Forecast Model

When a buyer who models rather than multiplies looks at an owner-run business, someone on its team builds a forecast. The only question is whose it will be. If the seller has none, or has a spreadsheet of growth rates nobody can trace to anything, the buyer's analyst builds the model from the statements, at assumptions chosen to protect the buyer. The seller then negotiates against a number it has never seen and cannot test, built on assumptions it would not have chosen.

Forecast Model changes whose model it is. It turns free cash flow to the firm into a projection the seller can defend: an integrated forecast built from the same drivers the history shows, whose discounted value is the governing figure and is reconciled to the market multiple. Given a model that reproduces the past, reaches cash and has been independently reviewed, a buyer has little reason to build its own. It starts from the seller's and argues with specific inputs.

Three statements, linked and tested

The model runs five years, monthly for the first and annual after. Its profit and loss, balance sheet and cash flow statement are linked and balance, its first year equals the approved budget, and when it is run on the last two years' actual drivers it reproduces their actual results.

Linking matters because cash lives in the balance sheet. A forecast of the profit and loss alone can show rising profit while the business runs out of money, since it never asks what the growth does to receivables, stock and equipment. Once the three statements are linked, that cannot be hidden. Every sale the model forecasts creates a receivable, every receivable is collected on the observed days, and the cash at the bottom of the cash flow statement is what is left.

The inputs are drivers. Customers, volume, price, margin, headcount, capital spending and days of working capital sit on one sheet, and every output is computed from them. A buyer who disagrees with the forecast can then disagree with a particular driver, the retention rate or the margin, and see what the disagreement is worth.

The model proves itself on the past. Run on the actual drivers of the last two years, its output is reconciled to the statements, and the framework expects it to land very close to actual EBITDA. A model that cannot reproduce the past is not used for the future, because its mechanics are wrong somewhere and nobody knows where. A buyer's first question about any model is whether it works, and the back-test gives an answer the buyer can check for itself.

Revenue built from customers

The first year's revenue is built from existing customers at their observed retention rate, plus new business at the historical win rate, and the five-year growth assumption does not exceed the trailing three-year rate without a documented reason.

This is the part of most sellers' forecasts a buyer trusts least. A line labeled "revenue growth, 8 percent" says nothing about where the revenue will come from. A revenue build says exactly where. It starts from the customer schedule built in Diligence Readiness on the second bridge, applies the retention history to it, and adds new business at the rate the business has actually won it. The model's revenue then moves when the schedule does, so a customer lost or won changes the forecast the next time it is run.

Anything beyond the history is kept apart from it. A new product, a price change or a new market is a separate line with its own label and evidence, and is never folded into the base growth rate. A buyer can then accept the base, which rests on the record, and discount the new line, which rests on a plan. If the two are blended, it has to discount both. Kept apart, the base growth rests on evidence of customers staying and new ones arriving at the observed rate, and the buyer has much less reason to cut it.

The cash flows govern, and the multiple tests the price

This is the first bridge where the discounted cash flow governs. The model's value, discounted at the cost of capital built in the valuation file, is the governing figure. The value implied by the EBITDA multiple for comparable transactions is computed beside it as pricing, the difference is explained line by line, and the DCF is never adjusted to close it.

The two are ranked this way because they answer different questions. A multiple is a price: what buyers have recently paid for businesses that look something like this one, under conditions that may or may not hold today. A DCF is a value: what this business's own cash flows are worth at a rate that reflects their risk. So when they disagree, the disagreement tells you something. In the illustration, a business forecast to produce $630,000 of free cash flow next year, growing at 3 percent and discounted at 12 percent, is worth $7.0 million on its cash flows. Comparable sales at 4.5 times its $1.4 million of EBITDA suggest $6.3 million. The $700,000 between them needs explaining. Perhaps this business has something the comparable businesses lacked, or the market is paying less than the cash supports, or one of the model's inputs is wrong. Each explanation is written down as a sentence. An input changes only with a recorded reason, and never to make the gap disappear.

The discount rate comes from the valuation file. It is a weighted cost of capital whose cost of equity is that of the marginal investor at this anchor, with no size premium and no company-specific premium on top. That can look generous to a buyer. Owner dependence and customer concentration are real risks, but the framework puts them in the cash flows, where they can be named and measured, and keeps them out of a premium on the rate, where they cannot. A buyer arguing for a higher rate then has to say which risk the higher rate is meant to carry, and the seller can point to where that risk already sits.

The model also states its terminal value method, its terminal growth rate and the multiple it is checked against, so every assumption behind the governing figure is visible. With all of it visible, the cash-flow value is the number and the price is tested against it, and the buyer loses its reason to discount the size of the figure.

Sensitivities, and a reviewer who did not build it

Last, the model shows how its value moves with revenue growth, margin, capital spending and the discount rate, and an independent reviewer has checked its mechanics, with every finding closed.

The sensitivity sheet shows the value at a set step up and down on each main input, and the combination of inputs that would take the value below the multiple method. That combination tells a buyer, and the owner, how wrong the forecast would have to be before the price stopped being supported, which makes it the line on the sheet most worth reading.

The review covers mechanics only. A reviewer who did not build the model checks the formulas, the links and the balance, and every finding is logged and closed. A model built by one person carries that person's errors, and a buyer's analyst who finds one broken link will distrust every other cell. After a clean review the buyer can test the model instead of rebuilding it.

Together these are the checks a buyer's analyst runs before deciding whether to start from the seller's model or build its own, and they run in order. Testing the revenue build of a model that does not balance is a waste of time.

What it does to value

Forecast Model ranks ninth on the bridge. It changes neither the cash flow nor the rate. What it changes is the spread of outcomes a buyer has to allow for. Every buyer's valuation carries some probability that the model behind it is simply wrong. A model that reproduces two years of history and has passed an independent review narrows that probability to the size of its back-test variance, and the margin of safety a buyer holds back narrows with it. That narrowing is the value the framework counts.

The framework leaves out the multiple itself, any step-up in it, and any explained line in the gap between the DCF and the multiple. Those are price, and they are recorded as price.

What has to be in place first

The model's first year is the budget, so it needs the budget from Budgeting, or it starts from a guess. Its capital spending, working capital and tax lines are linked to the replacement plan from Asset Stewardship, the working-capital peg from Working Capital Management and the stated tax rate from Tax Position, so those schedules have to exist. And it needs two people, both budgeted: a modeller, who can be the finance seat if it has the skill or an outside modeller if not, and a separate reviewer.

Knowing that it works

The model's record is its back-test variance, which compares the model's output on actual drivers with the statements for each of the last two years, at EBITDA, and should stay within the framework's tight tolerance. The gap between the DCF value and the multiple method is tracked as a percentage, with the share of it explained, and open review findings should be down to none.

Changing an assumption, the retention rate or the margin, and explaining why the value moved as it did, without the modeller in the room, shows whether the owner understands the model. On the discount rate, the lead argues for a higher one as a buyer would, and the owner defends each component from its source. If the owner cannot do both, the model still belongs to whoever built it.

Why it slips, and what holds it

A model built for one year and left untouched while the business moves on starts to miss without explanation. Deferred maintenance shows up in a forecast as an improvement in cash that is really a cost postponed.

What keeps the model honest is that it is retested every year. It is rolled forward each year from the new budget and tested again on the year just closed, which renews its record every year and makes a failed back-test a visible finding. And the capital spending line stays linked to the replacement plan, so a deferral that lowers capital spending in the budget has to appear as a change to the plan, which the owner signs, and cannot slip into the forecast as better cash.