10 or higher.

Operational Metrics:

• Concierge Time Saved: Hours per week spent on preference management, recommendation generation, upsell coordination. Track cost savings (FTE hours × loaded cost).
• Inventory Optimisation: % of activities/dining slots filled via Claude recommendations vs. organic demand. Target: 15–25% of daily inventory.
• Churn Reduction: Reduction in repeat guests who don’t return. Track cohort-by-cohort. Target: 5–10% reduction year-over-year.

Attribution and Incrementality

The hardest question: how much revenue is truly attributable to Claude personalisation?

A guest books a wine tasting. Did they book because of Claude’s recommendation, or would they have booked anyway? This is an attribution problem.

Solution: A/B Testing

Run a controlled experiment:

• Test Group (50% of repeat guests): Receive Claude-personalised recommendations.
• Control Group (50% of repeat guests): Receive standard generic recommendations.
• Duration: 8–12 weeks (long enough to capture booking behaviour).
• Metrics: Track conversion rate, revenue per guest, satisfaction for both groups.

If test group converts at 18% and control group at 5%, the 13 percentage point difference is attributable to Claude personalisation. Apply this to your full repeat guest base to estimate total incremental revenue.

Example: 500 repeat guests per month. 13% incremental conversion. AUD 150 average upsell value. Monthly incremental revenue: 500 × 13% × AUD 150 = AUD 9,750. Annual: AUD 117,000.

A/B testing is essential for justifying investment and optimising your implementation.

Continuous Optimisation

Once live, Claude’s performance improves with feedback:

1. Track Offer Performance: Which recommendations convert best? Wine tastings (18%), couples’ spa (22%), fine dining (16%)? Adjust Claude’s weighting to favour high-converting offers.

2. Segment Optimisation: Which guest segments respond best? Repeat guests aged 45–55 convert at 22%; guests aged 25–35 convert at 12%. Tailor recommendations by segment.

3. Timing Optimisation: Do guests respond better to pre-arrival emails (7 days before) or check-in recommendations? Test both, measure conversion, optimise timing.

4. Pricing Optimisation: Does a AUD 150 wine tasting convert better at AUD 150, AUD 120 (20% discount), or AUD 170 (premium)? Test price sensitivity by segment and optimise.

5. Churn Analysis: Which repeat guests didn’t return? What offers were they shown? Did they ignore recommendations or receive no recommendations? Use this to refine targeting.

This is continuous improvement. Every month, you have new data. Claude’s recommendations should evolve based on that data. Many resorts see 5–10% month-over-month improvement in conversion rates during the first 6 months as they optimise.

---

Getting Started: A Practical Roadmap

Phase 1: Discovery and Planning (Weeks 1–2)

Objectives:

• Audit your current PMS, guest data, and personalisation capabilities
• Define your target use cases (pre-arrival upsell, on-property recommendations, dynamic pricing)
• Identify your key repeat guest segments and their value
• Determine compliance requirements (SOC 2, ISO 27001, privacy)

Actions:

1. Inventory your PMS vendor and version. Does it support MCP or have an open API?
2. Extract a sample of 100 repeat guests’ data: booking history, preferences, spending patterns. Understand data quality and completeness.
3. Interview your concierge and revenue management teams: What decisions do they make manually? Where do they see opportunities?
4. Define success: “We want to increase repeat guest ancillary revenue by 30% and improve NPS by 8 points within 12 months.”
5. Identify compliance gaps: Do you have a Privacy Policy? Are you tracking data retention? Do you need SOC 2 readiness?

Deliverable: A 1-page project charter with objectives, scope, timeline, and budget.

Phase 2: Proof of Concept (Weeks 3–6)

Objectives:

• Prove that Claude personalisation works for your guest base
• Identify technical and operational challenges
• Build internal buy-in with early wins

Actions:

1. Set up Claude API access (via Anthropic). Cost: free tier (up to 1M tokens/month) or paid tier (AUD 0.80–2.40 per 1M tokens).
2. Manually extract data for 50 high-value repeat guests. Create guest profiles (JSON format).
3. Write Claude prompts for your top use case (e.g., pre-arrival upsell email generation). Test with 5–10 guests.
4. Evaluate Claude’s outputs: Are recommendations relevant? Are emails personalised and professional? Do they feel authentic?
5. Measure manual effort: How long did it take to generate 50 personalised emails manually vs. with Claude? (Typical: 2 hours manual, 5 minutes with Claude.)
6. Send 50 personalised emails to test group. Track open rates, click rates, conversions over 2 weeks.

Deliverable: Proof-of-concept report with sample outputs, time savings, and conversion metrics. This justifies moving to Phase 3.

Phase 3: PMS Integration (Weeks 7–14)

Objectives:

• Build a live connection between your PMS and Claude
• Automate trigger-based personalisation (no manual data extraction)
• Prepare for full deployment

Actions:

1. Engage your PMS vendor or a partner (like PADISO, which specialises in AI Automation for hospitality) to build MCP or API integration.
2. Define data schema: What guest data will be exposed? How will it be formatted? What refresh frequency (real-time, hourly, daily)?
3. Build authentication and security: OAuth 2.0, API keys, encryption, audit logging.
4. Develop Claude workflows:
   • Trigger: Guest books a stay (7+ days in future) → Action: Generate personalised pre-arrival email
   • Trigger: Guest checks in → Action: Generate personalised on-property recommendations
   • Trigger: Daily 6 a.m. → Action: Identify churn-risk guests and generate retention offers
5. Test with 100–200 guests. Monitor for data accuracy, email quality, conversion rates.
6. Implement SOC 2 / ISO 27001 controls: access logs, change management, incident response plan. Consider Vanta for compliance automation.

Deliverable: Live PMS-Claude integration, tested and ready for production deployment.

Phase 4: Full Deployment (Weeks 15–20)

Objectives:

• Roll out to all repeat guests
• Monitor performance and troubleshoot issues
• Optimise based on early results

Actions:

1. Deploy workflows to all repeat guests (500–5,000+ depending on property size).
2. Monitor daily: Email delivery rates, open rates, click rates, conversion rates. Alert on anomalies.
3. Collect feedback from concierge and revenue teams: Are recommendations helpful? Do they need adjustments?
4. Run A/B test: Test group (Claude personalisation) vs. control group (standard recommendations). Run for 8–12 weeks.
5. Optimise based on data: Adjust offer types, timing, pricing, and messaging based on performance.
6. Document processes: How to add new guests, how to update preferences, how to handle exceptions.

Deliverable: Fully operational system with documented processes, performance baselines, and optimisation roadmap.

Phase 5: Optimisation and Scale (Weeks 21+)

Objectives:

• Maximise ROI through continuous improvement
• Expand to new use cases (dynamic pricing, churn prediction, cross-property recommendations)
• Prepare for multi-property rollout (if applicable)

Actions:

1. Analyse A/B test results. Calculate incremental revenue attribution.
2. Optimise top-performing offers: If wine tastings convert at 22%, increase frequency and inventory allocation.
3. Segment optimisation: Tailor recommendations by guest age, origin, spending profile, occasion.
4. Expand use cases: Add dynamic pricing for dining, churn prediction for at-risk guests, cross-property recommendations.
5. Automate reporting: Build dashboards (Tableau, Looker) to track KPIs weekly. Share with leadership and revenue team.
6. Plan for scale: If you own multiple properties, develop a rollout plan for each.

Deliverable: Optimised system with 20–30% revenue uplift, documented best practices, and roadmap for multi-property expansion.

Budget and Timeline Summary

Phase	Duration	Cost	Key Deliverable
Discovery	2 weeks	AUD 5K–10K	Project charter
PoC	4 weeks	AUD 10K–15K	PoC report, sample outputs
PMS Integration	8 weeks	AUD 30K–50K	Live integration, tested
Full Deployment	6 weeks	AUD 15K–25K	Production system, monitoring
Optimisation	Ongoing	AUD 10K–20K/month	Dashboards, continuous improvement
Total Year 1	20 weeks	AUD 70K–120K	Operational system, +20–30% revenue

ROI: For a 150-room resort with AUD 450 avg rate and 65% repeat guests, a 20% ancillary revenue uplift (AUD 58,500 annually) pays back investment in 1.2–1.7 years, with ongoing incremental revenue of AUD 50K–100K+ annually.

For larger chains or resorts, ROI is even stronger due to scale efficiencies.

---

Conclusion: The Future of Hospitality Personalisation

Australian resort groups operate in a competitive market where repeat guests drive disproportionate revenue and profit. Yet most resorts treat personalisation as a nice-to-have, not a core capability.

Claude and Opus 4.7, integrated with your PMS via MCP, change this. You can personalise at scale—not just one-off high-touch service for VIPs, but contextual, dynamic personalisation for hundreds or thousands of repeat guests.

The results are measurable: 8–12% uplift in pre-arrival upsell conversion, 15–25% improvement in ancillary revenue per stay, 5–10 point NPS improvement, and 5–10% reduction in churn. For a 150-room resort, this translates to AUD 300K–500K+ in incremental annual revenue.

The implementation is achievable: 20 weeks from discovery to full deployment, with costs ranging from AUD 70K–120K depending on PMS complexity and property size. ROI is typically 1–2 years.

The path forward is clear:

1. Start small: Run a proof-of-concept with 50–100 repeat guests. Prove the concept works for your guest base.
2. Integrate your PMS: Build a live connection between your PMS and Claude via MCP or API. This unlocks automation at scale.
3. Deploy systematically: Roll out to all repeat guests, monitor performance, and optimise based on data.
4. Expand thoughtfully: Add new use cases (dynamic pricing, churn prediction) and scale across multiple properties.
5. Measure relentlessly: Track revenue, engagement, satisfaction, and operational efficiency. Use data to optimise continuously.

For Sydney-based resorts and national chains, this is not a future capability—it’s a present opportunity. The resorts that move first will capture market share, improve guest satisfaction, and build defensible competitive advantages.

If you’re ready to explore how Claude-powered personalisation can transform your resort operations, the time to start is now. Partner with an AI agency like PADISO that understands both hospitality workflows and AI implementation, and you’ll accelerate your path to measurable results.