Share-based compensation is the cost most likely to be missing from a private investor's model, and it is missing for a good reason: no cash leaves the company, so every instinct you have about costs points the wrong way. The bill is real anyway. It is paid in ownership, it is paid by you, and it shows up as a share count that keeps climbing while your slice of the same business keeps shrinking.
The awkward part is that most models do not skip it out of ignorance. They handle it in a way that quietly cancels it out, which is worse, because the sheet looks complete.
Three ways a retail model makes stock comp vanish
The first is taking adjusted or non-GAAP earnings at face value. Companies report both figures and the adjusted one usually excludes share-based pay, on the reasoning that it is non-cash. Build a price to earnings comparison on the adjusted number and you are valuing a business whose compensation bill has been partially deleted. On a company where stock comp runs at five percent of revenue against a twenty percent operating margin, that exclusion is a quarter of operating profit.
The second is using cash flow based free cash flow without a matching correction. Operating cash flow adds stock comp back as a non-cash charge. That is correct accounting, since the cash never moved. But if your free cash flow is operating cash flow minus capex, and you then value the equity on today's share count, you have taken the benefit of paying people in stock and never recorded the cost of it. Either deduct the expense from the cash flow or model the share count rising. Doing neither counts one side of the trade.
The third is the most common and the hardest to see: modelling everything correctly at the company level, then dividing by the share count that is on the screen today. A five year model divided by a share count that only exists in year zero has an error compounding through it.
Two rows that look identical and the column that is not there
It helps to see what a screen can and cannot tell you here.
Worth stating plainly, because it changes what you can do with the screen: there is no enterprise value column on this view, and no share count column either. If you want two names lined up on identical enterprise values, which is the comparison you actually need when debt and cash differ, this screen will not build it for you. You build it: market cap plus debt minus cash, both names, from the last balance sheet. That is a ten minute job and it is yours, not the screen's.
What the two rows do illustrate is how easily two companies can look interchangeable on the summary columns. Identical sub-scores across fundamental, technical and sentiment, similar composite figures, similar verdicts. Two businesses can sit that close on a summary line and have completely different dilution profiles, and none of the columns above would separate them. I mention these tickers only because they are the rows visible in the capture, and the point is the method, not either company.
Counting shares forward instead of arguing about the expense
The treatment I would use, and the one that survives contact with a real filing, sidesteps the accounting argument entirely. Do not fight about whether the expense is real. Count the shares.
Pull diluted weighted average shares from the last three annual reports. Compute the annual growth rate. That single number already nets grants against buybacks, since it reflects what actually happened to the count rather than what management said about either activity. Project it forward across your model horizon, then divide your equity value by the year five share count instead of today's.
Say gross grants run at about three percent of shares a year and buybacks retire about one percent. Net dilution is two percent annually. Over five years that compounds to roughly ten and a half percent more shares. Your per-share value is not ten percent lower, it is about nine and a half percent lower, because you are dividing by a bigger number. That is the whole adjustment. No opinion about non-GAAP required.
What the dilution costs a 6,000 dollar position
Numbers in dollars, because percentages hide things. You put 6,000 dollars into a name at what you believe is a fifteen percent discount to fair value, so your model says the position is worth about 6,900. Apply a nine and a half percent haircut to per-share value for five years of net dilution and fair value falls to roughly 6,250. Your discount was not fifteen percent. It was about four percent, which is inside the error bars of every assumption in the sheet.
That is the actual damage: not that you lose money on the position, but that you thought you had a margin of safety and you were trading on rounding error. The names where this matters most are the ones where stock comp is largest relative to profit, which tends to mean younger, faster growing, more heavily engineering weighted businesses. Which are, unhelpfully, exactly the names where the growth story makes the dilution feel like a detail.
The buyback line that is really payroll
One more check, because buyback announcements are where this cost gets laundered. A company announcing a large repurchase reads as capital returned to shareholders. Sometimes it is. Sometimes it is the company buying back the shares it just issued to staff, at market price, with your cash.
The way to tell them apart takes thirty seconds and it is not the press release. Take diluted weighted average shares this year against last year. If the count fell meaningfully, the buyback returned capital. If the count is flat or up despite hundreds of millions spent on repurchases, that spending was compensation settled in cash, routed through the share register. It is a legitimate way to run a business. It is not a return to you, and it should not be sitting in your model as one.
The check to run this week
Pick your largest holding. Open the last three annual reports, find diluted weighted average shares in each, and write the three numbers on one line. If the count is flat or falling, stock comp is being fully absorbed and you can leave your model alone. If it is compounding at two or three percent a year, go back to your valuation and divide by a bigger share count.
Then do the same for the next name on your shortlist before you buy it, because this is the rare check that is cheaper to run before you own something. A screen will happily rank two companies as equivalent on the columns it shows. The share count trajectory is not one of those columns, and on a five year holding period it is frequently the difference between the two.