Deal Analysis Masterclass

A 1,133 sq ft flat in Prime Central London, asking £1,721/sq ft. On its own that number means nothing — the whole judgement is whether that price leaves enough room, after every cost between buying it and selling it renovated, to be worth the risk.

The one figure everything else gets measured against is the GDV comparable — what similar, already-renovated properties nearby have actually sold for, per square foot. For Eaton House that came from 24 real comparable sales, and the figure used is the 70th percentile of that set (a deliberately conservative choice — not the average, and nowhere near the top of the range), which comes out at £3,155.33/sq ft.

Buying the property is never just the purchase price. Stamp Duty Land Tax (SDLT) — a UK tax on buying property, paid by the buyer — is charged on top, and legal fees on top of that.

SDLT is charged as a flat rate for a company buying residential property without developer's relief: 17% of the purchase price under the current non-natural-person rate. (A different real deal, covered in Part II, shows a case where a lower rate genuinely applied instead — worth knowing SDLT is not always one flat number, before you assume it is.)

The renovation cost here is a blanket assumption — a single £/sq ft figure applied to every property the pipeline scores, rather than a real, itemised build cost. That is a deliberate simplification for screening purposes (it lets the pipeline score dozens of properties automatically), not a claim that it is precisely accurate for any one of them. Part II shows what a real, line-by-line renovation budget actually looks like, on the exact same £/sq ft basis, for comparison.

Three pieces stack up: the net renovation cost itself, a contingency on top (money set aside for the renovation running over budget — standard practice, not optional), and VAT (UK sales tax, 20%) charged on the net cost plus contingency together.

Almost nobody buys and renovates a £2m+ property in cash. Most of the cost is borrowed, through a development finance facility — a loan specifically for buying and renovating a property, repaid from the sale proceeds.

The size of the loan is set by loan-to-value (LTV) — the percentage of the total cost the lender is willing to fund. The pipeline's default is 65%, meaning the borrower has to fund the remaining 35% themselves (this remaining share is the equity — more on that below).

On top of the loan itself, two further costs: finance fees (what the lender charges to arrange the loan — a percentage of the loan amount, plus a flat valuation and legal fee), and finance interest (the cost of borrowing the money for the term of the loan — here, a flat-rate approximation: the full loan amount, at an annual rate, for the fraction of a year the loan runs).

The gross development value (GDV) is the comp exit value applied to the property's size — what it should sell for once renovated, before any selling costs. From that, subtract the agent's fee (a percentage of the sale price, paid to whoever sells it) and exit legals (conveyancing costs on the sale) to get the net development value (NDV) — what actually lands in the bank.

Profit on its own — NDV minus total cost — tells you almost nothing, because it says nothing about how much was risked to earn it. A £445,000 profit on a £50,000 outlay is a very different proposition from the same profit on a £5m outlay. Four different margins answer four different versions of "was it worth it," because each divides profit by a different denominator.

Every number below is real — reproduced exactly from the pipeline's stored output for this property. Work through it in order; each step uses the answer from the one before.

Everything above is self-consistent — every number follows correctly from the one before it, and it reproduces exactly what the pipeline actually stored for Eaton House. That consistency proves the arithmetic is right. It proves nothing about whether the assumptions are right, because a formula can be applied perfectly and still rest on a simplification that doesn't hold in reality.

The only way to find that out is to test the same formula against a deal that has already actually happened, and compare the model's answer to what really occurred. That test was run once, against a real completed Mayfair deal — 9 Mount Row — covered fully in Part II. The result:

Notice the shape of that result: net profit % came out close. Return on equity came out a third too high. That's not a random error — profit and GDV, the two things net profit % depends on, were both roughly right; equity, the one thing unique to ROE, was the piece that broke.

Total cost was only 2.9% off — remarkably close. Equity was nearly 20% short. Three assumption differences, individually unremarkable, compound to produce that gap:

That last point is the one that actually moves ROE. The same 85.79% ratio, applied to a bigger base, produces a loan that's close in raw pounds to the real one (£6.76m modelled vs £6.57m real) but represents a larger share of the pipeline's own total cost: 78.96% geared, against the real deal's effective 74.52%. A few points of extra gearing doesn't move total cost much — debt is still debt, whichever side of the ledger it sits on — but it comes straight out of equity, the smaller number being divided into. Profit moved up a little (leverage cuts both ways); equity moved down by a fifth; divide one by the other and a 4-point gearing difference becomes a 32-point ROE difference.

The lesson isn't "the loan assumption was wrong," on its own. It's that ROE is the margin most exposed to compounding, second-order assumption drift — three unrelated choices, each defensible alone (how you treat a tax relief, which comp source you trust, how you back-solve a missing input), stacked on the one ratio that has both a shrinking numerator's-worth of leverage and a shrinking denominator running through it at once. Net profit % barely noticed. ROE moved by a third. That is exactly why a model that reproduces itself perfectly can still mislead on the one number everyone actually asks for first.

Eaton House's £445,426 profit sits at a single point — every input at its best estimate. Real deals don't happen at a single point; costs run over, sale prices move, interest rates change. Stress-testing the deal means deliberately flexing those inputs to see what breaks it, and what doesn't.

Run separately, after the original underwriting (this was genuinely later analysis, not part of the original pipeline output — worth being precise about that distinction when describing your own work), flexing build cost, exit value and interest rate together in different combinations:

The most useful finding isn't the range itself — it's which variable actually drives the risk. A 10% fall in exit value costs 79% of the profit. A 10% overrun in build cost costs 13%. A 200 basis-point rate rise costs 12%. Renovation is only £396,550 of a £3,061,213 total cost — the deal's risk sits almost entirely in whether the purchase price is far enough below the comp ceiling, not in whether the build runs to budget. That's a judgement about entry pricing, not construction risk, and it's the same judgement a lending or private equity desk makes before it commits to a structure.

Eaton House's renovation cost was one number — £350/sq ft, applied blanket, because the pipeline has to score dozens of properties automatically and can't cost each one properly. A real deal doesn't get that shortcut. 9 Mount Row's renovation cost is built from sixteen trade categories, each priced separately by whoever actually estimated the works.

On top of that: professional fees (£36,948) and monitoring fees (£16,700) — money paid to the people overseeing the build, not doing it — plus a separate, second breakdown of high-spec room-by-room allowances (Roof Terrace, Kitchen & Utility, Bathroom, Flooring, Doors, Partitions & Interior Architecture, Electrical, FF&E — furniture, fixtures and equipment) totalling £475,775, which is a specification-level cross-check against the trade breakdown above rather than a separate cost. Add it all together and the grand total, before VAT, is £1,172,589.

Worth sitting with that comparison for a second: the blanket assumption and the real, fully-costed build land within 5% of each other on this particular deal. That doesn't mean the shortcut is always safe — it means that on this one property, the average happened to hold. A screening tool that has to move fast across many properties can't itemise every one of them; the real question is whether the properties it flags as promising are then checked properly before real money moves, the way this workbook was.

Recall from Part I: a tranche is one slice of a loan, released at a different time, for a different purpose. Eaton House's model collapses the whole facility into one number, drawn on day one, for the whole term. Here is what a real facility actually looks like.

The £6 gap between the tranches and the headline figure is the workbook's own — it's labelled "£6 Difference, Tranche C" in the file itself, a rounding artefact its own builder flagged, not something to hunt for.

Tranche A and Tranche B — £5,434,950 combined — are the only tranches that are real cash to the project: money that pays for the purchase and the build. Tranche C (£131,414) is the lender's own arrangement and broker fees, 1% each of the gross loan, rolled into the facility so the borrower doesn't have to find that cash separately. Tranche D (£1,004,340) does the same for interest and exit costs — a reserve the facility carries for its own running cost, not money spent on the property at all.

The interest rate itself: 11% all-in, labelled in the workbook as "Apr '24 Base + 5.75%" — meaning a 5.75-point lender margin over the Bank of England base rate at the time, not the 3-point margin the automated pipeline assumes by default. That gap matters: it's part of why an attempt earlier in this material to reverse-engineer the pipeline's exact modelled result for this deal didn't converge — the real facility was priced on materially different terms than the pipeline's stated defaults.

£5,400,000 purchase price. SDLT of £721,250 — the figure from Part I's closing note, a correct banded calculation for a company claiming property-developer's relief, at rates in force before 31 October 2024, not the flat non-natural-person rate Eaton House's model assumes.

These are the real figures a formula got tested against in Part I, and came out badly wrong on ROE despite being close on net profit %. Everything in this section is the answer key that test was checked against.

The Analysis tab runs its own version of the same deal — the same purchase price, the same £3,500/sq ft exit assumption — and produces its own profit figure. Before being told the answer, work out whether it should match the Summary tab's £2,150,104, and if not, why not.

This tab splits the deal into £100,000 investment blocks and runs six alternate scenarios — sale prices from £9.5m to £12m — to show investors a spread of possible outcomes. Every scenario's net development value is calculated the same way. Find the fault in the formula before it's shown to you.

Every figure so far has answered some version of "what does it cost to buy and renovate this, and what will it sell for." The Rebuild Valuation tab answers something else entirely: if this building burned to the ground tomorrow, what would it cost to rebuild it from nothing? That's the figure a buildings insurance policy needs — the sum insured — and it has nothing to do with market value or renovation cost.

Notice this total — £1.5m to £1.8m — sits close to the renovation cost from earlier in this Part (£1.17m to £1.4m). That's a coincidence worth being alert to, not a relationship: one figure is what it costs to fully rebuild the structure from nothing (walls, floors, everything), the other is what it costs to renovate an existing building that's still standing. They answer different questions and happen to land in a similar range on this property; they wouldn't necessarily on another.

One more thing worth knowing rather than treating as settled: this tab's floor areas (26, 108, 66, 66, 55 m²) are rounded, while the Renovation tab's own CAD-laser-surveyed areas for the same floors are more precise (26.52, 101.86, 61.61, 56.72, 50.73 m²) and don't all match exactly. That could be a genuine small inconsistency between two tabs built at different times, or it could be that this tab was deliberately built on rounded, conservative figures for an insurance valuation, where a slightly higher sum insured is the safer error to make. Unlike the two audit exercises above, this one isn't resolved with the same confidence — flagged honestly rather than called either way.

817 sq ft, asking £1,400,000. One thing before you start: this property was scored on 28 July 2026, two days before the contingency default changed from 15% to 10% (the change Eaton House already reflects). Real pipeline output changes over time as its own assumptions get refined — this deal genuinely used 15%, not a mistake to correct.

1,475 sq ft, asking £1,500,000. Same 15%-contingency period as Hertford Street. No formula reminders this time — if a step doesn't come back to you, that's real information about which part of the chain to revisit, not a sign to guess.