That is not a buyback, it is a payroll run

That a large repurchase programme is capital being returned, a signal of confidence and a support under the stock. Gross buyback dollars get quoted in headlines, counted in shareholder yield screens and modelled as a reduction in share count.

Buybacks are not the problem. Doing the arithmetic on the gross number is.

At a meaningful set of companies, particularly in software and technology, the repurchase is mostly mopping up shares issued to employees. Stock compensation creates new shares, the buyback retires them, and the net count barely moves. Cash left the business, owners gained no additional claim on it, and the compensation expense never appeared where an investor would look for it.

The honest reading is that a portion of the buyback is an operating cost settled in the capital account. Sizing it means comparing dollars spent on repurchases against the change in diluted shares outstanding. Where the two disagree sharply, the difference is the real cost of paying people in equity.

What makes this a position rather than a complaint is valuation. Free cash flow gets quoted before this cost and after it in ways that shift a multiple materially, and screens ranking on shareholder yield pick these names up as generous when they are not.

Not as a sudden repricing. As a slow correction, arriving when growth slows enough that the equity can no longer carry the compensation.

The moment to watch is when a company has to choose between letting dilution run and spending cash it now needs elsewhere. Both options are visible in the numbers. Neither is in the narrative.

One ratio does the work. Take the dollars spent on repurchases over a period of several years, and divide by the fall in the diluted share count over the same period, valued at the average price paid. Call the result the net retirement rate. At a company genuinely returning capital it sits near one. At a company running a payroll through the capital account it sits far below one, and in the worst cases the share count is flat or higher after billions have been spent.

Run it over three to five years rather than one. A single year is noise: a large acquisition, a convertible settling, a founder selling down. The pattern is what matters, and the pattern is remarkably stable, because it is a function of how a company chooses to pay people rather than of anything cyclical.

Then restate the cash flow. Subtract stock compensation from free cash flow and recompute the multiple. In enterprise software this routinely moves a valuation by a third or more, and it moves it most at exactly the names that screen cheapest on the unadjusted number.

Where the screen points is not a secret, it is just unfashionable to say. Large capitalisation enterprise software is the centre of it: companies running stock compensation at a high teens to low thirties share of revenue, announcing multi billion dollar repurchase authorisations, and reporting diluted share counts that move very little. Salesforce, Workday and Snowflake are three of the most liquid places to see the mechanic working, which is why they are the ones worth pulling first. Not because the businesses are bad, but because the published shareholder yield and the real one are different numbers.

The long side is the better half of this theme and almost nobody writes it up, because it is boring.

Own the companies where the count actually falls. Apple has retired an enormous share of itself over a decade and the diluted count proves it. AutoZone and O'Reilly have done the same thing in a far duller industry, to the point that the share count reduction is most of the per share compounding. These are the names where the headline buyback and the real buyback are the same number, and where the shareholder yield on a screen is the one you actually receive.

The pair writes itself. Long a real retirer against short a gross spender with a flat count, matched on sector where you can, is a position in the accounting rather than in the economy. It does not need a recession, a rate move or a growth scare. It needs the market to eventually quote buyback yield net of issuance, which it does, slowly, every cycle.

The catalyst, where there is one, is a growth slowdown. Equity compensation is affordable while the equity is appreciating. When it stops, the company has to choose: let the count run and dilute, or spend real cash it now needs elsewhere. Both choices are visible a quarter or two ahead in the authorisation size and the issuance line, and neither is in the story the company tells.

There is a third expression for anyone who wants it. Several of these companies are issuing stock into their own buyback at valuations they will not see again. That is a quiet transfer from owners to employees, and the size of it is computable from two lines in the cash flow statement. It is the single best argument for reading the financing section before the headline.