Guest Memory for Property Rentals and OTAs

A guest should benefit from what the team learns while the stay is still under way. On a return visit, relevant preferences should already help the team prepare. A property manager should be able to carry a service commitment from the booking desk to the local team. A resident should be able to report an unresolved problem without reconstructing its history for every person involved.

GuestMemoryOS puts that continuity into daily hospitality service: preserve relevant knowledge, establish what is confirmed, identify who can act on it, and carry the outcome into the next interaction. Extending this working product into other rental settings creates different opportunities. A marketplace earns a booking fee. A manager retains an owner relationship. A letting agent earns a service fee. A landlord protects occupied days and avoids some turnover expenditure.

These eight studies connect those mechanisms to public operating evidence and transparent financial models. Each can be published as an individual opportunity study with its linked sources and model qualifications intact.

How to read the evidence

Verified benchmark means a figure or statement supported by an identified public source. The underlying businesses are recognisable through their published operating characteristics; their names are omitted from the narrative. Source links lead to the original, identifiable issuers.

Model assumption means an explicit input selected to show how the economics work. Cohort sizes, conversion changes, fees and cost assumptions are hypothetical unless specifically labelled as reported or derived from reported figures.

Model output means arithmetic conditional on those inputs. It describes potential value, before the remaining cost of implementing and operating GuestMemoryOS. It is neither a forecast nor an observed result for the benchmark business.

Pilot target means a proposed improvement to test against existing service. It does not describe the benchmark company's current service quality.

The commercial proposition

Use remembered and confirmed needs to improve matching, fulfilment and continuity. Measure whether that produces additional completed business or a genuine reduction in expenditure. Charge the programme with its full costs, including integration, staff effort and ongoing oversight.

These are commercial opportunity studies. No benchmark company is represented as an AFG customer, deployment partner or endorser.

What the evidence establishes

Guest history can improve booking outcomes

A major home accommodation marketplace reports online A/B experiments with a system representing past trips and recent guest behaviour. Long-term history alone increased uncancelled bookers by 0.31%. The combined long-term and short-term treatment delivered 0.55% more uncancelled bookers and 0.82% more uncancelled nights. These are relative changes, reported as statistically significant for the combined treatment. They are different outcomes and must not be added together. E1 Experimental results

The relevant conclusion is specific: guest history improved measurable accommodation booking outcomes in that operator's experiments. The implementation was a search recommendation system. Its results establish a credible mechanism for testing AFG's service memory proposition; they do not establish that AFG will reproduce those effects.

Shared knowledge can improve service productivity

A study of the staggered introduction of an AI assistant across 5,172 support agents found an average 15% improvement in issues resolved per hour. Effects varied substantially; the most experienced agents had small speed gains and small quality declines. E2 Research paper revised November 2024

That evidence supports testing whether accessible knowledge helps staff resolve unfamiliar requests. It does not justify applying 15% to a rental company's payroll. The studies below calculate specific tasks and include the effort required to review and maintain memory.

AFG must demonstrate its incremental value

GuestMemoryOS helps hospitality staff turn guest knowledge into preparation, relevant suggestions and service actions, then preserve useful learning for the next interaction. It was developed through live operation at Mandarin Beach Villa and is designed to work alongside existing CRM, property management, reservation and messaging systems through progressive integration. AFG's live use, internal cruise engineering simulations and residence review record are distinct from the controlled OTA conversion gains, residential retention gains and financial returns modelled here. A1 Product overview and A2 Current evidence and scope

The practical test is whether confirmed service knowledge adds value on top of an operator's existing customer records, recommendation systems and staff processes. Established operators already invest in these capabilities.

The chain that must be measured

A relevant memory becomes available. A staff member or authorised workflow acts on it. The guest or resident receives a better outcome. Completed business or avoidable expenditure then changes against a comparable control group.

Measure each link. More notes, more suggestions or faster replies alone cannot establish a financial return. Evidence is strongest for the search and service mechanisms; the retention and property operating effects remain hypotheses until tested in their own settings.

Eight models at a glance

Central scenarios | 12-month evaluation horizon | all improvements are assumptions

“Value” below is incremental contribution plus modelled cash savings, after any operating costs explicitly included in the case, but before the remaining programme cost. Different cohort sizes and currencies make these examples unsuitable for ranking businesses by attractiveness. They are independent scenarios and must not be summed.

Who receives the money

For studies 01 to 06, an intermediary generally earns only its own fees. Guest spending on accommodation is not interchangeable with intermediary revenue. For study 07, value is calculated at the property operating level and must be allocated under the owner's operating agreement. Study 08 concerns the landlord's own rent and operating expenditure.

Staff capacity is valued separately from cash. Where a case assumes 25% cash realisation, only a verified reduction in overtime, agency cover or contracted support would qualify. Without that expenditure reduction, remove the cash saving from the result.

The final sections show illustrative programme budgets and break-even limits, a validation approach and the evidence register.

01 Global accommodation OTA

Make the next stay easier to choose

A guest returns with a useful history: a preferred neighbourhood, a successful room configuration and a service requirement that was previously confirmed. Carrying that context into the next search could reduce the work needed to find a suitable property and complete another booking.

Verified benchmark. A listed global travel group reported 1.235 billion room nights, US$186.1 billion gross bookings and US$26.917 billion revenue for 2025. These are group figures covering several travel businesses; gross bookings also include services beyond accommodation. They establish scale and the distinction between transaction value and company revenue. O1 Annual reporting package, financial review

The group's major European accommodation platform is the recognisable reference for this study. Its public figures do not disclose the eligible returning-guest session volume or conversion rate assumed below.

Proposed Guest Memory workflow

A returning customer can confirm which previous preferences remain relevant. Search uses that context alongside current dates, price and verified property information. A service adviser sees the same confirmed requirements when the guest asks for help. The chosen property receives only the information needed to fulfil the request, subject to the appropriate permissions.

An old preference remains editable. A remembered requirement never substitutes for checking whether a particular property can meet it. A successful previous stay is a useful matching signal, not a guarantee that the same property suits every future trip.

Service improvement to test

For eligible enquiries, assume handling time falls from 15 to 13.5 minutes, with a further 0.5 minute required to review or maintain memory. That gives one minute of net staff time released per enquiry. Test resolution quality, repeated questions and booking completion alongside speed.

The search experiment in E1 supports the possibility of a small conversion improvement in this sector. The AFG treatment would still need to outperform the OTA's existing personalised search and service processes. E1 Evidence

The repeat booking outcome

The model concerns returning customers. Additional completed bookings therefore represent incremental repeat business within the tested session population. Repeat rate per unique guest should also be tracked, because one person may generate multiple search sessions.

Scope. This is an anonymised opportunity study based on public scale data. The workflow, service baseline and uplift are proposed, and no AFG deployment is asserted.

01 Financial model

Central assumptions and calculation

Assume 100 million eligible returning-customer sessions over 12 months, with 4.00% producing a completed, uncancelled accommodation booking. A 0.5% relative improvement changes that rate to 4.02%, an increase of 0.02 percentage points.

100,000,000 × 4.00% × 0.5% = 20,000 additional completed bookings.

Assume an average commissionable accommodation amount of US$750 per booking, excluding taxes and pass-through charges, and a net OTA fee of 15%. These are scenario assumptions, not a fee schedule inferred from group reporting.

20,000 × US$750 × 15% = US$2,250,000 incremental OTA revenue.

At an assumed 60% incremental contribution margin, after payment costs, additional acquisition expense, booking support and relevant incentives, the extra business contributes US$1,350,000 before programme costs.

Service capacity and cash

For 1 million existing-booking enquiries, one net minute saved releases 16,667 hours. At an assumed US$30 per hour, this is approximately US$500,000 of capacity. If 25% becomes actual lower overtime or contracted support spend, cash savings are US$125,000. The enquiry population excludes support costs already charged to the new bookings above.

Central value = US$1,350,000 contribution + US$125,000 cash savings = US$1,475,000.

The table holds the service cash saving constant to isolate conversion sensitivity. These are chosen scenarios, not statistical confidence limits.

How to prove the result

Randomise at guest level before eligible sessions begin. Compare completed bookings and contribution per assigned guest after cancellations mature. Deduplicate sessions, retain the control group's existing personalisation, and measure displacement across the group's other brands. A shift between group brands earns only the net difference.

A short rise in search clicks is insufficient. The decision should use incremental completed business, service quality and full operating costs. At the central assumptions, the remaining annual programme cost must be below US$1,475,000 to produce a positive return; the limit falls to US$1,350,000 if staff capacity produces no cash reduction.

02 Home rental marketplace

Protect a booking before a mismatch becomes a cancellation

A home rental can involve several independent parties: a traveller, a host, local support and a marketplace. A confirmed need can lose meaning as it passes between them. Guest Memory could help maintain that context, identify an unresolved requirement before arrival and record who has accepted responsibility for resolving it.

Verified benchmark. A global host marketplace reported 533 million Nights and Seats Booked, US$91.273 billion gross booking value and US$12.241 billion revenue in 2025. The volume measure combines stay nights with seats for experiences and services, net of cancellations and alterations; it is not a count of rental reservations. Revenue is generally recognised at check-in. O2 Annual report, key business metrics and revenue

These distinctions matter. A cancelled reservation may be replaced on the same platform, and some fees may already be retained under the applicable cancellation terms. Saving the original reservation does not necessarily create a wholly new platform fee.

Proposed Guest Memory workflow

Before a repeat stay, ask the guest to confirm relevant requirements. Match them against current property records. Route an unresolved issue to the host with a clear deadline, preserve the host's response, and ensure support staff can see the resolution. Where the property cannot meet the requirement, use accurate alternatives and the normal customer choice process.

The outcome sought is a stay that works as promised. A cancellation reduction is useful only if it reflects fewer avoidable failures and completed stays that guests actually want.

Service improvement to test

Select 1 million reservations where a confirmed requirement creates an actionable pre-arrival workflow. Assume avoidable cancellations in this defined category fall from 2.0% to 1.8%. That is a 10% relative reduction in the category, or 2,000 fewer cancellations. The 2.0% baseline is an assumption requiring measurement.

The pilot should report issue resolution before arrival, host response time and guest satisfaction. It should also track refunds and serious service failures, so keeping a booking does not conceal an unresolved problem.

Future repeat business

A better completed stay may support another booking. For illustration, a separate cohort of 100,000 previous bookers gaining one percentage point in completed repeat booking rate would produce 1,000 additional repeat bookings. That potential is calculated separately in the financial model below and is excluded from the main cancellation model.

Scope. The mechanism is a proposed extension of service continuity. The public source establishes marketplace economics, not AFG cancellation or loyalty performance.

02 Financial model

Value retained platform business

The central scenario prevents 2,000 cancellations. Assume only 50% represent incremental platform business after allowing for guests who would have rebooked on the same marketplace and fees otherwise retained. That leaves 1,000 economically incremental completed reservations.

Assume US$800 of commissionable accommodation value per reservation and 13% net platform revenue on that amount. Both are assumptions; the percentage is not calculated from the issuer's gross booking value, which includes other charges and has different timing.

2,000 × 50% × US$800 × 13% = US$104,000 incremental platform revenue.

At an assumed 65% incremental contribution margin, after ordinary transaction and fulfilment support costs, contribution is US$67,600. The intervention must be incremental after refunds, credits and alternative bookings are reconciled.

Avoided escalation work

Assume each avoided cancellation removes 20 minutes of exceptional escalation work, net of all additional manual prevention effort across the tested cohort. That releases approximately 667 hours. At US$30 per hour, the capacity equivalent is US$20,000. With 25% converted into reduced contracted support or overtime, cash savings are US$5,000.

This exceptional escalation work is separate from the normal completed-booking costs in the contribution margin. System integration and ongoing programme costs remain to be deducted.

Central value = US$67,600 contribution + US$5,000 cash savings = US$72,600.

The separate repeat-booking illustration yields 1,000 × US$800 × 13% = US$104,000 revenue, or US$67,600 contribution at the same assumed margin. It is excluded because a longer follow-up and careful overlap adjustment are needed.

How to prove the result

Randomise eligible guests and follow all reservations through completion, cancellation, replacement and refund. Use net platform contribution per assigned guest as the financial outcome. Where hosts serve both groups, measure possible spillovers or randomise suitable host clusters.

Preserve normal cancellation rights and guest choice. A fee retained from a dissatisfied guest is not evidence of better service. The central remaining programme break-even limit is US$72,600, falling to US$67,600 if staff capacity produces no cash saving.

03 Holiday rental search marketplace

Make returning demand less expensive to reach

A rental search marketplace can create value when an existing customer returns through its own app, website or permitted communication. The economic question is whether remembered needs make that route useful enough to reduce avoidable paid reacquisition spend while preserving completed bookings.

Verified benchmark. A Berlin-founded vacation rental group reported €151.7 million of Marketplace IFRS revenue and €15.8 million of Marketplace adjusted EBITDA for 2025. The issuer described a deliberate reduction in marketing expenditure and a focus on marketing efficiency. These are segment figures; the expanded group also contains other businesses and an acquired property manager. O3 FY 2025 issuer results, segment highlights

The reported improvement supports the commercial relevance of acquisition efficiency. The issuer did not attribute it to AFG or to the particular memory workflow proposed here.

Proposed Guest Memory workflow

After a completed stay, retain guest-confirmed details that can improve a later search: suitable destination areas, group configuration and preferences the guest wants reused. When the customer returns, offer a relevant starting point and current availability. Permitted reminders can connect a stated travel intention with available properties.

The marketplace should evaluate these suggestions against its existing recommendations. Relevance has to improve the customer experience sufficiently to generate additional useful direct engagement. A reminder alone does not establish that paid advertising can be reduced.

Service improvement to test

Measure the time required to produce a shortlist meeting confirmed requirements, completed repeat bookings per eligible guest and support contacts caused by unsuitable suggestions. For a defined population of 100,000 enquiries, assume two net minutes saved per enquiry. That releases 3,333 hours, equivalent to €100,000 at an assumed €30 per hour.

The financial model assigns no cash saving to that capacity. The measurable cash benefit is reduced marketing expenditure after the incremental cost of communicating with returning customers.

The repeat booking outcome

In the central model, total completed repeat bookings stay at 200,000. Their acquisition route changes. This is valuable repeat business at lower acquisition cost; it does not claim that 6,000 channel shifts are 6,000 additional stays.

A separate growth illustration in the financial model below shows how to evaluate additional repeat bookings. It must be reconciled with the acquisition model before any combined claim is made.

Scope. This is a proposed acquisition-efficiency study using public segment evidence. Direct share, customer acquisition costs and booking effects are assumptions.

03 Financial model

Reduce expenditure while preserving bookings

Assume a cohort generating 200,000 completed repeat bookings in 12 months. The direct or owned-channel share rises from 50% to 53%, with overall bookings and net fee revenue unchanged.

200,000 × 3 percentage points = 6,000 bookings shifted from paid reacquisition to owned channels.

Assume the company can remove €45 of paid media cost per shifted booking. This must be marginal avoidable expenditure, measured using a holdout or spend experiment, rather than average marketing expense divided by bookings.

6,000 × €45 = €270,000 avoided paid media spend.

Deduct €30,000 of additional annual campaign delivery, creative, consent management and communication costs across the cohort. The remaining annual programme cost is considered separately.

Central value = €270,000 avoided spend − €30,000 additional campaign costs = €240,000.

The platform's accommodation fees remain unchanged in this model. Commission paid by an owner is a different economic flow and is not counted as a saving for the marketplace.

All scenarios include the same €30,000 campaign cost. With no incremental channel shift and no paid spend reduction, the campaign would cost €30,000, before the remaining programme expense.

Separate repeat booking growth

If a separately measured treatment creates 1,000 additional completed repeat bookings, an assumed €100 of recognised net fee revenue per booking gives €100,000 revenue. At 60% incremental contribution after acquisition and fulfilment costs, value is €60,000. This is an alternative growth illustration, excluded from the central total.

How to prove the result

Run a guest-level memory experiment alongside a controlled paid media reduction. Compare completed bookings, fee revenue, contribution and actual spend against controls. A change in last-click attribution without lower expenditure produces no cash saving.

Include unsubscribes, complaint rates and suitability of recommended homes. Require sustained booking volume and a reduction in actual marketing outlay. The central remaining programme break-even limit is €240,000, with no payroll saving included.

04 Full-service vacation rental manager

Protect the owner relationship behind the inventory

A manager's relationship with a property owner depends partly on whether the organisation remembers commitments and resolves recurring issues. Better continuity could keep an owner from leaving, preserving future management fees and avoiding the need to replace that inventory.

Verified benchmark. A large North American vacation rental manager reported 5.080 million nights sold, US$1.856691 billion gross booking value and approximately US$910.5 million revenue in 2024. Reported gross booking value per night was US$365, including specified fees and estimated taxes. O4 Historical annual report, key metrics

The business was acquired in May 2025. The 2024 filing is used as a historical operating benchmark, rather than a statement of the present organisation's size. O4a Acquisition completion announcement

Revenue includes several services. Dividing reported revenue by gross booking value would not yield a reliable management commission rate. The model uses a separately stated fee assumption.

Proposed Guest Memory workflow

Connect the guest's confirmed service issue to the relevant property record. Preserve the owner's authorised instructions, the local manager's actions and the guest's verified outcome. Before the next stay, surface unresolved property matters and relevant guest requirements to the responsible team.

For the owner, a service review can explain what happened, what was resolved and which issues still need attention. This makes the relationship less dependent on a single staff member remembering a history of conversations.

Service improvement to test

On 100,000 existing owner-service contacts, assume four net minutes saved through accurate issue history and fewer repeated explanations. That releases 6,667 hours. Measure reopened issues and completion of commitments alongside time saved. Owner retention should be evaluated over a full renewal or departure cycle.

This cannot prevent an owner selling a home, withdrawing it for personal use or responding to market conditions. The target is a measured reduction in avoidable departures related to service continuity.

Guest repeat business

The retained-owner model preserves supply; it does not assume that all guests staying in retained homes are newly acquired guests. Additional guest repeat stays are a separate opportunity, evaluated on net management fees and adjusted for bookings displaced elsewhere in the manager's portfolio.

Scope. The retention benefit is hypothetical. Public financial data establishes the operating model and historical scale, rather than the effectiveness of the proposed workflow.

04 Financial model

Count only the fees that would otherwise be lost

Assume 10,000 eligible managed homes and an annual owner retention rate improving from 85% to 86%. The baseline and improvement are assumptions. This produces 100 retained owner contracts relative to the control case.

Assume each retained home generates US$50,000 annual commissionable accommodation rent, excluding taxes and pass-through charges, and a 20% management fee. That gives US$10,000 annual fee revenue per home. It is not derived from the benchmark company's revenue ratio.

Allow only six incremental months per retained contract within the first evaluation year, because departures and interventions occur throughout the year.

100 × US$50,000 × 20% × 6 ÷ 12 = US$500,000 retained fee revenue.

At an assumed 35% incremental contribution margin, after the service costs required to keep operating those homes, contribution is US$175,000. Replacement properties and displaced demand must be reconciled against the control scenario.

Service capacity and cash

100,000 contacts × 4 minutes ÷ 60 = 6,667 hours, approximately. At US$30 per hour, capacity is US$200,000. If 25% reduces overtime or contracted support, cash savings are US$50,000. These contacts relate to the existing portfolio, excluding the extra service costs already charged against retained-home contribution.

Central value = US$175,000 contribution + US$50,000 cash savings = US$225,000.

Service cash savings are held constant. A full 12 months of future fees would require contracts and demand to persist; that later value is excluded.

Separate guest repeat illustration

For a separate cohort of 100,000 previous booking households, a one percentage point increase in completed repeat stays produces 1,000 stays. At four nights, US$300 commissionable rent per night and a 20% fee, that is US$240,000 fee revenue, or US$84,000 contribution at 35%. It is excluded from the main total to prevent overlap with retained inventory and displaced bookings.

How to prove the result

Compare owner cohorts over 6 to 12 months or the appropriate contract cycle. Track departure reasons, retained fee months, costs and guest outcomes. The central remaining programme break-even limit is US$225,000, or US$175,000 without realised staffing savings.

05 Curated luxury villa agency

Carry the successful holiday into the next booking

For a villa booking, the remembered relationship often belongs to a family or travelling group. A useful record can preserve confirmed room arrangements, service preferences and the reasons a previous property suited the party. The next adviser can begin with that understanding and ask what has changed.

Verified benchmark. In a published executive interview, an Asia-focused luxury villa business reported hosting more than 80,000 guests a year, predominantly families and groups, and described customers returning to its villas. O5 Executive interview published by its payments provider

That is a self-reported historical operating statement, not an audited 2025 guest total. It counts people, not booking households. It does not disclose a repeat booking rate, commission margin or average booking value. Those inputs are explicitly assumed below.

Proposed Guest Memory workflow

After a stay, an adviser records which arrangements the lead guest wants remembered. Before suggesting another property, confirm the current party and priorities. A shortlist can then explain why each available villa fits, using current property facts rather than assumptions about an unchanged family situation.

Once booked, the villa team receives the relevant service brief and confirms what it can deliver. The booking adviser can see whether the key promises were fulfilled and use the outcome in the next conversation.

Service improvement to test

For 2,000 booking files, assume the combined enquiry, briefing and handover workload falls from 120 to 80 minutes, with 10 minutes of additional review and memory maintenance. The net improvement is 30 minutes per file, or 25% of the assumed original workload.

Measure verified requirements delivered on arrival, requests repeated by the guest and amendments caused by an inaccurate handover. These targets describe a test, not the benchmark operator's existing service.

The repeat booking outcome

Assume a defined cohort of 2,000 previous booking households has a 12-month completed repeat booking rate of 20%. Improving it to 22% means 40 additional stays. The cohort is an independent planning assumption; it is not estimated by dividing the reported guest count by an invented party size.

The agency earns its contracted fee. The property owner receives the remaining rent and bears property costs. Both may benefit from an extra stay, but their revenues must be shown separately.

Scope. This is a hypothetical repeat-stay opportunity using a recognisable public business profile. No individual guest story or financial result is presented as an actual AFG case.

05 Financial model

Turn forty additional stays into agency contribution

Assume US$12,000 commissionable accommodation rent per completed stay, excluding tax and separately purchased services, and a 20% agency fee. The model's booking value and fee are commercial assumptions requiring confirmation from contracts and actual bookings.

2,000 households × 2 percentage points = 40 additional completed repeat stays.

40 × US$12,000 = US$480,000 incremental accommodation rent.

US$480,000 × 20% = US$96,000 incremental agency revenue.

At an assumed 50% incremental contribution margin, after adviser time, incentives, payment costs and delivery obligations, agency contribution is US$48,000.

Owners would receive US$384,000 of the accommodation rent after the assumed agency fee, before their property operating costs. That amount is not agency income or profit and is excluded from the agency's return.

Service capacity and cash

2,000 existing booking files × 30 net minutes ÷ 60 = 1,000 hours. At an assumed US$25 per hour, capacity is US$25,000. A 25% reduction in actual seasonal cover or overtime would save US$6,250. The files exclude the variable work on the additional stays already charged to their contribution margin.

Central agency value = US$48,000 contribution + US$6,250 cash savings = US$54,250.

The table holds staffing savings constant. These values are additional to the existing business, conditional on the assumed improvements.

How to prove the result

Assign previous booking households to the memory workflow or normal service. Follow both through the next comparable travel season. Compare completed repeat stays and net agency contribution, controlling for destination, availability, party size, rates and promotions.

At seven nights per assumed stay, the central case requires 280 net additional sold villa nights. Verify that they occupy otherwise unsold dates, or deduct the contribution of bookings they displace. A booking moved from one villa in the same agency to another produces only the net difference.

The remaining annual programme break-even limit is US$54,250 under the central assumptions, falling to US$48,000 without realised staffing savings. Additional concierge spending and lifetime value are excluded.

06 Residential letting agency

Make ongoing management worth choosing

For a landlord, the agency relationship may begin with finding a tenant and expand into ongoing property management. Remembered instructions, accurate maintenance histories and clear ownership of commitments could make that additional service easier to demonstrate and more reliable to deliver.

Verified benchmark. A London lettings business founded in 1981 reported £111.0 million lettings revenue in 2025. It disclosed 43% of its portfolio as fully managed at year end and a portfolio of more than 32,000 tenancies at 31 January 2026. These dates and populations differ. O6 FY 2025 results, operational highlights and segment review

The issuer reported growth in management cross-selling and investments in its operating platform. That supports the economic relevance of management adoption; it does not establish an AFG effect or imply the agency lacks existing customer records.

Proposed Guest Memory workflow

Apply the same continuity concept to landlord and resident service. Keep property defects, authorised landlord instructions and resident contact preferences in their appropriate records. When a staff member changes, the next person can see unresolved work, previous contractor visits and the current commitment.

A landlord considering management services can receive a clear account of the service proposed, its price and the responsibilities it covers. Uptake should follow demonstrated value and explicit agreement.

Resident service knowledge belongs in service delivery. It should not be used to infer housing eligibility or to vary treatment based on protected personal characteristics. Access arrangements and maintenance instructions require current confirmation.

Service improvement to test

For 50,000 existing management contacts, assume three net minutes saved through usable issue history and fewer internal handoffs. That releases 2,500 hours. Evaluate reopened repairs, contractor revisits and completion of agreed actions as well as handling time.

The central model tests a two percentage point increase in management uptake among 10,000 eligible existing letting-only accounts. It does not treat all 32,000 reported tenancies as eligible prospects.

The equivalent of repeat business

Here, repeat business means retained landlord instructions and recurring management relationships. A separate retention illustration is provided in the financial model below. Resident continuity can also benefit landlords through lower voids, but the landlord's rent is a separate economic flow from the agency's fees.

Scope. This is a proposed service and management-adoption study. All intervention effects, cohort sizes and fee assumptions require validation.

06 Financial model

Count the additional management fee

Assume 10,000 eligible letting-only accounts. Additional management uptake of two percentage points produces 200 upgraded properties relative to normal service.

Assume £2,000 monthly rent, an additional management fee equal to four percentage points of collected rent above the existing letting service, and six incremental fee months per upgraded property during the first evaluation year. All are assumptions, with amounts excluding VAT.

200 × £2,000 × 6 × 4% = £96,000 additional first year agency revenue.

At an assumed 50% incremental contribution margin, after the extra property management workload, contribution is £48,000. The existing letting fee is excluded because it was already being earned. At twelve full fee months, the same upgrades would generate £192,000 additional revenue, conditional on continuing tenancy, collection and contract retention; that later run rate is excluded from the first year calculation.

Service capacity and cash

50,000 contacts × 3 net minutes ÷ 60 = 2,500 hours. At £25 per hour, capacity is £62,500. If 25% reduces overtime or external support, cash savings are £15,625. The contacts concern the existing managed portfolio, excluding the additional workload on the upgraded properties.

Central value = £48,000 contribution + £15,625 cash savings = £63,625.

Staffing savings are held constant. No rent increase is needed to create the modelled fee growth.

Separate retention illustration

For 5,000 already-managed accounts, a one percentage point improvement in landlord retention preserves 50 accounts. At £2,000 monthly collected rent, an assumed 8% existing management fee and six incremental months, this preserves £48,000 fee revenue, or £24,000 contribution at 50%. This separate cohort is excluded from the central total.

How to prove the result

Compare eligible account cohorts or matched branches. Track signed management agreements, collected incremental fees, retention, actual service costs and resident outcomes. Follow fee collection through the full evaluation horizon, including cancellations and unpaid rent.

If lower vacancy benefits a landlord, report that benefit separately. The remaining programme break-even limit for the central agency case is £63,625, or £48,000 without realised staffing savings.

07 Serviced apartment operator

Make a useful extension easy to arrange

An extended-stay guest may need extra days, a different servicing schedule or a return visit connected with a project. Guest Memory could help the team recognise those needs in time to offer available dates and carry confirmed arrangements into the extension.

Verified benchmark. A Singapore-listed lodging trust reported S$161 revenue per available unit and 80% occupancy for FY 2025 in its relevant operating portfolio. Its presentation specifies that this RevPAU measure concerns properties under management contracts, including those with minimum guaranteed income, and excludes master leases, rental housing and student accommodation. The population includes serviced residences and hotels. O7 FY 2025 presentation, slide 12

The benchmark does not disclose an AFG extension effect or a serviced-apartment-only rate. Dividing the reported S$161 by 80% gives a rough S$201.25 occupied-unit revenue reference using rounded inputs. The model selects S$200 as its own nightly rate assumption.

Proposed Guest Memory workflow

Preserve the guest's confirmed servicing preferences and any stated need to extend. Check current availability and price before making an offer. If accepted, update the reservation and make the agreed arrangements visible to the next shift. The system should retain the outcome, including a declined or unsuitable offer, so the guest does not receive repeated irrelevant prompts.

For a corporate guest, an authorised booker and the guest may have different responsibilities. Keep their instructions clear and route any extension approval through the customer's usual process.

Service improvement to test

For 2,000 eligible extension-related service cases, assume ten net minutes saved through fewer repeated confirmations and handoffs. That releases 333 hours. Measure accepted extensions, accurate fulfilment and the number of unresolved issues at handover.

An extension is a further purchase within an existing stay. A later return trip is a separate repeat booking and should be measured independently. This model values only the extension nights.

Who receives the value

The calculation is at the property operating level. An asset owner, lease operator and management company may receive different shares under their contracts. A manager cannot claim the whole accommodation revenue as its own fee income.

Scope. This is a proposed 1,000-unit serviced-apartment scenario using a mixed lodging portfolio as a rate and occupancy reference. It is not an extrapolation to every asset in that trust.

07 Financial model

Sell only nights that would otherwise remain empty

Assume a 1,000-unit portfolio with 365,000 available unit nights and an 80% baseline occupancy. That leaves 73,000 unsold unit nights over the year, although availability on the exact requested dates must be checked.

Assume 10,000 eligible extension decisions, with an incremental acceptance improvement of one percentage point after allowing for extensions that would happen anyway. This produces 100 additional extensions. At seven net additional occupied nights each, the result is 700 extra unit nights after displacement is deducted.

700 × S$200 = S$140,000 incremental accommodation revenue.

Assume S$60 variable cost per added night, covering incremental servicing, utilities, distribution and other guest-related costs. The incremental contribution is 700 × (S$200 − S$60) = S$98,000 before programme cost and any further contractual allocation.

The extra 700 nights increase occupancy by approximately 0.192 percentage points, from 80% to 80.192%, with available inventory unchanged. The annual unsold-night total alone does not prove that all extensions can be accommodated profitably.

Service capacity and cash

2,000 cases × 10 net minutes ÷ 60 = 333 hours, approximately. At an assumed S$30 per hour, capacity is S$10,000. If 25% reduces actual overtime or external cover, cash savings are S$2,500. The saved administrative work is separate from the variable cost of serving the additional nights.

Central value = S$98,000 contribution + S$2,500 cash savings = S$100,500.

Staffing savings are held constant. Extending a stay postpones its final departure clean; it does not automatically eliminate a clean. No checkout-clean saving is included.

How to prove the result

Randomise eligible guests, or corporate booking accounts where appropriate, before extension decisions. Compare net occupied nights and contribution after accounting for other guests displaced, group properties substituted and any discounts.

Evaluate accuracy of confirmed arrangements and guest satisfaction. Track later repeat stays separately, without attributing ordinary extensions to future loyalty. The central remaining programme break-even limit is S$100,500, or S$98,000 without realised staffing savings.

08 Residential rental landlord

Reduce the avoidable cost of a resident leaving

For a landlord, the most useful retention model starts with the days a home would otherwise stand empty and the work needed between residents. Better service continuity could reduce some avoidable departures when maintenance histories, promises and unresolved issues survive changes in staff.

Verified benchmark. A large US single-family rental landlord reported a 76,819-home same-store portfolio, 22.8% annual turnover, 96.8% average occupancy and US$2,450 average monthly rent for FY 2025. Its quarterly same-store turnover operating expenses, net of resident recoveries, sum to US$39.650 million for that year. The rent and cost schedules are unaudited supplemental operating disclosures. O8 FY 2025 supplemental report, pages 5, 17, 21 and 26

The cost sum is derived from reported quarterly amounts: US$10.268m + US$11.641m + US$9.618m + US$8.123m. Turnover capital expenditure is separately disclosed and excluded from this model.

Proposed Guest Memory workflow

Preserve a usable record of the resident's reported problem, completed visits, remaining work and agreed contact arrangements. When the resident follows up, the next staff member can act from that history. Before a renewal discussion, confirm whether earlier commitments have been resolved.

The purpose is service delivery. Household profiling is unnecessary for the model. Eligibility decisions, rent policy and the resident's ability to leave are outside the proposed treatment.

Service improvement to test

A proposed target is a 10% relative reduction in repeat contacts about the same unresolved issue, alongside accurate completion and resident satisfaction. Faster closure in a system does not count if the underlying problem remains.

The financial target is a fall in annual turnover from 22.8% to 21.8%, a one percentage point reduction, equivalent to approximately 4.39% fewer turnover events. This is a hypothesis, not a disclosed service baseline or forecast.

The equivalent of rebooking

A continuing resident provides repeat rental business. The incremental value is not the entire next year's rent: a replacement tenant would usually pay rent too. The model counts only the avoided vacant days and a defensible portion of turnover operating expenditure, after extra costs.

Scope. This study uses the reported portfolio size and expense base. The turnover change, vacancy period, avoidable cost share and intervention expense remain assumptions.

08 Financial model

Anchor the model in disclosed rental operations

76,819 homes × 1 percentage point = 768.19 fewer turnover events over a comparable 12 months. Fractional events represent an expected portfolio result; realised events will be whole numbers.

Assume each prevented turnover avoids 14 vacant days. At the reported US$2,450 monthly rent, daily rent is US$2,450 × 12 ÷ 365 = approximately US$80.55.

768.19 × 14 × US$2,450 × 12 ÷ 365 = US$866,266 additional rent, rounded. Allow 5% for marginal collection losses and occupied-day costs: incremental rent contribution is US$822,952.

Use an avoidable share of the actual cost base

The reported annual turnover rate implies about 17,515 turnover events. Dividing the derived US$39.650 million turnover operating expense by that event estimate gives approximately US$2,264 per event. This is a derived average using a rounded turnover rate, not a quoted cost for each move-out.

Assume 75% of proportional turnover operating expenditure is avoidable in the evaluation year. The remaining 25% allows for work or cost that persists. No capital expenditure saving is counted.

US$39.650m × (1 percentage point ÷ 22.8%) × 75% = US$1,304,276 cash expense reduction.

Deduct US$100,000 of additional annual resident-service intervention expense across the cohort. The remaining technology and programme cost is still to be deducted.

Central value = US$822,952 rent contribution + US$1,304,276 avoided expenditure − US$100,000 = US$2,027,229, using unrounded calculations.

How to prove the result

Use a resident or community comparison spanning a full turnover cycle. Separate service-related departures from relocation, purchase and affordability changes. Compare occupied days, actual invoices, collection losses and all intervention expense. Rental rate differences between retained and replacement residents must also be reconciled; the model assumes no net difference beyond the vacancy effect.

Some work may merely be deferred. Follow future expenditure before describing it as a lasting saving. The remaining programme break-even limit is approximately US$2.027 million under these assumptions. No additional staff-time saving, capitalisation uplift or property valuation gain is included.

Programme budgets and break-even

Translate potential value into a purchase decision

Each case gives a value before the remaining cost of GuestMemoryOS. Call that value V. Let C be the complete remaining cost for the same evaluation period, including licence, integration, training, data preparation, operating oversight and additional staff effort not already deducted in the case.

Net benefit = V − C. Return on cost = (V − C) ÷ C.

The budgets below are illustrations, not AFG prices or quotations. They show what the central scenarios would mean at a stated all-in remaining cost. All benefits and costs cover the same 12-month evaluation horizon; the retained-contract models already allow six incremental fee months in year one.

For example, study 05 gives a 117% return on an assumed US$25,000 remaining programme cost: US$29,250 divided by US$25,000. That is conditional on the repeat stays, contribution margin and cash saving actually occurring.

What could change the decision

Lower contribution margins, additional incentives, unavailable inventory, overlap with existing tools, implementation delays or incomplete guest adoption can reduce value. A failed intervention can have zero benefit or a negative effect. The lower scenarios are sensitivity choices, not guaranteed floors.

The central value is also the break-even ceiling for the remaining programme cost. A commercially attractive price would normally leave a margin for uncertainty. Replace the assumptions with the prospective customer's own verified unit economics before agreeing a budget or publishing a customer-specific return.

Validation and measurement

Establish the baseline before the treatment

Define eligible guests, properties or accounts, the observation period and the existing service process. Measure the baseline event rate, contribution per transaction and actual cost of the tasks addressed. Record exposure to the new workflow and whether the relevant memory was accurate, current and actionable.

Use a single principal financial outcome for each experiment. Suitable choices are contribution per assigned guest for an OTA, net acquisition spend at preserved booking volume for a marketplace, retained fee months for a manager, and occupied-day contribution plus avoidable expenditure for a landlord.

Compare against the service already in place

Randomise at the level that limits spillovers: guest, booking household, owner account, property or community. Keep pricing, availability and normal service comparable. Where randomisation is impractical, use matched comparison groups and describe the weaker causal certainty.

A short operational trial can establish whether information reaches the right team and actions are completed. Booking effects require cancellations to mature. Repeat stays, owner retention and resident turnover usually need a longer follow-up aligned with their commercial cycle. Set duration and sample size from the baseline rate and the effect worth detecting, before reviewing outcomes.

Reconcile the full economic effect

Count completed, paid business after refunds, cancellations and displacement. For intermediaries, use their own net fees. Deduct incremental fulfilment, incentives and communication costs. Verify whether staff time leads to lower spending or simply more available capacity.

Reconcile transfers between properties, brands and channels. Avoid counting the same booking as a conversion gain, retained cancellation, repeat stay and marketing saving. Use one transaction-level benefit ledger with a clear counterfactual and an explicit rule for allocating overlapping effects.

Measure service quality alongside money

Track correct fulfilment of confirmed requirements, repeated questions, reopened cases, resolution time and guest or resident satisfaction. Keep confirmed facts distinct from inferences. Test the relevance and age of information, provide a way to correct it and limit access to the people who need it for service.

Report effects with uncertainty and the number of eligible accounts or guests. A result should be called demonstrated for AFG only when the specific workflow produces a measured improvement against the comparison group, with acceptable service quality and positive value after full costs.

Publication claims that the evidence can support

Public evidence supports the proposition that useful guest history can improve accommodation search and that accessible knowledge can improve some service work. The eight studies explain how related operational changes could create value in distinct rental businesses. Their financial figures remain conditional models until the corresponding commercial results are measured.

Evidence register

Research and product scope

E1 Guest journey search experiments. Operator engineering publication, accessed 16 September 2026, Results section. Supports the reported relative changes in uncancelled bookers and nights, and the role of long-term and recent guest behaviour. The underlying research paper was submitted in June 2026. Engineering results and research paper. Evidence type: operator-reported online A/B tests. Transfer to AFG remains unproven.

E2 Service productivity study. Brynjolfsson, Li and Raymond, Generative AI at Work, revised 6 November 2024. Supports the 5,172-agent sample, average 15% improvement in issues resolved per hour and heterogeneous effects. Author manuscript. Evidence type: empirical study of staggered deployment; different product and industry setting.

A1 AFG product scope. Public product page, accessed 16 September 2026. Supports the description of a governed memory layer and intended progressive integration with existing business systems. Product page. Evidence type: vendor description, not a certification of integrations or financial performance.

A2 AFG operating proof material. Public proof page, updated 21 September 2026. Covers live development at Mandarin Beach Villa, its public residence review record, the operating snapshot as at 21 September 2026 covering activity since go-live on 14 July 2026, and completed internal cruise simulations and a simulation of a 600-room hotel in Dubai. Proof page. Evidence type: company-reported operation and simulation testing, with separately linked property reviews. Detailed simulation results are available on request to verified cruise-line representatives.

Rental distribution benchmarks

O1 Global travel group. 2026 annual reporting package containing FY 2025 results. Room nights and gross booking definitions: printed annual report pages 31 to 32, PDF pages 156 to 157. Revenue table: PDF page 157. Supports 1.235 billion room nights, US$186.1bn gross bookings and US$26.917bn revenue. Annual reporting package. Group scale; not the eligible cohort or fee rate used in study 01.

O2 Global host marketplace. FY 2025 Form 10-K. Key Business Metrics and Revenue sections. Supports 533m Nights and Seats Booked, US$91.273bn gross booking value, US$12.241bn revenue and metric definitions. Annual filing. Platform scale; not a reservation count or study-specific cancellation rate.

O3 European vacation rental group. Issuer announcement dated 19 March 2026, Business segment highlights. Supports €151.7m Marketplace IFRS revenue, €15.8m adjusted EBITDA and the stated marketing-efficiency focus. FY 2025 results announcement. Segment results, with adjusted EBITDA a non-IFRS measure. No assumed acquisition cost is represented as disclosed.

Property operating evidence

Managers and agencies

O4 North American vacation rental manager. FY 2024 Form 10-K, Key Business Metrics table on printed page 72, definitions on page 73 and revenue disclosures. Supports 5.080m nights, US$1.856691bn gross booking value, approximately US$910.5m revenue and reported US$365 gross booking value per night. Historical annual filing. Historical scale and definitions; no management commission is inferred from the aggregate revenue ratio.

O4a Acquisition status. Company announcement dated 1 May 2025. Establishes completion of the acquisition and the historical status of the standalone 2024 benchmark. Completion announcement.

O5 Asia luxury villa agency. Published executive interview hosted by its payments provider, accessed 16 September 2026; no publication date shown in the retrieved interview. Supports the statement of more than 80,000 guests annually and the qualitative description of returning villa customers. Executive interview. First-person company testimony, not audited financial data; no repeat rate or household count is disclosed.

O6 London residential letting agency. FY 2025 results announcement dated 5 March 2026. Pages 1 to 4 give portfolio and management adoption facts; the 32,000-plus tenancy figure is dated 31 January 2026 in footnote 7. Page 4 gives lettings revenue. Results announcement. Supports commercial scale and recurring management economics, not the assumed fees or adoption effect.

Property operators and landlords

O7 Singapore-listed lodging trust. FY 2025 results presentation dated 29 January 2026, slide 12. Supports S$161 RevPAU and 80% occupancy, including the relevant management-contract scope and exclusions. Results presentation. The S$201.25 reference is arithmetic using rounded reported inputs; the S$200 scenario rate is an assumption.

O8 US residential rental landlord. Q4 2025 Earnings Release and Supplemental Information, released 18 February 2026. Pages 5 and 17 cover turnover; page 21 gives FY same-store homes, rent and occupancy; page 26 gives quarterly turnover operating expense net of resident recoveries. Operating supplement. The US$39.650m annual turnover operating cost is summed from four reported quarters. The implied cost per turn is derived; the 75% avoidable share is assumed.

Model provenance. All other financial inputs, cohorts, margins, service targets and improvement rates are stated scenario assumptions. Calculations use unrounded intermediate values; displayed totals may differ slightly from sums of rounded components. USD, EUR, GBP and SGD amounts remain in their original scenario currencies, with no exchange-rate conversion.