Forecast Models

Two playbooks: the integrated three-statement model and the bottom-up operating model. They share a projection discipline, given once below, and diverge sharply after it — the three-statement model is the corporate-finance build whose test is internal consistency, the operating model is the growth-equity build whose test is whether the unit economics work. The method in SKILL.md applies throughout.

---

Projection discipline

Tie period zero to actuals first. Before projecting anything, reconcile the starting period to reported figures and show the reconciliation. A forecast built off a base that does not tie to the accounts is wrong from the first column, and the error compounds silently across every year.

Drivers, not growth rates. Revenue is built from operating quantities — units and price, customers and ARPU, headcount and utilisation, volume and take rate. "Revenue grows 15%" is an output, never an input. Where a genuine growth-rate assumption is unavoidable, mark it and say what would have driven it.

State the granularity and why. Monthly where cash timing decides the answer, quarterly where seasonality does, annual otherwise. Do not project monthly for five years; the false precision costs credibility and hides nothing useful.

Name any structural break in the history. If the historical period contains an acquisition, a business-model change, a move between premises, a pricing reset or a pandemic distortion, whole-period averages are not a valid base. Say where the break falls and project off the post-break period, stating the shorter sample.

Close on sensitivity. Both playbooks end by naming the three assumptions that most move the output and the level at which each changes the conclusion.

---

Three-Statement Model

Income statement, balance sheet and cash flow, fully linked. Internal consistency is the deliverable — a three-statement model that does not tie is not a model.

When: "three-statement", "integrated model", "build me a forecast", "project the financials", "what does the balance sheet look like in year three", "will it need financing".

Phases:

1. Architecture. State the revenue recognition basis — volume × price, subscription, project completion. List every linking formula you will use: net income to retained earnings, D&A to the cash flow and the PP&E roll, debt schedule interest to the income statement, closing cash to the balance sheet. Then mark each line as assumed or derived. Derived lines must never be independently assumed later; that is how models stop tying.
2. Income statement. Revenue with its build; less COGS to gross profit and margin; less SG&A and R&D to EBITDA and margin; less D&A to EBIT and margin; less interest from the debt schedule to EBT; less tax at the effective rate, not the statutory rate, to net income and EPS. Where the effective and statutory rates differ materially, say why.
3. Balance sheet. Assets: cash, receivables at DSO × revenue / 365, inventory at DIO, PP&E as prior plus capex less D&A, intangibles. Liabilities: payables at DPO × COGS / 365, accruals, debt from the schedule. Equity as prior plus net income less dividends.
4. Cash flow. Operating: net income plus D&A plus or minus working capital movements. Investing: capex and any M&A. Financing: debt draws and repayments, dividends, equity issuance. Net movement to closing cash.
5. Supporting schedules. Working capital with each day-count assumption stated and compared to the historical level; debt schedule with opening, draws, repayments, interest and closing; PP&E roll; retained earnings roll.
6. Integrity table. Report the figure, every year, never a tick:

Check	Y1	Y2	Y3	Y4	Y5
Assets − liabilities − equity
Closing cash, balance sheet − cash flow
Interest expense − debt schedule interest
D&A, income statement − PP&E roll
Tax expense − (EBT × effective rate)

Cash is a plug only while it is positive. If the plug goes negative the company has a financing need, not negative cash. Route the shortfall to a revolver draw, carry the interest, and say in one line what facility the model assumes exists. A model showing negative cash has substituted an impossibility for a conclusion.

Circularity. Interest depends on debt, debt depends on cash, cash depends on interest. Either compute interest on the opening balance and say so, or iterate to convergence and say so. State which, rather than leaving the loop unremarked.

Required tables: the three statements; each supporting schedule; the integrity table.

Challenge test: the three assumptions that most move closing cash or the financing requirement, with the level at which the company needs new capital.

Inputs to gather: company name and business description; current-year revenue and growth; gross margin, historical if available; EBITDA margin, current and target; capex as a percentage of revenue or absolute; existing debt balance and rate; working capital dynamics and payment behaviour; dividend or capital return policy; the drivers that actually move revenue.

---

Operating Model and Unit Economics

A bottom-up build for a growth business, where the question is whether the economics work at scale rather than what the business is worth. Top-down TAM × market share is not an acceptable revenue build here.

When: "unit economics", "LTV/CAC", "cohorts", "burn and runway", "does this scale", "operating model", "when do we break even", "how much do we need to raise".

Phases:

1. Revenue architecture. Select the build that fits and say why: SaaS as cohort ARR from logos × ACV with gross and net retention; marketplace as GMV × take rate with buyer and seller cohorts; e-commerce as orders × AOV with repeat rate; services as headcount × utilisation × bill rate; subscription as subscribers × ARPU with churn.
2. Unit economics. Current against target, with the definition stated for each so the numbers can be checked:

Metric	Definition	Current	Target
CAC	Fully loaded S&M / new customers acquired
Gross margin per customer	Revenue − direct COGS, per customer
LTV	Gross margin per customer ÷ churn rate
LTV / CAC
CAC payback	CAC ÷ monthly gross margin per customer
Net revenue retention	(Prior cohort revenue + expansion − churn) ÷ prior
Gross churn	ARR lost to cancellation ÷ opening ARR

Three cautions that decide whether these numbers mean anything. LTV as gross margin over churn is a perpetuity — it assumes the customer relationship runs forever at today's churn, which overstates high-churn businesses badly; cap it at a stated horizon, three to five years, and show both. Separate blended from paid CAC; blended CAC falls whenever organic acquisition rises and flatters a deteriorating paid channel. Measure NRR on a fixed cohort, not on total revenue, or new customer growth reads as expansion.

The conventional thresholds — LTV/CAC above 3x, payback under 18 months — are benchmarks for comparison, not targets the company must be shown to hit. Report the actual figures and whether they are improving or deteriorating with scale. The trend matters more than the level. 3. Cohort analysis. For each acquisition cohort: revenue in years one, two and three in absolute terms and as a percentage of initial; cumulative gross margin by year; and the year cumulative margin exceeds CAC. Whether later cohorts pay back faster than earlier ones is the single most informative line in this analysis. 4. P&L projection. Monthly for year one, quarterly for years two and three: new customers, total customers, revenue, gross profit, S&M, R&D, G&A, EBITDA, burn, closing cash. Carry a hiring plan by function showing how headcount drives each cost line, rather than growing opex as a percentage of revenue. 5. Cash and runway. Current monthly burn; runway at current burn; runway at projected burn including planned growth spend; and the revenue level or date at which the business reaches breakeven under the base case. 6. Scenarios. Bear as churn up three points with growth down thirty percent, base, and bull as NRR up five points with growth up thirty percent. Report ARR at year three, gross margin, months to breakeven, and total cash required to reach it. The last figure is the one that determines the next raise. 7. Investment signal. The ARR level or unit-economic milestone at which this becomes attractive, and the single gating question that would resolve it.

Required tables: unit economics; cohort table; the P&L projection; scenario summary.

Challenge test: the churn rate at which LTV/CAC falls below the cost of acquiring the next customer, and how far that sits from current churn.

Inputs to gather: company name and business model; current ARR or run rate; customer count; CAC and payback if known, with whether CAC is blended or paid; gross and net churn; gross margin; monthly burn and cash on hand; the growth levers being funded; current headcount by function.