Shares outstanding or diluted weighted average: which one goes in a DCF

Two tags answer two different questions. dei:EntityCommonStockSharesOutstanding is a point-in-time count, taken off the filing's cover page, dated a few weeks after the fiscal year ends. us-gaap:WeightedAverageNumberOfDilutedSharesOutstanding is an average across that whole year, and it already has the dilutive effect of options, RSUs and convertibles baked in. A DCF wants one of them on purpose, and using whichever one happened to load first is how a correct enterprise value turns into a wrong price per share.

This has already cost us a live bug. On 29 August 2026, a company in our pipeline that does not tag the cover-page count hit a lookup that silently kept the previous company's figure instead of flagging the gap — 24.3bn shares, Nvidia's count, carried into Shopify's valuation. It priced Shopify at €5.10 a share instead of €94.97. The output looked like a number, not like an error, which is exactly the failure mode the first guide on this site describes.

The two tags, and the question each one answers

dei:EntityCommonStockSharesOutstanding — how many shares exist right now, basic count, no dilution added, dated to the cover page of the filing itself.

us-gaap:WeightedAverageNumberOfDilutedSharesOutstanding — the average share count the income statement actually used, across the fiscal year, with dilution already included.

For a DCF built on a fiscal year's free cash flow, the diluted weighted average is usually the right denominator, because the numerator and denominator then cover the same period: the same convention a company uses to get its own diluted EPS. The cover-page count is the right one when you need today's share count specifically — after a buyback, a merger, or a capital raise that happened since the fiscal year closed, when the trailing average is simply out of date. Neither tag is wrong. Picking one without checking which question it answers is.

Filter both to annual forms — 10-K for domestic filers, 20-F or 40-F for foreign ones — and read the end date on every fact you keep. The two tags were never designed to be compared to each other automatically.

Trap one: two large filers never tag the cover-page count at all

Twenty companies, the same ones surveyed for the free cash flow guide. Two of them — Alphabet and Meta — have no dei:EntityCommonStockSharesOutstanding fact anywhere in their SEC filings. Not an empty array, not a stale one: the tag is entirely absent from their full company facts file, checked directly rather than through the summary endpoint that can miss dimensioned data.

This is the trap that produced the Shopify bug above. Shopify does not tag this field either. When a field a filing simply cannot supply is treated the same as a field the code forgot to re-fetch, the previous answer in the batch quietly becomes this one's answer. The fix is the same one the FCF guide landed on: a missing field is blanked and flagged, never carried over.

What 20 large filers actually show, read on 19 October 2026

Company	Cover count	Diluted average	Date gap	Difference
Apple	14.776bn	15.005bn	20d	−1.52%
Microsoft	7.426bn	7.453bn	23d	−0.37%
Nvidia	24.300bn	24.514bn	26d	−0.87%
Amazon	10.735bn	10.827bn	28d	−0.85%
Alphabet	never tagged	12.230bn	—	—
Meta	never tagged	2.574bn	—	—
Tesla	3.752bn	3.528bn	23d	+6.36%
JPMorgan	2.697bn	2.781bn	31d	−3.04%
Exxon	4.167bn	4.419bn	4,414d	stale match
Johnson & Johnson	2.410bn	2.429bn	40d	−0.80%
Walmart	7.972bn	8.022bn	39d	−0.62%
Procter & Gamble	2.324bn	2.423bn	31d	−4.05%
Coca-Cola	empty answer	empty answer	—	—
Pfizer	5.686bn	5.713bn	50d	−0.47%
Intel	4.995bn	4.530bn	20d	+10.26%
Chevron	1.995bn	1.856bn	37d	+7.51%
AT&T	7.001bn	7.179bn	28d	−2.49%
Boeing	0.785bn	0.762bn	23d	+3.02%
General Electric	1.049bn	1.068bn	15d	−1.80%
Merck	2.472bn	2.507bn	31d	−1.38%

Coca-Cola repeats the same empty-array pattern the FCF guide found for its cash flow tags: a 200 response, a valid units object, nothing inside it. Checked four times, empty every time, on both tags here.

Trap two: the gap is not noise, it is buybacks and issuance

Fourteen of the twenty companies above show the cover-page count below the diluted average — the ordinary signature of a company buying back stock faster than it issues it, so the most recent count is smaller than the year's average. Apple, Procter & Gamble and JPMorgan all sit in the one to four percent range from buybacks alone.

Four companies run the other way, and each has a reason that shows up in the filings, not in the arithmetic. Chevron is +7.51%: it closed the all-stock acquisition of Hess during the period the cover count reflects. Intel is +10.26%: external capital raised into the company increased the share count after the fiscal year the diluted average covers. Tesla and Boeing sit between three and six percent for the same reason, new shares issued after the period the average was computed over.

The rule that fixes it. A gap above a few percent is not an error to average away. It means a buyback, a merger or a raise moved the count after the diluted average's period ended, and the fix is to know which direction it moved and say so next to the figure — not to silently prefer whichever tag returned a number first.

Trap three: when the two "last facts" are not even the same year

Exxon's row above reads 4,414 days, just over twelve years, because that is what naively taking the newest fact for each tag actually does when one tag was abandoned. Exxon's cover-page count is filed every year through FY2025. Its diluted weighted average under us-gaap:WeightedAverageNumberOfDilutedSharesOutstanding has exactly three annual facts, and the newest is FY2013.

Exxon did not stop having a diluted share count. It stopped tagging one under this exact element. The closest replacement in its own filings is WeightedAverageNumberOfSharesOutstandingBasic, which has 39 annual facts running through FY2025 — but that is the basic count, not the diluted one, and treating it as equivalent repeats the same mistake in a different place.

Comparing a FY2025 cover count against a FY2013 diluted average produces a confident-looking percentage from two numbers twelve years apart. The script behind this page checks the gap between the two dates before computing anything, and refuses to publish a figure once that gap passes about fourteen months — the same discipline as leaving a field blank rather than guessing.

What a correct lookup does, in order

Ask for both tags. Never assume the cover-page count exists — two of twenty companies here don't have it.
Keep only annual forms, and read the end date on every fact, not just the value.
Before comparing the two tags, check that their newest facts cover the same, or nearly the same, fiscal year. A multi-year gap means one tag was abandoned, not that the share count jumped.
Default to the diluted weighted average for a DCF built on that year's free cash flow, so the numerator and denominator share a period.
When a known buyback, merger or capital raise happened after that period closed, prefer the current cover-page count instead, and say which one you used and why.
When a tag has no facts at all, leave the field blank and flag it. Never carry over the value from the company processed just before it — that is the exact bug this page opened with.

Check this yourself

Every figure above comes from one script hitting SEC EDGAR, and both the script and its output are published rather than described:

shares-outstanding-tag-survey.json, the full result for all twenty companies, including the ones with no facts at all.
shares-outstanding-tag-survey.py, the script that produced it. Run it and the numbers regenerate. If it disagrees with this page, the script is right and the page is stale.

Read from SEC EDGAR on 19 October 2026 · figures are annual filings only · corrections welcome at hello@fairvalueradar.com

Want the arithmetic without the plumbing? The calculator applies this rule to any SEC filer and shows you the tag and the date behind the share count it used.

Open the calculator