How to justify the rent you put into your model

In an investment calculation almost everything comes from documents: the price from the contract, the payment schedule from the same place, building charges from the management company's tariff. Only one figure you supply yourself, out of your head — the rent. It multiplies through every year of ownership, so it drives the result more than any other line. This page is about turning it from an opinion into a number you can defend. Where to look for demand data and tenant segments is covered in the guide to rental demand in Phnom Penh; here we work with observations you have already gathered.

What "a defensible rent" means

Defensibility is decided not by the source and not by how neat the number looks, but by whether you can explain where it came from to an outsider. A defensible rent has three attachments: the list of comparables it was derived from, the named adjustments bringing them to your unit, and the class of evidence behind each observation.

It pays to label the classes explicitly, because in a spreadsheet they all look the same — just numbers in a column.

The gap between asking and achieved rent is its own important subject, covered in the guide to rental demand. For modelling purposes one practical consequence is enough: observations of different classes must not be averaged together. A mean of two leases and three listings is not a rent but a blurred quantity of unknown meaning.

Selecting comparables: narrow first, widen second

Instinct says start broad to gather more observations. The correct order is the reverse: start with the narrowest possible circle and widen it one attribute at a time, noting every departure.

1. Same building, same layout type. The ideal comparable is the neighbouring unit of the same configuration. Adjustments are minimal, and the building, security, lifts and amenities are identical by definition.
2. Same building, different type. An adjustment for size and bedroom count appears.
3. Neighbouring building of the same class and age. Adjustments for common facilities and the building's reputation are added.
4. Same tenant segment elsewhere. The weakest level: only the demand profile matches.

The attributes to check on every observation: type and number of rooms, floor area, floor level and outlook, condition of the finish, presence and quality of furniture and appliances, lease length, date agreed. And separately what is included in the rent: utilities, internet, cleaning, parking, servicing. That last item is the usual explanation for "strange" gaps between two apparently identical units.

One warning: competing units inside your own building are not comparables — they are competitors. Identical units in one building listed at the same time compete for the same tenant, and that shows up not in the rent in your table but in the time it takes to let.

Normalisation: bring everything to one base

Before adjusting, observations must be brought to a single definition of rent, otherwise you are comparing different things. There are three axes.

• Services included. An all-inclusive rent and a rent plus utilities are different quantities. Bring every observation to your own convention, adding or removing the cost of included services at tariff rather than by feel.
• Term. A monthly rent under an annual lease, under a six-month lease, and nightly letting are three different markets with different cost structures. They cannot share a table even with adjustments.
• Concessions at signing. Rent-free months, a discount for the first period, a paid move or part of the furnishing at the owner's expense all reduce the effective rent while leaving a flattering figure in the contract. The method for converting to an effective rent is set out in the guide to rental demand.

Adjustments: direction matters more than size

Ready-made coefficients do not exist, and any table of "plus so many percent per floor" is invented unless it was derived from your own observations. What works instead is the paired comparison: take two observations differing in exactly one attribute, and the difference between them is your adjustment — for your building and your segment. Where no pair exists the adjustment stays qualitative: you know the direction but not the size, and that should be recorded honestly.

Two rules for working with adjustments. First: an adjustment you cannot explain in one sentence should not be applied. Second: if the total adjustments to an observation exceed the spread between observations, the unit is not comparable — exclude it rather than stretching it into place.

Getting to the final number

1. Normalise every observation to your own definition of rent.
2. Apply the adjustments and record each one on its own line, not as a single total.
3. Drop observations whose adjustments turned out excessive, and drop listings altogether if you have at least three confirmed observations.
4. Look at the spread of adjusted values. A narrow spread suggests a homogeneous set; a wide one suggests you have mixed different things and should return to selection.
5. Take the median rather than the mean: the mean is too sensitive to a single outlier, and small samples never lack outliers.
6. Shift the result down according to the margin rule below.
7. Write an assumption card: the number, the class of evidence, the count of observations, the range, the date.

The quality of your own data

Before the number goes into the model, judge the set it came from. Four questions, each needing an answer recorded next to the figure.

• How many observations. One is an anecdote, two a coincidence. A meaningful conclusion starts where several independent observations say the same thing.
• What share is confirmed. A set of listings with one signed lease and a set of signed leases with one listing give very different confidence at the same average.
• How independent they are. Five units let by one agency in one building on one template are effectively one observation, not five.
• How recent they are. A lease signed long ago describes the conditions of that moment, not today's. An old observation is not corrected — it is downweighted.

The output of this assessment is not a verdict of good or bad, but a decision about which point in the range you take into the model.

When comparables are scarce

This is the normal situation for a new building, an unusual layout, or a segment where lettings are simply infrequent. The bad answer is to take the single number you found and call it the rent. The workable approaches are these.

1. Widen in steps and record which step you stopped at. Whoever reads your model — including you a year from now — should see that the support is not from the same building.
2. Anchor on the tenant profile rather than geography. If you know who will rent it, look for comparables where that type of tenant lives, even in another part of the city.
3. Use indirect signals. How fast similar units let, how many stand empty in the building, how neighbouring owners behave on price — none of this gives a rent, but it tells you which half of the range you are in.
4. Test with demand. If the unit can already be offered, a trial listing is the most honest source: the response to a specific figure in a specific week beats any reasoning.
5. Move from a point to a range. With a thin set the rent is an interval; the lower part goes into the model, and the decision is tested for robustness at a value lower still.
6. Shorten the horizon of the assumption. Justifying the rent for year one is easier than for year five. For distant years it is more honest to hold the same rent flat than to build in growth you cannot evidence.

How much margin to allow

There is no universal percentage, and any figure named here would be invented. There is a rule: the weaker the evidence, the lower the point in the range you take. A dense set of confirmed leases lets you take the middle. Mixed data, mostly listings, or isolated observations — the bottom only.

Test whether the margin is sufficient not by comparing yields but with a single question: how far can the rent fall before it stops covering costs. That level is calculated in the guide to rental break-even. If the gap between your rent and that level is narrow, the margin is inadequate however carefully the comparables table was assembled.

And separately: a margin on the rent does not substitute for a void assumption. Being wrong about the price of a month and being wrong about the number of paid months are two different errors, and covering both with one reduced figure means losing track of what you are actually assuming.

An illustrative comparables review

The figures below are notional units chosen for the arithmetic. They are not rents in Cambodia, not market data and not a guide to the size of adjustments: in a real calculation adjustment sizes are derived from your own paired comparisons.

Subject unit: two bedrooms, mid floor, furnished, twelve-month lease, utilities paid by the tenant.

What happens next. Observation C is a listing and the only one standing out on the upside; with three leases available it cannot be the backbone, though it usefully marks the upper bound of asking. That leaves 102, 97 and 98. The median is 98, the spread is narrow, all three are in the same building, two of the three are recent.

The model takes 97: the set is small, one observation is over six months old, and at that level of data quality you take the lower part of the range rather than the middle. The assumption card reads: rent 97 notional units per month; three confirmed leases in the same building; adjusted range 97–102; most recent observation two months old; checked July 2026. It is that card, not the number itself, that makes the assumption checkable — and it is the card you carry into the cash-flow table.

What to record beside the rent in the model

• The number and the currency.
• The class of evidence: leases, confirmations or listings.
• The count of observations and how many are in the same building.
• The adjusted range, not only the point.
• The date of the most recent observation.
• What the rent includes and who pays utilities.
• The lease length it refers to.
• On a separate line, the void assumption, so that it never merges with the rent.

From there the rent lives in two places: in the year-by-year cash flow, where it is multiplied by months, and in the stress test, where it is deliberately worsened to find the boundary.

Frequently asked questions

What makes a rental assumption defensible?

Not the source in itself, but whether you can explain where the number came from to an outsider. A defensible rent comes with three attachments: the list of comparables it was derived from; the named adjustments that bring them to your unit; and the class of evidence behind each observation — a signed lease, a deal confirmed by a manager, or an advertisement. Without those three it stays an opinion, however tidy it looks in a spreadsheet.

What counts as a comparable unit?

One let to the same type of tenant on comparable terms. The comparison covers not only size and bedroom count but what is included in the rent: furniture and appliances, utilities and internet, cleaning, parking, lease length and prepayment. The practical rule is to narrow first — usually to units inside the same building — and only then widen, noting every departure, because each one becomes an adjustment.

What if there are almost no comparables?

Widen the search in steps and record honestly which step you stopped at: the same building first, then neighbouring buildings of the same class, then the same tenant segment in another location. Each step weakens the conclusion and that should be written down. If the set is still thin, the rent stops being a point and becomes a range: the lower part goes into the model, and the decision is tested for robustness at a value lower still.

How much margin should you allow if you are unsure?

The margin is set by a rule rather than a universal percentage: the weaker the evidence, the lower the point in the range you take into the model. With signed leases and a dense set you can take the middle; with mixed data or mostly advertisements, take the bottom. And test the margin not by comparing yields but by asking how far the rent can fall before it stops covering costs.

Sources

NovAsia editorial corpus on letting and on preparing an investment decision · practice supporting owners in Phnom Penh · checked July 2026. There are no market indicators on this page: rents, yields, vacancy levels and market adjustment sizes are not quoted, and the numbers in the illustrative section are invented and expressed in notional units. Adjustment sizes for your own calculation are derived from your own observations and must not be carried over from this page. This content is for general information only, is not individual investment, legal or tax advice, and contains no promise of results.