DIRECTIONAL — no registered model · 8 September 2026
A new model, with the original preserved
Model 02 was built afresh using the property-development-finance-expert workflow alongside finance-scenario. It is not a renamed copy or an approved feasibility. Case A holds the original 40-key hotel and 1,000 m² office GLA scope and starting assumptions constant. Case B separately costs a 24-key hotel and 600 m² office first phase.
Open expert-built model 02 ↗ · Download model 02 (Excel) · Download original model 01 (Excel)
Both downloads are public frozen copies dated 8 September 2026. The original Google workbook is unchanged. The new Google Sheet is publicly readable; download a copy to change inputs. Read the architect brief (PDF).
The comparable result is unchanged
At constant September 2026 prices, both same-scope cases give R140.196m pre-finance cost, R3.379m annual net operating income and 2.41% yield, excluding land and VAT. The original published 2.07% mixed starting-price income with future escalated cost. It is not directly comparable with the new nominal ratio.
Model 02 Case A gives R164.245m nominal pre-finance cost, R4.031m annualised full-ramp NOI and 2.45% yield on nominal cost. It applies 6% construction escalation to actual monthly spend and a new, unverified 4% revenue/fixed-cost growth hypothesis. This is a change in calculation basis, not evidence of improved demand or profitability. See Comparison and Summary!B13:B25.
What the expert-led rebuild adds
- 240 monthly periods, dated S-curve construction spend, preconstruction fee timing, opening ramp and actual operating days.
- Monthly loan draws, equity-paid construction interest, amortisation and opening-year liquidity. The failed 60% loan-to-cost screen remains visible; the main cash flow uses a separately income-capped loan.
- A separately priced smaller phase retaining common infrastructure and fixed operating costs until savings are evidenced. Smaller does not automatically mean viable.
- Live demand, cost and funding sensitivities; labelled scenario snapshots that flag stale when inputs change.
- Separate VAT exposure and land opportunity-cost stress, plus explicitly hypothetical terminal-value return diagnostics.
The smaller phase needs a different operating proposition
Case B costs R109.704m before finance and produces only 0.20% nominal yield with the full common infrastructure and fixed hotel budget retained. Reducing that fixed hotel budget to 60% of Case A lifts yield to about 1.94%. This is an unverified staffing sensitivity, not an achieved saving. See Summary!C14:C25 and Live_Levers!A18:B21.
Funding does not rescue the operating case
Case A’s hypothetical 60% loan-to-cost debt gives only 0.29x stabilised debt-service cover. The selected income-capped loan is about R22.12m; even this has only 0.59x opening-year cover during ramp-up. Peak cumulative equity is about R145.75m excluding VAT and land. See Summary!B27:B38.
The illustrative project XIRR is −5.05%, excluding land, VAT and tax. This assumes a hypothetical marked-value realisation after ten operating years at a 10% capitalisation rate, less 2% realisation costs. It is not a sale recommendation, a valuation of the whole holding or a verified exit. No undeveloped residual land value is included. See Summary!B40:B46.
What to test next
- Obtain operator-funded / ground-lease terms and compare capital, income surrendered, infrastructure obligations and retained control.
- Obtain achieved hotel trading and a costed staffing model. Even 60% occupancy and R1,800 starting ADR gives only about 5.60% nominal yield in Case A.
- Ask the architect, QS and engineer for genuinely standalone phases and a reconciled hotel cost-per-key / area schedule.
- Compare demand-backed professional accommodation and conditional residential only once rights, access and servicing dependencies are priced.
Neither priced own-build case supports commitment under the provisional 10% screening hurdle. Unpriced alternatives are not proven winners. The eight-week programme and architect brief remain the evidence-gathering route.
Evidence and checks
Sources: current Haarties status and model registry; original workbook Inputs, Cost, Operations and Funding; the current programme, architect brief and SG/zoning source pack. No new achieved trading, QS quotation, tax opinion or lender offer was introduced. Site/legal/services gates remain unresolved.
68 summary outputs matched a separate Python implementation; monthly debt and timing controls passed. Reversible input-change tests confirmed both cases and live levers recalculate, and that scenario snapshots flag stale. Seven spreadsheet review pages were visually inspected. This was a sequential, non-independent review, not professional sign-off.