Two units, two different listing problems, and one question underneath both: can these leases be covered, and if so from where.
Six things have been established since the deep dive. They sharpen the priorities rather than change the diagnosis.
Both calendars are confirmed clear of any iCal or channel-manager sync, acted on within hours of the call. So the visibility test is not pending — it has been running for the whole measurement window. The result: 11,118 first-page impressions on the 2BR and 5,146 on the Heritage Home, with no sync in place.
That is a useful finding. Sync was not what was suppressing distribution, and the 2BR's weak performance has a different cause entirely (section 4). It also means reconnecting later carries less risk than feared. The live question is no longer whether to disconnect, but how to manage the double-booking exposure that manual reconciliation creates while the channel stays open.
Confirmed at 45 on both units. It is blocking precisely the long-stay bookings that produced the two strongest months of the year, and it blocks the mid-term strategy in section 9 before it starts. Raising it is now a first-week action rather than housekeeping.
The Heritage Home carries a 3-night minimum on Thursday and Friday check-ins against a 2-night default. The 2BR has a Thursday customisation set to 2 — the shape of a rule configured and then partially reversed. Both are consistent with PriceLabs residue rather than intention. The clear-out instruction from the call was right; these are two of the places to look.
The new-listing promotion on the 2BR reads Inactive — 20%, 0 views, 6 used, 4 remaining. It expired on the three-review threshold rather than the ten-booking one, so the four remaining uses are stranded: the qualifying condition cannot be un-met. There is no custom promotion tool on the listing at all, in the discounts panel, on per-date selection, or on the promotions route, which renders empty.
The plan should not hold a slot open for a promotion that is coming. The available levers are base price, per-date overrides, and the six standard discounts — all of which sit at zero on the 2BR.
The Heritage Home is on, with the R3,500 floor and a R10,750 ceiling as agreed. The 2BR is still off, running a manual R2,450 base with a R2,650 weekend rate and hand-set overrides. An unfinished action rather than a decision.
Rose Street, De Waterkant is the Heritage Home — three bedrooms, six beds, two baths. Colosseum, Unit 705 St George's Mall is the 2BR. The rent figures attach as shown below.
| Unit | Rent | Utilities | Total |
|---|---|---|---|
| Colosseum — 2BR, Unit 705 St George's Mall | R23,625 | R740 | R24,365 |
| Rose Street — Heritage Home, De Waterkant | R34,500 | R2,200 | R36,700 |
| Per month | R61,065 | ||
| Per year | R732,780 |
| Unit | Rate | Nights to break even | % of a 30-day month |
|---|---|---|---|
| Heritage Home | R2,800 | 13 | 44% |
| 2BR St George's | R2,200 | 11 | 37% |
These breakeven figures are a floor, not the real number. They count rent and utilities only. The populated rent-to-rent calculator gave a break-even occupancy of 62% on the Heritage Home at a R3,500 base — against the 44% shown above. That implies roughly R65,000 of monthly cost on that unit, some R28,000 more than rent and utilities, and at the R2,800 actually being achieved rather than the R3,500 modelled it would push break-even towards 78%.
Two possibilities, and they lead different places. Either there is a substantial operating cost base not reflected here — cleaning labour, linen, consumables, wifi, insurance, platform commission, a management fee, an owner drawing, debt service — in which case the position is materially worse than sections 2 and 9 describe. Or the calculator was run on assumptions worth revisiting, perhaps portfolio-level rather than unit-level. Section 2 and the revenue plan in section 9 should be rebuilt on the completed calculator for both units before either is relied on for the February decision.
Add cleaning labour, linen, consumables, wifi, insurance and platform fees and the real floor rises from the figures above. The rent-to-rent model on the Heritage Home was already showing a break-even occupancy above 60% at a R3,500 base across a full twelve-month cycle. That is the structural finding, and it matters more than any single listing fix: at achievable Cape Town winter occupancy this unit does not clear its own cost, and no pricing tool changes that.
The Heritage Home shows 40% occupancy. Airbnb measures that against available nights — 8 booked of 20, with 8 blocked and 12 unbooked. Rent is charged on all 30. Against the nights that actually have to be paid for, occupancy was 27%.
The Airbnb figures understate the business, and by an amount I can't yet quantify. Booking.com reservations are reconciled onto the Airbnb calendar by hand, which strips those nights out of Airbnb's view of occupancy, revenue and impressions. Some of the Heritage Home's eight blocked nights are likely Booking.com stays. The market benchmarks also compare against hosts whose calendars aren't hand-blocked, so the gap to market is real but narrower than it appears. One input resolves this: the Booking.com night count per unit for the last three months.
This is an annual-cycle business, not a monthly one. December and January have to fund June through August, and the forward book already carries R148,154 for the rest of 2026. The Heritage Home's January-to-August position was a net loss of roughly R72,000 before the rent waiver closed the gap — which means the waiver, not the trading, is what balanced the year so far. That is not repeatable.
One booking in May at R11,500, two in September at R16,847 — R28,347 for the year, against six cancellations across October, December and January. Six cancellations to three completed stays is the signature of Booking.com's no-guarantee card policy letting people hold dates for free, and the cancelled ones were peak season.
| Stage | Value |
|---|---|
| First-page search impressions | 5,146 |
| First-page impression rate | 52.1% |
| Page views | 1,130 |
| Search to listing | 21.96% |
| Listing to booking | 0.88% |
| Overall conversion | 0.19% |
| Versus similar listings in the area | 0.26% below |
Occupancy 40% of available nights, down 41.2% on the prior 30 days. Three check-ins. Page views down 38, wishlist additions 54 and down 27. Achieved rate R2,800, which is below the listing's own R3,500 floor, because the 30% monthly and 15% weekly discounts cut into it.
4.00 in the window, with no five-star reviews. Both recent reviews scored 4. Michael's breakdown is the tell: cleanliness 5, check-in 5, communication 5, location 5, value 5 — accuracy 4, overall 4. His written review was unreservedly positive and specifically praised the Table Mountain view. Airbnb attached a photo tip to the accuracy score.
Straight fives, warm text, 4 overall means the stay was good but wasn't the stay the page promised. Antonio's review mentions three levels; unflagged stairs on a heritage property would alone explain an accuracy ding, and the construction crane visible in the terrace photos is a second candidate. The chain runs: page oversells, guest recalibrates on arrival, 4-star review, 4.00 public rating, next guest bounces off the page, 0.88% conversion.
Every day in October sits at a flat R3,500 — Smart Pricing pinned to its minimum because demand isn't there. September is already discounted into the R2,200s. January prices itself up to R5,500–6,400, so the mechanism works; October is simply on the mat. This is the proof that price is not the constraint on this listing. Every rand off from here is margin lost with no booking gained.
The R1,116 cleaning fee has no short-stay reduction configured, adding roughly 16% to a two-night stay at floor price and over 30% at the rates actually being achieved. The field exists and is empty.
| Stage | Value |
|---|---|
| First-page search impressions | 11,118 |
| First-page impression rate | 45.9% |
| Page views | 302 |
| Search to listing | 2.72% |
| Listing to booking | 1.99% |
| Overall conversion | 0.05% |
| Versus similar listings in the area | 0.68% below |
The previous listing was deleted outright, taking its review history with it. That single fact accounts for every anomaly: no data before mid-August, four reviews on the listing against 58 at host level, a burned-out new-listing promotion, and the 2.72% click-through.
Airbnb is serving 11,118 first-page impressions — more than double the Heritage Home's, on a listing a month old — to a card showing almost no reviews and no Superhost badge, in a CBD market where competitors show dozens of reviews and most of the comparable stock is Superhost. There is nothing on that card to stop the scroll. Once they land, the page converts at 1.99%, more than twice the Heritage Home's rate. The page works. The search card doesn't.
There is a clock on this. Those impressions are a new-listing distribution boost, and it decays. Every week at 2.72% burns visibility that will not come back. The reviews are not recoverable, so the only route is generating new ones fast, inside the boost window. This is the most time-sensitive item in the plan.
Base R2,450, weekend R2,650, Smart Pricing off, with hand-set overrides showing R2,600 to 10 September and R2,200 after. Airbnb's price tips suggest R1,600–1,900, putting the listing 20–30% above where its model sees the market. No cleaning fee, no extra guest fee, no fees of any kind — and no weekly, monthly, last-minute, early-bird or length-of-stay discount set.
Smoke alarm marked unavailable. A duplicated section in the description. A safety warning in the location copy that belongs in pre-arrival messaging rather than on the public page, where it deters the click it is meant to reassure.
| Setting | Heritage Home | 2BR St George's |
|---|---|---|
| Smart Pricing | On | Off |
| Floor / ceiling | R3,500 / R10,750 | n/a — base R2,450 |
| Cleaning fee | R1,116, no short-stay rate | R0 |
| Weekly discount | 15% | 0% |
| Monthly discount | 30% | 0% |
| Minimum nights | 2, but 3 on Thu and Fri | 2 throughout |
| Maximum nights | 45 | 45 |
| Advance notice | Same day, 5pm cutoff | Same day, 6pm cutoff |
| Preparation time | None | None |
| Booking window | 6 months | 12 months |
| Restricted check-in / out days | None | None |
| iCal / channel sync | None — deliberately cleared | None — deliberately cleared |
One unit charges R1,116 to clean and the other nothing. One offers deep length-of-stay discounts and the other none. One prices dynamically and the other by hand. The booking windows differ by six months. None of this reads as deliberate positioning of two different products, and the incoherence is itself a finding: these are settings that accumulated rather than settings that were chosen.
The public host card shows 58 reviews, a 4.66 rating and one year of hosting. No Superhost badge. That rating is displayed to every guest who opens either listing, in a market where most comparable CBD stock carries 4.8 or higher, so it is working against conversion on both units at once. The listing-level rating on the Heritage Home is 4.85, and accuracy is the weakest category — consistent with the diagnosis in section 3. The likeliest drags at host level are the forced-refund guest and the upset guest at the second property.
The host rating will not move quickly, and the plan should not assume it does. With 58 reviews at 4.66, lifting the displayed average to 4.8 takes roughly 40 consecutive five-star stays. That is a year of work, not a quarter. Superhost is therefore a 2027 objective rather than a lever available now.
The 2BR's own listing rating is the opposite case: with only four reviews, it moves substantially on a handful of stays. That asymmetry is the whole argument for concentrating the review push on the 2BR rather than spreading it — the fast-moving number is there, and the boost window is there too.
Yusuf is confirmed as her son, with calendar and messaging access only and no settings rights. The business started in May 2025, consistent with the tenure shown on the profile, so the earlier question about a possible account reset is closed.
One loose end: a dashboard modal reported that "the primary host confirmed this price adjustment to account for the service fee change." If it was neither Hani nor Yusuf, it is most likely Airbnb's own service-fee migration worded confusingly. Low priority, worth one question.
Cover runs to theft, fire, public liability and SASRIA. No business interruption, no building cover. Seven weeks were lost to the ceiling repair and that income was uninsured. On a portfolio already running at a loss, a repeat is not a setback, it is the end of the venture.
Recovering 51 days of rent — 20 in June plus all of July — was the win. AirCover is not a second bite: it covers guest-caused damage and liability, its income protection attaches to bookings cancelled because of guest damage, and a landlord's building repair is not a covered event. Filing windows are measured in days from checkout. Worth confirming with Airbnb directly, but nothing should be planned around recovering it.
Colosseum ends 30 April 2027. Rose Street ends 1 December 2027. Notice is 60 days as calendar months, which puts the real decision deadlines at end of February 2027 and end of September 2027.
February is the date to diarise: that is when another twelve months has to be committed on a unit that does not currently cover its rent. Rose Street expiring on 1 December is worse placed — handing the unit back at the start of peak season forfeits the only weeks that earn properly. If any renegotiation is available, moving that end date past February is worth more than a rent reduction.
Subletting is permitted under the lease, which is the significant de-risker. Worth knowing that a subletting clause and express permission to short-let are treated differently, and that body corporate conduct rules sit above the lease regardless of what it says.
A branch manager has limited discretion. Ad-hoc rent reductions and early releases need sign-off that isn't reachable across the desk, which is why the early-termination request was refused and August and September were billed anyway — process rather than malice. The upside is that an organisation that size has escalation paths, a complaints function and a reputation to protect. The waiver only came after legal advice; assume the same of anything else worth asking for, and escalate above branch level rather than arguing at it.
The commercial rate-in-the-rand is roughly 2.35× residential, and commercial properties also lose the residential rebate on the first R620,000 of value. Reclassification triggers on thresholds around 180–210 nights let or advertised, or on non-primary-residence status.
The critical point for an arbitrage operator: the rates bill goes to the registered owner, not the tenant. Whether it reaches Hani at all depends entirely on the lease. The real exposure is that the listing triggers the landlord's reclassification, at which point the landlord has every incentive to end the arrangement. Verify the current status of the draft by-law before budgeting for it.
On leases, rates and insurance this report is reading documents and public information, not advising. The business interruption cover needs a broker who writes short-term-let policies in South Africa, which is a narrow market. Any lease renegotiation or exit costing needs someone qualified on the actual wording.
Each tier contains the one above it. Minimum protects the position and stops obvious losses; it does not grow revenue. Medium is where most of the available revenue sits. All-in is the full build, and it is a job rather than a side activity.
Roughly 3–4 hours in week one, then about an hour a week.
The highest-value single click in the plan. It currently blocks the long stays that produced the two strongest months of the year, and it blocks the mid-term strategy entirely.
Audit the first five photos against what a guest actually walks into. Add the three-level layout and any stairs to the description explicitly. Remove or reshoot terrace photos showing the construction crane. This is the root cause of the 4-star reviews and sits upstream of everything else on that listing.
Tick the smoke alarm if one is fitted. Delete the duplicated description section. Move the safety warning out of the public location copy into pre-arrival messaging.
Currently declining Friday weekend arrivals, the highest-intent traffic in this market. Airbnb's own panel estimates it misses 25% more searches.
The field is empty. R1,116 flat adds over 30% to a two-night stay at achieved rates.
2BR from 0% to roughly 10% weekly and 20% monthly. Heritage Home from 15%/30% down to 10%/20%, because 30% on top of a floor price is what drags the achieved rate below the floor. On the 2BR these don't only cut price — they change which searches the listing appears in, which is the stage that's broken.
Expect stability rather than uplift. Smart Pricing biases downward and will sit near the floor in a soft market; the floor is the protection.
The largest uninsured exposure in the portfolio. A broker who writes short-term-let cover.
If they are Booking.com stays, opening them creates a double booking.
Manual calendar reconciliation without a written record is how double bookings happen, and a host cancellation costs the penalty plus a ranking hit — on a listing that has just lost its review history and cannot absorb another setback.
Everything above, plus roughly 6–8 hours a week. This is where the revenue plan in section 9 becomes achievable.
Drop the minimum to 1 night, take gap fills, price toward Airbnb's suggested R1,900, over-deliver, and ask for the review every time. The goal is review count, not revenue. Only run it while the stays can genuinely be serviced — volume at 4 stars is worse than no volume, and host-level rating is already 4.66.
List both units for 2–3 month stays on Property24 and SafariNow. The lease being in the company name provides the documentation those channels expect. This is the single biggest lever in the revenue plan, and it depends on action 1.
Facebook groups, the nomad and expat communities, relocation agents, and the corporate housing desks at the larger employers and film production companies. Post the Airbnb links too, since clicks from social are a traffic signal to the listing as well as a direct route — but the higher-value ask in these groups is the 2–3 month stay, because that is what converts an empty winter into a fixed monthly figure. Read each group's posting rules first; promotional posts get members removed and she'd lose the channel.
Costs nothing and compounds. This is the foundation of the direct channel in the All-in tier; without it, that tier has nothing to sell to.
Custom rule sets, seasonal rates and stay restrictions running to 2028. Two are already located. Airbnb stores these as per-date overrides on the calendar grid rather than in the settings panel, so this needs a month-by-month sweep rather than a settings check.
Agree a referral fee structure. Overflow referrals cost nothing to receive and monetise relationships that already exist in the buildings.
The disconnection test has run and distribution was not suppressed, so judge the channel on its own numbers instead: R28,347 for the year against six cancellations. If it stays, the manual log in action 10 is not optional.
Three asks in order of value: move the Rose Street end date past February 2027; a rent reduction reflecting a soft market; and written clarity on who carries a rates reclassification.
Every occupancy and revenue figure in this report stays an estimate without it.
Advertise it on Property24 and through the relocation and corporate-housing channels, and find out what it actually fetches on a 6–12 month furnished let. This costs nothing but a listing, and it produces the single most useful number missing from the plan: what the floor is worth. Subletting is permitted, so this is a live option rather than a hypothetical — see section 9.
The star explanation already sits in the guidebook and the checkout information, which is the right place for it — both are post-booking, guest-only surfaces, so there is no cost to conversion. Two refinements. Guidebooks are lightly read, so the checkout information is the surface carrying the weight; make sure the wording there is an invitation to raise problems rather than a request for stars. And note the result: the explanation was in place and two 4-star reviews still came in, which says the education is not the binding constraint.
The missing message is mid-stay, on day one or two, asking whether everything is as expected. It is the only point in the sequence that catches a fixable complaint while it can still be fixed, and it is worth more than anything said after checkout. Pre-arrival carries the expectation-setting — the stairs and three levels, the construction next door, any street noise — which is where accuracy is actually protected.
Everything above, plus roughly 15–20 hours a week for three months. Treat this as a job, and commit only if the numbers in section 9 justify it.
Website, payment gateway, review widgets pulling the Airbnb reviews across, and a booking engine. The only route that removes platform commission, and the only asset in this plan that survives the leases ending.
Not occasional posting — a content cadence with the property as the subject. Bringing her daughter in is sensible, but it needs a brief and a schedule rather than goodwill.
January at R5,500–6,400 for a three-bed heritage home with Table Mountain views is probably under market. Check AirDNA first. Do not raise a 4.00-rated listing into peak season — fix the rating, then reprice, in that order.
December's profit is May's cash flow. Without a model that ring-fences the peak surplus against the winter deficit, a good December gets spent and the same crisis arrives every June.
Deposit forfeit plus penalty, set against twelve more months of losses at current trading. Also price a conversion to long-term sublet at cost, which caps the downside without breaking the lease. This sits in the All-in tier because it needs real numbers, and those numbers are the point of the exercise.
From 1 July 2027, at roughly 2.35× residential. Verify the by-law's current status and get the lease position in writing.
The target is R61,065 a month, R732,780 a year. Short-term letting alone has not reached it and, on the break-even numbers in section 2, is unlikely to. So the plan has to stack channels rather than optimise one — and it has to be honest about whether the total clears the obligation.
| Layer | Low | High | Depends on |
|---|---|---|---|
| Airbnb — Heritage Home | R22,000 | R45,000 | Rating recovery; heavily seasonal |
| Airbnb — 2BR | R8,000 | R30,000 | Review count inside the boost window |
| Mid-term 2–3 month stays | R18,000 | R38,000 | Raising the 45-night cap first |
| Booking.com | R0 | R8,000 | Cancellation rate; channel decision |
| Direct bookings | R0 | R12,000 | Guest list; 3+ months to build |
| Host referral income | R0 | R4,000 | Building relationships; not reliable |
Read the low column, not the high one. At the low end the stack reaches roughly R48,000 against R61,065 — a shortfall of about R13,000 a month. At the high end it clears comfortably, but the high end assumes peak season, a recovered rating and a mature direct channel all at once, which will not happen before December 2027.
The honest conclusion: short-term letting is unlikely to cover both leases through the winter of 2027, whatever is done to the listings. Mid-term letting is the layer that changes the arithmetic, because it converts unpredictable nightly demand into a fixed monthly figure at low turnover cost. It is also the layer currently blocked by a single setting. That is why raising the 45-night cap sits at the top of the action list rather than in the middle.
A 2–3 month corporate or digital-nomad let on the Heritage Home at R30,000–38,000 sits at or slightly below its R36,700 cost — but with near-zero turnover cost, no cleaning labour, no marketing spend and no cancellation risk. The 2BR at R18,000–24,000 against R24,365 behaves the same way. Neither produces profit alone. Both convert a loss with high variance into a small loss with near-total certainty, and that is what makes the winter survivable while the listing fixes take effect.
Mid-term should be the default for June to August and the exception for December to February, not a year-round policy.
Subletting is permitted under the lease. That makes converting a unit to a long-term furnished tenancy a front-line strategic option, not a fallback, and it has been buried in the exit costing until now.
The logic is simple. Rent-to-rent arbitrage only earns its keep if short-term revenue beats what the unit would fetch on a long let by enough to cover the extra cost, the extra risk and the very considerable extra work. On current trading it does not. If a unit can be sublet at or near what she pays for it, she converts an uncertain monthly loss into a certain zero — and gets her time and attention back for the people-development business, which may be the more profitable use of both.
Three things make it more attractive here than it first looks. The units are already furnished, and furnished long lets in Cape Town command a premium over unfurnished with corporate and expat tenants. There is no turnover cost, no cleaning labour, no platform commission and no cancellation risk. And it removes the rates reclassification exposure in section 7 entirely, because a long-term residential tenancy does not trigger it.
Three constraints that decide whether it works.
Lease runway. A sublease cannot outlast the head lease. Colosseum runs only to 30 April 2027 — about seven months — which is too short for most good long-term tenants, though it fits a 6-month contract or academic let well. Rose Street runs to 1 December 2027, roughly fourteen months, which is a normal lettable term.
Tenant default. She stays fully liable to Trafalgar while depending on a subtenant. Under the PIE Act, removing a defaulting tenant in South Africa is slow and expensive. Proper vetting, a deposit, and a written sublease are not optional, and the sublease terms should mirror the head lease.
Written consent for this specific arrangement. A subletting clause permits the principle. Get Trafalgar's sign-off on the actual sublease before signing a tenant.
The channels in the table above are not all compatible on the same unit at the same time. A full-time tenant and an Airbnb listing are mutually exclusive; mid-term blocks the peak weeks it occupies. So the portfolio should run different strategies on the two units rather than every strategy on both — and the runway and earning power point in opposite directions.
| Unit | Strategy | Why |
|---|---|---|
| Rose Street — Heritage Home | Short-term through peak, mid-term through winter | Genuine STR demand — 52.1% first-page placement, 21.96% click-through, and January priced at R5,500–6,400. At reasonable peak occupancy this unit substantially out-earns any long let over December to February. Do not hand those weeks to a tenant. Fill June to August with a 2–3 month let instead. |
| Colosseum — 2BR | Short-term now inside the boost window, then convert | Weaker STR prospect — no review history, no Superhost badge, and Airbnb's own model reads the market 20–30% below her rate. Seven months of runway suits a 6-month furnished let almost exactly. Use the boost window to build reviews and take peak season, then convert around February rather than renewing into another loss-making year. |
That split answers the February decision before it arrives. It keeps the unit that can actually earn in short-term letting, stops the bleeding on the one that can't, and it does not require both bets to come off.
Her host profile lists her work as people development, which is the one other income stream visible from outside. This section cannot be finished without it and any others named and quantified. There are three specific questions rather than one:
One caution, stated plainly. Using profitable businesses to subsidise a loss-making lease is a decision with a deadline attached, not an open-ended arrangement. The February 2027 notice date on Colosseum is where that decision has to be made on numbers. Subsidising through to then in order to protect the peak season is defensible. Subsidising indefinitely because the exit is unpleasant to price is how one problem becomes two.
Named channels rather than categories. Verify each group's posting rules before the first post — promotional posts get members removed, and losing a group is harder to undo than waiting a day.
These book whole apartments for a month or more, at a fixed rate, with one point of contact and no guest acquisition cost. They are the highest-value version of the mid-term strategy.
South Africa introduced a digital nomad visa in 2025, which is steadily increasing the pool of people looking for one-to-six-month furnished stays in precisely these two neighbourhoods. The mid-term demand is structural rather than a fad, which is what makes it worth building for rather than dabbling in.
Before building anything, audit what contact detail already exists — WhatsApp threads, email, enquiries that never converted, past direct bookings. 58 stays have happened. Some of those people are repeat visitors to Cape Town and a handful will book again if asked. This is the cheapest revenue in the entire plan: a flyer with a defined offer, a real deadline, and a direct rate that beats what they paid on Airbnb while still netting her more.
This is lawful under POPIA, with one condition. Past guests are existing customers, so section 69's opt-in requirement doesn't bind her: she may market her own similar services to people whose contact details she obtained in the context of a sale, provided each person has had a reasonable opportunity to object, free of charge and without unnecessary formality. In practice that means every message identifies her, gives contact details, and carries an easy opt-out — and any opt-out is honoured immediately. People who only ever enquired and never stayed are not customers, and they need consent first.
The platform line she must not cross. Do not use Airbnb's messaging to solicit direct bookings, and do not offer an off-platform rate to a guest with a live Airbnb enquiry. Diverting bookings off the platform is a clear terms violation, it is detectable in message content, and she is not in a position to risk a listing. Reaching a past guest on an email address or WhatsApp number they gave her directly, after their stay, is a different thing — and the reason she is on the right side of it is that Airbnb never gave her those details in the first place.
Direct bookings with deposits solve the immediate cash-flow problem, which is why they are worth doing now. Two conditions, both learned from what has already gone wrong this year.
Airbnb masks guest contact details and they stop being retrievable after a period, so anything not captured at the time is gone. This is a standing process, not a one-off export.
These three are in the All-in tier for a reason: they take months to produce anything and they depend on the guest list above existing first. Sequence matters. The list is the asset; the website is where it converts; Instagram is what feeds it; the referral WhatsApp group monetises the relationships in the buildings. Building the website before there is a list to send to it is the common and expensive mistake.
The one exception worth doing early is a single landing page with real photos, the direct rate, and a way to enquire. It costs a weekend, it gives the flyer and the group posts somewhere to point, and it works long before a full booking engine does.