Capex Discipline

In many small businesses, the line between an expense and a capital purchase moves with the year. In a good year a large repair is expensed, because the business can absorb the hit to profit and the tax deduction is welcome. In a lean year the same kind of repair is capitalized and depreciated, because profit needs protecting. When the business is being prepared for sale, almost every purchase gets described as growth, since growth spending is not deducted from the earnings a buyer pays for. None of this is dishonest, but it leaves a record of capital spending nobody can read. A buyer who cannot read it rebuilds it from the ledger and deducts whichever figure is larger.

Capex Discipline fixes the line. It decides what counts as capital spending, splits three years of it into the spending that keeps the business running and the spending that grows it, and reconciles the result to depreciation, so that the deduction a buyer takes is the one the records support.

A policy that does not bend

Everything starts from a written capitalization policy, approved by the outside accountant, applied consistently, with the last three years booked under it or restated to it.

The policy sets a dollar floor, below which everything is expensed, and useful lives by asset class, so that depreciation follows the asset and not the business's need for profit. It also sets a test that separates a repair from an improvement, illustrated with examples from the business's own purchases so that the people applying it can see where the line falls in their own work.

The restatement is the uncomfortable part. Across the three years, every item above the floor that was expensed and every item below it that was capitalized is reviewed, and the reclassifications are scheduled with their effect on EBITDA. The effect often runs against the owner, because repairs capitalized to protect a lean year come back into the expenses. The owner pays that cost for a reason. Once the policy holds, capital spending and EBITDA can no longer be traded against each other, and a buyer has no grounds to doubt either one.

Every item classified

The second piece of work is a schedule of every capital item above the floor in the last three years, each with its invoice, the asset it relates to, and a classification of maintenance or growth with a one-line reason.

The items come from more places than the fixed-asset account. The repairs account and the large lines of the expense accounts get searched too, because something capital in nature that was expensed is still part of the record a buyer will rebuild. Each item is matched to its invoice and classified against a written test. Maintenance replaces or restores capacity the business already had. Growth adds capacity, a new line or a new location. An item that does both, such as a faster machine replacing a worn one, is split by its documented cost.

The schedule stays current only if each new item is classified when it is bought. The purchase approval carries the classification and the asset it relates to, so the record is kept as the spending happens. The alternative is someone else rebuilding it years later from invoices whose purpose nobody remembers.

Maintenance reconciled to depreciation

The third test is whether average yearly maintenance spending over the three years comes close to annual depreciation. Where it does not, each asset class behind the gap needs an explanation: either a catch-up of maintenance deferred in the past, or assets that have outlived the lives the books gave them.

Depreciation matters here because it is what a buyer falls back on when the maintenance figure cannot be trusted. Over a long enough period, a business that keeps its assets in working order spends roughly what they depreciate, so a large gap in either direction raises a question. Maintenance well below depreciation suggests deferral, a saving that flatters the earnings now and gets paid for later. Maintenance well above it suggests assets kept running long past their book lives, or a catch-up after years of putting work off. Both can be legitimate. The explanations go in a table by asset class, with the gap in dollars and the reason beside it, and the outside accountant traces a sample of them to invoices and to the asset register.

A buyer who accepts the documented maintenance figure deducts it. A buyer who does not will choose a figure of their own, and their choice will not favour the seller. That is the discount this work removes: the one a buyer applies to the size of the deduction.

The other half of the split

Most of the attention in this driver goes to the maintenance figure, because that is what a buyer deducts. The growth figure matters as much for a different reason: it is where the business's reinvestment gets measured.

In the framework, growth is worth something only where the business earns more than its cost of capital on what it reinvests. Showing that takes two numbers, how much the business reinvested and what the reinvestment earned. The first comes largely from this schedule. Every purchase classified as growth, tied to its asset, is a dollar of reinvestment that can later be traced to the revenue and profit it produced. The register of growth projects and their returns, built later on this bridge under Reinvestment Return, starts from these classifications.

So the owner has a reason to classify honestly that has nothing to do with satisfying a buyer. A maintenance purchase called growth flatters the maintenance figure today. Later it shows up in the reinvestment record as growth spending that produced no growth, and it drags down the measured return on everything the business reinvests. A growth figure padded with maintenance makes the business look worse at the one thing the next anchor will pay for. The split has to be right in both directions to be any use.

What it does to value

Most of what this driver does is a price effect, and the framework keeps it separate. Reclassifying spending that has not changed, for example moving a purchase from maintenance to growth because the documents now support it, changes the deduction a buyer takes and leaves the cash the business pays exactly where it was. The framework records that in its price column and does not count it as value.

What does count is spending avoided. A policy that stops maintenance items being capitalized twice, or bought early because there was cash to spare, changes what the business spends every year, and that recurring saving, discounted at the cost of capital, is the driver's value. Because its main effect falls on what a buyer believes and only modestly on what the business pays, the driver ranks low on this bridge. It still has to be done: without it, the next driver's replacement plan and the next anchor's valuation have no reliable record to start from.

What has to be in place first

The capital spending history is a reading of three years of reviewed accrual statements from Books, and the split cannot be made on cash-basis books. The outside accountant has to be engaged to approve the policy and test the schedule, with the restatement work added to the review engagement and budgeted. And there has to be a purchase approval step for anything above the floor, a form or a field in the purchasing system, because the classification has to be captured by someone at the time.

Knowing that it works

The driver is tracked by the share of capital items above the floor that are classified with an invoice; by maintenance spending against depreciation, by year and by asset class, where the unexplained gap in dollars should be zero; and by the share of each month's capital purchases classified at approval, which should be all of them.

The written test holds if the owner, given five items from the schedule with their classifications hidden, classifies them against it and reaches the same answers. A disagreement means either the test or the classification is wrong, and whichever it is gets fixed before an inspector arrives. The owner's understanding shows in saying, in two sentences and from the table, why maintenance spending differs from depreciation.

Why it slips, and what holds it

What this driver guards against is the policy bending to flatter a year. Two safeguards hold it. The policy sits in the review engagement and changes only by written amendment, dated, approved by the accountant and applied backward, so a change cannot lift one year's earnings. And classification is a required field on every purchase approval, filled in by the person approving the spend, which removes the temptation to reclassify at the year end once the earnings are known.