Sensitivity Analysis on a Property Model: Choosing What to Flex

How to run sensitivities that inform a decision rather than fill a page, from choosing which inputs deserve a range to presenting the result in a form people act on.

Sensitivity analysis identifies which assumptions actually drive an outcome and how far each can move before the decision changes. Done well it concentrates diligence spending where it matters. Done badly it produces a grid of numbers that nobody reads, usually because every input was flexed by the same arbitrary percentage. The steps below produce the first rather than the second.

Before stage one

Sensitivity work is only meaningful on a model whose structure is sound, because flexing an input in a model with hard-coded numbers will quietly fail to propagate. Confirm that inputs are separated from calculations and that the checks reconcile before you begin, otherwise you will be testing the spreadsheet rather than the deal.

• A completed base case with every assumption sourced and dated.
• A clean input structure so that changing one cell flows through the whole model.
• A stated decision threshold: the return, cover or price at which your answer changes.
• Agreement on which outputs matter, so the sensitivity reports those and not everything.
• An honest list of which assumptions are evidenced and which are judgements.

Stage 1: Identify the candidate inputs

List every assumption that a reasonable person could set differently. Typically this includes achievable rent, vacancy, the cost base and its escalation, the capital schedule, the exit capitalisation rate, the letting period for vacant space, and the cost and amount of debt. Exclude anything fixed by contract, since flexing a contracted rent is not a sensitivity but a different deal.

• Deliverable: A candidate list separating assumptions from contractual facts.
• Owner: Analyst with the decision maker.
• Approval gate: Nothing contractual appears on the list of things to flex.

Stage 2: Rank by influence before building any grid

Move each candidate through a plausible range one at a time and record the effect on the outputs that matter. The result is a ranking, and it is usually dominated by three or four inputs while the rest barely register. This ranking is the entire point of the exercise: it tells you where diligence money and negotiating effort should go.

• Deliverable: A ranked list of inputs by effect on the key output, with the range used for each.
• Owner: Analyst.
• Approval gate: The ranking is produced before any two-way table is built.

Stage 3: Set ranges from evidence, not from symmetry

Flexing everything by ten per cent each way is convenient and wrong, because plausible ranges differ by input. The exit capitalisation rate might reasonably move a point in either direction; achievable rent might move much less; a letting period might plausibly double. Set each range from what the evidence supports and from how wrong similar assumptions have been before.

• Deliverable: A range per input with a one-line justification for its width.
• Owner: Analyst, challenged by whoever assembled the evidence.
• Approval gate: No range is set by applying a uniform percentage across different kinds of input.

Stage 4: Build two-way tables only for the top pair

Take the two highest-ranked inputs and build a grid showing the key output across both ranges. Two variables are usually enough to show the shape of the risk and few enough for a reader to absorb. Mark the cells where the decision changes, because the reader's eye should go to the boundary rather than to the numbers.

• Deliverable: A two-way table with the decision boundary marked clearly.
• Owner: Analyst.
• Approval gate: The table reports the output that drives the decision rather than every available metric.

Stage 5: Build a small number of named combined cases

Single-variable sensitivity understates risk because adverse events correlate. Build three or four named cases instead: a base case, a downside in which rent, voids, costs and the exit yield all move adversely together, an upside, and where relevant a case built around a specific risk such as the loss of the largest tenant. Each should be a story someone can recognise rather than an arbitrary combination.

• Deliverable: Three or four named cases with their assumptions listed on one page.
• Owner: Analyst with the decision maker.
• Approval gate: The downside case moves several assumptions at once and is described in plain words.

Stage 6: Find the break points rather than reporting outcomes

Turn the analysis around and solve for the point at which the answer changes: the rent at which cover fails, the occupancy at which cash runs out, the exit yield at which the required return is missed, the delay at which the project stops working. Break points are far more useful than outcomes, because they can be compared directly against evidence and against the asset's own history.

• Deliverable: A break point table stating the level at which each key test fails.
• Owner: Analyst.
• Approval gate: Every break point is compared against a historical or evidenced figure for that asset.

Stage 7: Translate the analysis into actions

A sensitivity that changes nothing was wasted effort. Each influential input should produce an action: more diligence on the rent evidence, a retention negotiated from the price for the capital works, a lower loan amount to lift cover, a longer assumed letting period, or a condition in the contract. Write the action beside the finding.

• Deliverable: An action list linking each sensitivity finding to a specific step in the transaction.
• Owner: Decision maker.
• Approval gate: Every input in the top three has an action or an explicit decision to accept the risk.

Stage 8: Present the analysis in one page

The output for a reader is short: the ranked drivers, the two-way table with the boundary marked, the named cases, the break points, and the actions. Everything else stays in the file. A committee that receives twelve tables will read none of them, while a committee that receives one page with a marked boundary will ask the right question.

• Deliverable: A one-page sensitivity summary attached to the investment memo.
• Owner: Model owner.
• Approval gate: The summary can be read and understood in under five minutes.

Handovers between stages

The analysis has two audiences and they need different things. The diligence team needs the ranked drivers, because that tells them where to spend the budget. The decision maker needs the break points and the named cases, because those answer whether to proceed. Send each the part they will use, and keep one file that contains everything so the two versions never drift apart.

How you know it is finished

The work is complete when you can state, in one sentence each, the three assumptions that carry the decision, the level at which each one breaks the deal, and what you are doing about them. If the break points sit inside the range the asset has actually experienced in the past, the deal needs restructuring rather than more analysis. Assumptions age, so date the analysis and refresh it before any decision taken significantly later.

Domira builds the numbers on a Saudi property with you before you commit: rent and cost evidence, condition and capital expenditure, the appraisal and the sensitivities that decide whether the deal survives a bad year. The investment decision stays yours, and so does the advice of your licensed tax and legal advisers.