A developer's cash flow is lumpy on purpose — the forecast has to match it

A rolling forecast for a business with no steady month

A development budget can look healthy on an annual view and still run out of cash in a specific month if a construction draw and a development-charge payment land in the same window a pre-sale deposit was supposed to cover. A rolling cash flow forecast, refreshed against actual draw requests and actual sales or lease-up velocity, is the tool that surfaces a gap while there is still time to arrange a bridge or delay a discretionary cost — a 13-week cash flow forecast is the short-horizon version of the same discipline, stretched here across a construction timeline instead of a normal operating cycle.

Finding a shortfall in month two of an eighteen-month build is a planning exercise, solved with a phased land purchase or an earlier pre-sale push. Finding the same shortfall in month fourteen, with trades already on site and a closing date already marketed, is a crisis. The forecast's entire value is moving that discovery from the second scenario to the first, which only happens if it is rebuilt on real numbers every few weeks rather than treated as a document produced once at financing and filed away.

Two audiences, two numbers, one project

Your construction lender wants financial statements that show the project earning as it is built — percentage-of-completion accounting recognizes revenue and profit along the way, which is what a covenant test usually expects to see. Your tax return generally recognizes that same profit when a sale closes instead, following the completed-contract-style timing detailed on our developer tax services page. Neither number is wrong; they answer different questions for different readers, and our job is presenting both without letting either audience mistake one for the other, or letting an owner mistake the lender's number for spendable cash. A project that looks profitable on the covenant statement can still be cash-negative for months at a stretch, particularly right after a phase closes and before the next phase's deposits start arriving, and it is the cash position, not the accounting profit, that determines whether the next draw request or the next payroll clears.

Margin belongs to the phase, not the portfolio

A blended company-wide margin can hide one badly priced phase inside two strong ones, which is exactly backward for a business where the next land purchase decision depends on knowing which phase actually performed. We turn the job-costing detail from your books into a phase-by-phase margin report — land cost, capitalized soft costs, direct construction, allocated staff time — so a pattern shows up while there is still a next phase to price differently, not after the portfolio-wide number has already absorbed the miss. Comparing a completed phase against its original pro forma, item by item, also tells you whether an escalation clause or a contingency line actually did its job — information worth having before the next contract is negotiated on the same assumptions.

The capital stack decides who gets paid, and in what order

A senior construction lender, a mezzanine or preferred-equity provider, JV partner capital, and sometimes a US lender's participating debt can all sit in one project's capital stack, each with its own priority, its own fees, and its own claim on distributions, sometimes negotiated deal by deal rather than following one house template. We model the waterfall before the first draw is requested — who is repaid first, what triggers a capital call to JV partners when costs run over, and how a delay or a tariff-driven cost increase changes who is exposed — and rebuild that model every time a real assumption changes, rather than treating it as a one-time exercise from the financing memo. A capital call triggered by a cost overrun is also a moment where JV agreements are read closely rather than assumed: some require calls to be proportional and mandatory, others allow a partner to dilute rather than fund, and knowing which regime governs before the call is issued avoids a dispute layered on top of the overrun itself.