Budget stock compensation in shares, not dollars

Every quarter, software companies publish two versions of their profit. One follows US accounting rules and includes the cost of the shares and options they grant to employees. The other, usually labelled adjusted or non-GAAP, adds that cost back. Analysts mostly use the second. Many boards do too, because it is the number in the investor presentation and often the number executive bonuses are measured against.

I understand why the adjustment exists. Stock compensation is a non-cash expense, its accounting value depends on assumptions about volatility and forfeitures, and excluding it makes companies with different grant practices easier to compare. But I have come to believe the adjustment does harm inside the companies that rely on it, and that the harm is greatest in the years when a software company moves from growth to profit.

What the add-back hides

Stock compensation is not free because it avoids cash. It is paid in ownership. When an employee's restricted stock units vest, the company issues new shares, and every existing shareholder owns a slightly smaller piece of the business. A company that pays an engineer $200,000 in cash and $100,000 in shares has spent $300,000. The last third is paid by its owners rather than its bank account.

When I joined Brindlewood in 2021, stock compensation was 19 per cent of revenue. Our free cash flow that year was negative 6 per cent of revenue; with stock compensation counted, the real cost of running the business was closer to negative 25 per cent. The share count rose 4.6 per cent during the year, net of shares withheld to cover employees' tax. Nobody had concealed any of it. It was all in the filings. But the measures management discussed each week, adjusted operating margin and free cash flow, were both stated before the cost, so the cost was nobody's daily concern.

The same issue distorts the rule of 40, the familiar shorthand that a software company's revenue growth rate and profit margin should together exceed 40. Brindlewood's 2025 score on the usual basis is 36: 17 per cent growth plus a 19 per cent free-cash-flow margin. Deduct stock compensation and it falls to 26. I publish both, and I lead with the lower one, because it describes what a shareholder actually receives. In 2021, measured the same way, our score would have been 8.

Why dollar grants fail when the price falls

Most software companies set equity grants in dollars. An engineer at a given level receives, say, $60,000 of stock a year, converted into units at the share price on the grant date. It sounds fair and it is easy to explain. It also means that the number of shares a company issues rises as its share price falls, which is precisely when issuing them costs existing holders the most.

We met that problem head-on. Our shares fell from $88 in November 2021 to $31 by October 2022. Granting the planned dollar values in 2023 would have required almost three times as many shares as the year before, and gross grants would have reached 6.1 per cent of shares outstanding. The proposal on the table went further and would have topped up existing awards to restore their lost value. The motive was decent. Our engineers had been told to expect value they no longer had, and competitors were recruiting them.

We adopted a share budget instead. The Compensation Committee now approves, before any other part of the pay plan, the maximum number of shares the company may grant in the coming year: 3 per cent of shares outstanding in 2023, 2.5 per cent in 2024 and 2.25 per cent in 2025, with a ceiling of 1.5 per cent on net dilution. Within the budget, grants are set in share bands by level, so employees bear some of the price risk that every other owner carries. We moved part of junior engineers' pay into cash, which costs about $14 million a year and which I count as the price of the policy.

It was not painless. Engineering attrition rose from 11 per cent to 16 per cent in 2023 before falling back to 10 per cent in 2025, and some managers felt that finance had overruled them on their own people. Stock compensation is now 10 per cent of revenue, and net dilution was 1.1 per cent in 2025.

A buyback that only stands still

The second distortion appears once companies begin to generate cash. Many announce share repurchases and describe them as returning capital to shareholders. Often the repurchases do little more than absorb the shares issued to employees in the same period. A company that issues 3 per cent of its shares to staff and buys back 3 per cent has not returned capital to anyone. It has paid its employees in cash by an indirect route, and the buyback is part of the cost of compensation.

I would ask any board to split a buyback into two parts when it reports one. The first is the amount needed to offset dilution from awards, which belongs with compensation. The second is anything beyond that, which genuinely reduces the share count and should be judged against other uses of capital, acquisitions included. At Brindlewood our $120 million authorisation was explicitly sized to cover the first and do a modest amount of the second, and I report the two separately each quarter.

There is a serious counter-argument, and I have heard it from compensation consultants and directors I respect. Software companies compete for engineers with richer rivals, equity aligns employees with owners, and a share budget can push talented people towards companies willing to dilute more freely. All true. My answer is that a budget does not forbid generous grants; it forces the board to decide how generous, in the currency shareholders actually pay. A company can still choose 4 per cent a year. It should simply choose it in the open.

For directors on a Compensation Committee, three numbers do most of the work. Gross burn, the shares granted in a year as a share of those outstanding. Net dilution, after forfeitures and shares withheld for tax. And stock compensation per employee, which shows whether a rising expense comes from more people or from richer grants. A committee that approves those numbers before it approves the salary budget is governing the cost. One that reads them afterwards is merely recording it.