How a subcontractor builds a cash flow forecast that is actually right

Almost every subcontractor forecast is a spreadsheet of monthly revenue with a payment assumption bolted on the end, and almost every one of them is wrong in the same direction: it says the money arrives sooner than it does.

That is not pessimism about your customers. It is structural. A specialty sub bills once a month on somebody else’s calendar, gets paid thirty to seventy-five days after that, and has five to ten percent of every dollar withheld until a job finishes that may be a year past your own last day on site. Meanwhile payroll runs every week.

A forecast that ignores that is a revenue projection, not a cash forecast. Here is the sub-side version.

Why revenue and cash are further apart for you than for anyone else

Three facts make a specialty sub’s cash position unusual, and they compound.

You bill on a calendar you do not control. Your draw goes in by a GC’s cutoff, typically the twentieth to the twenty-fifth, covering work through month end. Then it waits for certification. Every GC’s calendar is different, and a sub running fifteen jobs is running fifteen calendars.

Your costs front-run your billing by weeks. Field payroll is weekly. Material gets bought ahead of installation. On a job where you install in the first week of the month, you have carried that cost for sixty to ninety days before the money lands.

A slice of every dollar does not arrive at all for a long time. Retainage is withheld on each draw and released on triggers you mostly do not control. On a long job you are financing the GC’s comfort at an amount that can exceed your net profit on the work.

Put together: you can be profitable on every job, growing, and still miss payroll. Growth makes it worse, because every new job consumes cash before it produces any. That is the specific failure a forecast exists to see coming.

The four inputs, and nothing else

You do not need a model. You need four facts per job, kept current.

One: contract value and billed to date. Contract value including approved change orders — an approved CO that never made it onto the schedule of values is revenue you are not going to bill. Billed to date tells you the remaining contract to bill, which is the ceiling on everything that follows.

Two: how much of the remaining work lands in each of the next three months. This is your only genuine estimate, and it comes from the schedule and the superintendent, not from dividing the balance by the months left. A job with three weeks of rough-in in November and nothing in December bills that way. Whether your lines bill percent complete or by unit changes how predictable this number is: unit-price lines move with installed quantities you can count, percent-complete lines move with judgment that a GC can disagree with.

Three: this GC’s cutoff and this GC’s actual lag. Two separate dates. The cutoff is published in the subcontract. The lag — submission to money in the bank — is not published anywhere, and it is the number that makes the forecast real. More on it below, because it is the one most subs skip.

Four: retainage held, per job. Not a percentage — the dollars, including on jobs that finished. This is a separate asset with its own timing and it does not belong in the next thirteen weeks.

Four facts per job. If you have those, the arithmetic is addition.

Build it in weeks, thirteen of them

Monthly buckets hide the problem. Payroll does not care that the month nets positive; it cares about Friday. A sub that forecasts monthly can show a comfortable quarter and still have three weeks in it where the account goes through the floor.

So: thirteen weekly columns, rolled forward every Monday. Thirteen weeks is the useful horizon because it is long enough to see a trough coming while you can still do something about it — delay a material buy, pull a draw forward, call the bank before you need the money rather than after.

On the money-in side, each expected draw lands in a single week: the week of the cutoff plus that GC’s actual lag, rounded to the week. One job with one GC produces one row.

On the money-out side, put in what actually leaves and when it leaves. Field payroll weekly, with burden — and burden is most of why payroll costs more than people expect, which is the subject of the labor burden rate. Office payroll, usually semi-monthly. Supplier invoices on their terms, which is where your open purchase orders matter, because a committed PO is cash that is already spoken for before any invoice exists. Then rent, insurance, equipment notes, taxes.

The mistake that makes forecasts useless: billed is not collectible

Here is the one that sinks most of them. A sub bills a draw, puts the full amount in the forecast for sixty days out, and moves on. Then the GC certifies ninety percent of it, and the forecast was wrong by ten percent of your largest inflow — in the week you were counting on it.

Keep three numbers, never one:

• Drawn — what you submitted. A request.
• Certified — what the GC agreed to. The first number you can responsibly forecast on.
• Received — what arrived.

The rule is simple. Before certification, a draw is an estimate in your forecast. After certification, it is close to a fact. If drawn and certified diverge regularly on a given GC, that gap is itself a forecastable pattern, and you should be applying it rather than hoping this month is different. And if a draw is certified and the money is still not coming, that is a specific and diagnosable problem, not a forecasting one.

Retainage gets its own line, with its own date

Do not spread retainage across your thirteen weeks and do not leave it in the general receivable pile. It behaves differently from everything else: it is withheld automatically, it is released on events rather than on terms, and the trigger is usually somebody else’s milestone. What actually triggers release runs through the gates in order.

Keep one schedule of retainage held by job and by GC, including every closed-out job, with your honest best estimate of the quarter it arrives. Then assume nothing in it lands inside the next thirteen weeks unless you have a specific reason.

Two things usually fall out of that schedule at once: your total retainage held is a large number next to your line of credit, and some of it is on jobs that finished long ago and that nobody is chasing, because the person who would chase it is busy with live work.

Measure each GC’s real lag. This is the whole thing.

If you do one thing off this page, do this. For every GC you have billed in the last two years, compute the average days from draw submission to cash received, and the spread around it.

It takes an afternoon from records you already have, and it changes three decisions. Bid markup, because a GC at ninety days is consuming your working capital and should pay for it. Which work you chase, because capacity spent on a slow payer is capacity you do not have for a fast one. And your forecast, which stops being a guess the moment the lag is measured rather than assumed.

The spread matters as much as the average. A GC who always pays in fifty-two days is easier to plan around than one averaging forty-five across a range of twenty to a hundred. Predictable beats fast.

What to actually do with it

A forecast is only worth building if a bad one changes a decision. The ones it should change:

• Time the material buy. If week seven is the trough, a large order lands in week nine instead.
• Negotiate terms at buyout, on the forecast. Pay-when-paid, retainage percentage and a retainage reduction at fifty percent completion are all buyout conversations, and they are easier with a number in front of you.
• Call the bank early. A line of credit is cheap and available when you ask twelve weeks out and expensive or unavailable when you ask on the Wednesday.
• Decline work you cannot finance. The job that sinks a growing sub is usually the one it won. Knowing you cannot fund the first ninety days of a large award is a legitimate, and rare, reason to pass.

How SubMark holds the inputs

Be clear on what this is: SubMark does not have a cash flow forecasting module, and there is no projection screen in it. Anyone who tells you their construction software forecasts your cash is describing something that still needs your schedule and your judgment.

What SubMark does is hold the four inputs in one place, current, instead of in four workbooks and one person’s memory. SubMark keeps:

• Contract value, billed to date and total paid on every job, with approved change orders added to the schedule of values automatically — so remaining-to-bill is a current figure rather than a reconciliation.
• Drawn, certified and received as three separate states on every draw, which is the distinction the forecast depends on.
• Retainage in dollars per draw, totaled across every job, including finished ones, with the final retainage billing flagged as a release rather than new revenue.
• A draw calendar with per-job cutoffs, submission methods and reminders across GCs, so the submission dates your forecast is built on are real dates.
• Job costing with budget against actual by cost type, where labor lands at wage cost (not burdened) as the daily labor logs are entered, and material costs pull in from QuickBooks bills and purchases. Not a live feed, and days rather than a monthly close.
• Open purchase orders shown as committed cost on the job, separately from invoiced cost — cash spoken for before the invoice exists.

The honest limits worth naming: job costing is per job and grouped by cost type, and labor by phase needs phase tracking switched on. On the QuickBooks side, bills and purchases pull in on the hour if you turn it on, while draws and POs are pushed to QuickBooks with a click, one at a time — nothing is pushed on its own. And T&M tag hours are billing, not cost, so they do not appear in your job-cost actuals.

The thirteen weeks are still yours to build. The difference is whether you spend Monday morning assembling the numbers or reading them.

Dedicated pages: draw tracking, retainage tracking and job costing. Field crews and the subs you invite are free and unlimited, which matters when the people producing your cost data are foremen — see how pricing works.

Profit is an opinion until the job closes. Cash is a fact every Friday. Forecast the one that can put you out of business.