The branch P&L lands on the 14th. Twenty-four files funded, volume up over the prior month, and the number at the bottom is smaller than the month you funded nineteen. Your regional calls it a good month. You spend the drive home trying to work out which line moved, and by Thursday you have stopped trying.
That gap — quoting your locked pipeline to the dollar while being unable to explain your own bottom line — is not a math problem. It is a reading problem. A branch P&L is four layers in a fixed order. Walk them top to bottom and most of the mystery goes away.
The short version
- Four layers, always in this order: revenue in basis points on funded volume, origination comp, operating cost, branch net.
- Two of those layers move with volume. Two bill by the calendar. Confusing them causes almost every bad read of a month.
- Compute your break-even file count — monthly fixed cost divided by contribution per file. It belongs on a wall.
- Revenue lands at funding, not at lock. This month's P&L is grading last month's work.
- Write expected revenue on every file the day it locks, log the actual at funding, and categorize every gap.
Layer one through four, in a fixed order
Strip away the corporate template and every branch P&L reduces to the same stack: revenue, origination comp, operating cost, branch net. Nothing else is load-bearing.
Run one branch all the way through so the numbers stay attached to something. Say your branch funds $6.0M across 24 files — a $250,000 average loan — and corporate credits you 250 basis points of revenue. Your LOs sit at 110 bps. Fixed cost, everything in, is $60,000 a month.
Revenue: basis points on funded volume, recognized late
Branch revenue is priced in basis points on funded loan amount. At 250 bps, a $300,000 loan books $7,500 and a $150,000 loan books $3,750 — for the same application, the same conditions, the same processor hours, the same closing. Loan-size mix swings a month harder than unit count does, and nobody in the building feels it happening.
Three things about the top line that get underweighted:
Locked is not funded. Your lock desk report is a revenue forecast multiplied by pull-through, and pull-through is never 100%. Cancellations, denials, the borrower who goes quiet, the file that dies at appraisal — none of that revenue exists. Read the locked pipeline as booked revenue and you will spend money you have not earned.
Revenue lands at funding. A purchase file typically runs 30 to 45 days from application to closing. The month your team ground through conditions is not the month the revenue shows up. Praise and panic on the current number and you will do both at exactly the wrong time.
Processing income is a real line. If the branch charges a processing fee and employs its own processors, the fee is branch revenue and the salaries are branch cost — and they belong on the same page, so you can see whether fulfillment pays for itself. Let corporate net those out quietly and you lose that visibility permanently.
The line that scales and the line that just ticks
LO comp is the cleanest line you have. Comp plans are set in advance in basis points and, by rule, cannot vary deal by deal with the terms of the loan. So origination comp tracks funded volume in a near-fixed ratio. Revenue minus LO comp is your contribution: 140 bps here, or $3,500 on a $250,000 file. It should be boringly stable, and when it is not, something specific is eating it.
Overhead is the opposite animal. Ops payroll, rent, the tech stack, licensing, E&O, and the corporate allocation all bill by the month whether you fund thirty files or twelve. Overhead has no per-file cost. It has a monthly cost that gets divided by however many files you happened to fund.
The allocation is the line most managers never interrogate. Three questions, and you are entitled to ask all of them. What is the allocation actually buying? Is it a flat monthly number or a rate on volume? And if it is a rate, at what volume does it step? Get those in writing once and you can model your own year instead of receiving it.
Know your break-even file count
Here is the most useful number on the page, and it is not on the page — you have to compute it.
Break-even file count = monthly fixed cost ÷ contribution per file. In the running example, $60,000 ÷ $3,500 = 17.1. Call it 18 files. Every file up to the eighteenth pays rent. Every file after the eighteenth is nearly pure margin, because the office is already bought.
That is the operating leverage in this business, and it explains the month you could not read. Twenty files and thirty files are not a 50% difference in branch net. They are more than a four-times difference. It also reframes a recruiting conversation: an LO who funds four files a month is not a quarter of a producer once you are above break-even. Those four files land almost entirely on the bottom line.
Post the number. Ops should know the file count where the branch starts making money, and where the month stands against it on the 20th — not on the 14th of the following month, from a PDF.
The estimate-versus-actual ledger
Contribution per file is where branches quietly bleed, because the number you plan with is the number at lock and the number you get is the number at funding.
Between lock and clear-to-close, small things happen. A pricing concession to save a deal. A lock extension because the file sat waiting on a condition. A tolerance cure at the table. A fee waived to keep an agent happy. Each is defensible on its own. Each is a person being reasonable under pressure.
Fourteen percent off one file's top line does not feel like a crisis. Run it across the book and break-even moves from eighteen files to twenty-four — six extra files a month just to stand still, and nobody ever wrote that down anywhere.
Then there is the one that arrives late. An early payoff inside the EPO window claws commission back months after funding, and if it books as a general branch expense, the file that caused it stays profitable in your records forever.
The discipline is four steps, and it is mostly about refusing to cheat:
- Write the expected revenue once, at lock. Lock is the natural fixed point.
- Never revise it. The estimate is a record, not a forecast to be managed.
- At funding, log the actual next to it.
- Categorize every gap: concession, extension, cure, waiver, clawback.
Do the same on the cost side — budgeted overhead against actual, monthly. None of it is sophisticated. It is bookkeeping with a memory, and most branches skip it because the estimate lives in the LO's head and the actual lives in a corporate report three weeks late. That is the argument for keeping the P&L behind the same login as the pipeline, which is how MAVYN handles it: financials render for owner and sales-manager seats only, enforced at the database row level, because the branch book is not the LOs' business.
A monthly average tells you what happened. A per-file number tells you where.
Two hours, Monday morning
Pull last month's funded list. For every file, write four numbers: actual revenue, LO comp, contribution, and the expected revenue from lock day. Sort by contribution and sit with the bottom five — ask what they share by source, by product, by originator, not by name. Then take the five files where actual came in furthest under expected and get a one-line reason for each. The file-level mechanics are worked through in per-loan profitability.
One list, one morning. You will know more about your branch's economics than the monthly package has told you all year.