Eighty-two per cent is a sentence, not a number. Ask an owner where it came from and you get one of two answers: the rent that came in this month, or the manager's feel for how full the place looks at dinner. Neither of those is occupancy. One is a lagging financial indicator and the other is a mood.
This matters because of how a PG actually loses money. It is almost never a dramatic emptying. It is one bed on the second floor that emptied on the 3rd, was never advertised because nobody wrote it down, and got refilled on the 5th of the following month. That bed did not cost you two days of rent. It cost you a full month and two days. And by the time it shows up anywhere, it shows up as a rent total that looked slightly small, which everybody explains away in four seconds.
Rent collected tells you what happened. Bed occupancy tells you what is about to happen. Most PG owners only track the first one.
Occupancy is a bed-level number, not a building-level feeling
The unit a tenant occupies is a bed. Not a property, not a floor, not a room. Everything downstream — the rent, the agreement, the deposit, the notice period, the vacancy — belongs to a bed, and a system that stops at the room is already lying to you.
Take a triple-sharing room with two tenants in it. In a room-keyed rent sheet that room is let. In reality it is 67 per cent occupied and one-third of its earning capacity is sitting there costing you electricity. Repeat that across a twelve-room property and the gap between what is let and what is occupied is an entire month of somebody's salary.
So the inventory has to run three levels deep — property, room, bed — with a label a human can say out loud. Bed 3, Room 204, Kothrud. The day there is a dispute, the argument is never about the room.
- Which property and which room the bed sits in, with its sharing type — single, double, triple, dorm, studio
- Who is on it today, from what date, under which agreement
- What that person actually pays, which is not always what the room card says
- When they are due to leave, if they have told you
- Whether the bed can be sold at all right now
The three states a bed can be in, and the one everyone forgets
Most owners run a two-state model in their head: occupied or empty. That model is why beds go missing. There are four states, and the damage lives in the two nobody records.
- Occupied — a tenant is on it and it is earning.
- Vacant — it is empty, it is sellable today, and every day it stays like this is a day of rent you will not get back.
- Reserved — you promised it to someone on a phone call. It is not earning and it is not available. In most PGs this state exists only in the manager's memory.
- Under maintenance — the cot is broken, the wall is damp, the fan is gone. This is not vacancy you can sell, and counting it as vacancy makes your fill rate look worse than it is.
The forgotten one is reserved. A reserved bed is the most expensive object in a PG: it earns nothing, it blocks a paying enquiry, and nobody is accountable for it because it was never written anywhere. Somebody said yes on a call three weeks ago, the prospect went quiet, and the bed is still being held out of a sense that it would be awkward to sell it now.
A bed you promised on a phone call is neither vacant nor occupied. It is invisible. Invisible beds are where lost months and double-allotments both come from.
Notice periods: knowing about a vacancy before it happens
A vacancy you learn about on move-out day is already a loss. A vacancy you learn about thirty days out is a task. The whole point of a notice period is to convert one into the other, and almost every PG wastes that conversion because notice is a conversation rather than a state.
The tenant tells the warden on a Tuesday. The warden tells the owner on Friday if he remembers. Nobody advertises the bed, because the bed does not look empty yet — it looks occupied, right up until it looks empty and stays that way.
Treat notice as a lifecycle stage the tenant record actually moves through: enquiry, active, on notice, vacated. The moment a tenant enters notice, their bed should appear on a forward-dated availability list with the exact date it frees up. That list is what you send to brokers. That list is what you show a walk-in who wants a bed from the 1st. It is the single highest-value screen in a PG and it costs nothing to maintain except the discipline of recording the date somebody told you they are leaving.
It settles the other half of the notice argument too. If the agreement carries a notice period and the tenant leaves short of it, that is a known deduction on a known date rather than a shouting match on the stairs.
Turnaround: the days between move-out and move-in that eat your margin
Turnaround is the number of days between one tenant leaving a bed and the next one sleeping on it. Almost no PG measures it. It is the most controllable loss in the building.
The arithmetic is unforgiving because rent is monthly and vacancy is daily. A bed at 9,000 rupees a month is 300 rupees a day. Say ten beds turn over in a month and each one sits idle nine days between tenants — cleaning, a mattress that needed replacing, a key nobody could find, a deep-clean that waited for the weekend. That is ninety bed-days, which is 27,000 rupees, which is a month of one staff salary, gone into nothing. Run those numbers on your own rent and your own turnover before you decide it is a small problem.
Once you can see turnaround per bed, the fix is usually boring and immediate: start the cleaning and repair the moment notice is given rather than the moment the tenant leaves, keep one spare mattress and one spare lock, and stop letting a Sunday move-out mean a Wednesday relist.
Double-allotment, and the WhatsApp message that caused it
Here is how it happens, every single time. On Wednesday the manager tells a walk-in that bed 3 in 204 is free from the 1st. On Thursday the owner tells a broker the same thing, because the owner is looking at the same mental picture and nobody updated it. On Sunday two people arrive with bags and both of them have a WhatsApp message that says yes.
There is no version of this that ends well. You refund one of them, you lose the referral, and if the loser had already paid a token you have also just taught them that you are disorganised with their money. Meanwhile the bed sat blocked for a week for a tenant who never existed as a record.
The root cause is not carelessness. It is that two people were allowed to commit the same inventory because the inventory had no single copy. The fix is structural: one bed record, one status, changed by whoever commits it, visible to everyone else the moment it changes. Reserving a bed should be an action against the bed, not a sentence in a chat.
Rent that follows the bed, not the room
Room-keyed rent sheets break for a reason that has nothing to do with software: two tenants in the same room routinely pay different amounts. One joined at last year's rate. One opted for meals and one did not. One took the AC room surcharge, one is on the lower bunk at a discount, one has a laundry add-on. The room has one number on the notice board and five different realities inside it.
So rent belongs on the agreement — the record that ties a tenant to a bed with a move-in date, a monthly rent, a deposit and a notice period — and the add-ons belong on the invoice as their own lines: rent, food, maintenance, utilities, and whichever amenities that tenant actually took. When you bill that way, two things stop happening. Nobody argues about what they were told at move-in, because the agreement says it. And nobody has to reconstruct a tenant's history from a year of chat messages when they leave.
It also changes collection. If each agreement carries its own rent and due date, the monthly invoice writes itself, and a tenant on a UPI AutoPay mandate has the rent debited on the due date with a numbered receipt against it instead of being chased on WhatsApp on the 7th, the 9th and the 12th. A collection process that runs on the owner's willingness to nag is a process that quietly gets softer every month the owner is tired.
There is a side effect worth naming: when rent is per-bed and invoiced, an empty bed stops being invisible. A bed that should have produced an invoice and did not is a hole in a numbered sequence, and holes are visible in a way a slightly-small bank balance never is.
Seasonality: exam season, job season, and planning for both
Every PG has a shape to its year and almost no owner can state it in numbers. A student PG near a coaching cluster empties around exams and refills before the new session. A working-professional PG churns with the hiring cycle and again after appraisals, when people who got a raise go looking for a 1BHK. Both empty out around the big festival travel weeks, and both get a wave of enquiries the week after.
You cannot plan for a season you have never measured, and memory is a bad instrument — owners remember the year June was terrible and forget the three years it was fine. What you need is dull and cumulative: bed-nights occupied, month by month, for one full year. After one year you know whether your June is genuinely weak or just vivid.
- Set the notice period so that departures cluster before your strong intake weeks, not after them
- Do deep maintenance in your known trough instead of during peak fill
- Push renewals a month before the season that historically takes your tenants away
- Discount early and briefly in a known weak month rather than late and permanently across the year
The weekly number an owner should actually look at
Not the occupancy percentage. That number is an average, and averages are where problems go to hide. Five things, once a week, ideally on the same day so you build a series rather than a snapshot.
- Beds occupied today over beds sellable today — with maintenance beds pulled out of the denominator and listed separately, so a broken cot cannot masquerade as weak demand
- Beds on notice, each with its vacating date — this is next month's revenue risk, visible today
- Beds reserved, each with the date the tenant is supposed to move in and the name of whoever committed it — anything reserved more than a few days without a token is fiction, release it
- Average turnaround days for every bed that turned over in the last thirty days, and the worst single one
- Rent due versus rent collected this month, and the count of tenants overdue rather than the amount — one large overdue and eleven small ones are different problems
Five numbers, and you can read them in the time it takes to finish a cup of tea. Do this for two months and you will find at least one bed that has been quietly out of service since a repair nobody finished, and at least one that has been reserved for a person who stopped answering their phone in April.
None of this requires a system when you run one property with eighteen beds and you are on site every day. You are the system, and you are a good one. It stops working at the second property, or the first month you are away, or the day your manager leaves and takes the whole picture with him — because none of it was ever written down at the level where the money actually lives.
The test is simple. Ask, right now, without calling anyone: how many beds are empty tonight, and for each one, since which date. If the honest answer takes more than a minute, you are not running at 82 per cent. You are running at a number you have not met yet.