Automatic Data Processing ADP

Reference table

1. What the business is

ADP is a Delaware corporation headquartered in Roseland, New Jersey, listed on NASDAQ, with approximately 67,000 employees, led by chief executive Maria Black and — since July 2025 — chief financial officer Peter Hadley, who joined ADP in 2002 and served as Corporate Treasurer from October 2022 to June 2025. The fiscal year ends 30 June.

Two reportable segments, from the March 2026 10-Q ($M, nine months ended 2026-03-31):

Read that table and the shape of the business is clear. Employer Services — payroll, tax filing, HR software, time and attendance — is two-thirds of revenue at a 37.4% pre-tax margin and it is expanding. PEO Services, in which ADP becomes the co-employer of record for small and mid-sized clients' worksite employees, is a third of revenue at a 13.4% margin and it is contracting: margin down 110 basis points over nine months and 120 in the March quarter, with pre-tax earnings falling 2% on 6% revenue growth.

The PEO deterioration has a named cause and it is structural, not cyclical. From the 10-Q: nine-month operating expenses rose "due to an increase of $232.8 million of PEO Services zero-margin benefits pass-through costs to $3,427.2 million from $3,194.4 million", plus "$31.9 million due to an increase in costs related to workers' compensation coverage and state unemployment taxes for worksite employees." Zero-margin pass-throughs are revenue that carries no profit at all, so when they grow faster than the rest of the segment, reported revenue growth flatters and margin compresses mechanically. PEO's headline 6% growth therefore overstates its economic contribution, and this is the least-good line in the company.

The float, and why it is the most interesting line in the file

From the March 10-Q, verbatim: "Total revenues for the nine months ended March 31, 2026 include interest on funds held for clients of $999.4 million, as compared to $881.3 million... The increase in interest earned on funds held for clients resulted from an increase in our average client funds balances of 7.3% to $40.2 billion... coupled with an increase in our average interest rate earned to 3.3%... as compared to 3.1%." For the March quarter alone: $403.9 million on average balances of $48.3 billion at 3.3%.

Three things follow and only the first is usually said out loud.

It is nearly pure margin. ADP incurs essentially no incremental cost to earn it, so its contribution to pre-tax earnings is close to its full dollar value. $999.4 million against nine-month pre-tax earnings of $4,472.5 million is 22.3% of pre-tax profit from 6.1% of revenue.

It is a rate bet the income statement does not label as one. ADP invests the float in a laddered portfolio "out to five years (in the case of the extended portfolio) and out to ten years (in the case of the long portfolio)", which damps but does not remove the sensitivity. A 100-basis-point fall in the earned yield is roughly $400 million of annual revenue and close to $400 million of pre-tax earnings — about 7% of the fiscal-2026 pre-tax total of $5,730.3 million. The ladder means this arrives over years rather than quarters, which is a real mitigant and is why the score is 3 and not higher.

It grows with employment and wages. Balances rose 7.3% year on year, against pays per control of +1%, so most of the balance growth is wage inflation and new client wins rather than headcount. In a genuine employment downturn, the float shrinks at the same time the yield falls.

The acquisitions and the debt actions the payload does not carry

WorkForce Software, acquired in fiscal 2025 for $1,165.1 million net of cash acquired (vendor acquisitionsNet FY2025 = −$1,165.1M, an exact match to the filing's cash-flow line). The 10-Q names its "integration costs" inside research and development and its intangible amortisation inside depreciation and amortisation. No revenue contribution is disclosed anywhere in the archive.

Two financing actions in fiscal 2026 postdate every statement in the payload's income and cash-flow blocks and both are in the 8-Ks:

• 2026-05-04 (8-K filed 2026-05-07): an underwriting agreement to issue $1,000,000,000 of 5.000% senior notes due 2036. This is visible in the balance sheet — long-term debt rises from $3,974.7M at 2025-06-30 to $4,964.1M at 2026-06-30 — and it is the reason net debt is not lower.
• 2026-06-26 (8-K filed 2026-06-26): a $5.7 billion 364-Day Credit Agreement and a $3.5 billion Five-Year Credit Agreement, which "replaced the Company's prior $4.55 billion 364-day facility... and the Company's prior $3.5 billion five-year facility." The 364-day facility grew by $1.15 billion, 25%. These facilities exist to backstop the commercial paper programme that funds client-fund timing mismatches; the fiscal-2026 cash-flow statement shows $4,769.5 million of net commercial-paper repayment in the first nine months, which is why shortTermDebt in the payload falls from $4,769.5M to zero. Neither 8-K is an operating event; both are recorded here because the highest-yield check in this programme is the one that looks for corporate actions postdating the statements, and on ADP the answer is that there are two and neither changes the thesis.

2. The fiscal-2026 result, and how the payload reports it

Fiscal 2026 closed on 30 June and was reported on 29 July, six days before this dive. The fiscal-2026 10-K is NOT in the archive — the most recent annual report is fiscal 2025, filed 2025-08-06 — and the 8-K carrying the fourth-quarter results contains only the cover page and exhibit index, not the release itself. So fiscal-2026 quarterly and segment detail comes from the vendor payload and the earnings calendar; nine-month figures, segment margins and all operating commentary come from the 10-Q filed 2026-04-30. Where the two overlap they agree, and where they do not the filing is used and the difference is stated.

Five straight years of margin expansion, from 17.3% to 20.1%, on revenue growth that has settled into a 6-7% band. That is the whole story of this company in one table.

The quarterly earnings record is the tightest beat pattern in this programme and it deserves a comment:

Five consecutive beats, every one between 1.3% and 2.1%. That is not an operating surprise; it is a company that guides to a number it is confident of clearing by two points. It is a mark of quality and of predictability, and it is emphatically not a source of upside — a stock does not re-rate on a 1.9% beat.

The revenue result also cleared, and by a similarly small margin: consensus est.revenueAvg for FY2026 was $21,910.6M against an actual $21,947.5M, +0.17%. That agreement is itself a useful finding: the estimate basis and the reported basis match, so ADP does not carry the DE-class estimate-basis mismatch this programme has found elsewhere.

3. Cash generation, and a capex definition that changes mid-payload

This is the section where the vendor file needs the filing to be readable at all.

The arithmetic that proves it. For FY2023 the vendor's capex of $571.6M equals the filing's capex of $206.3M plus additions to intangibles of $365.3M — to the dollar. For FY2024, $208.4M + $355.0M = $563.4M — to the dollar. For FY2025 the vendor's $168.7M equals the filing's capital-expenditure line alone, with the $378.3M of intangibles additions no longer deducted. The definition changed between the FY2024 and FY2025 rows of the same field in the same file.

The consequence is that the payload's own free-cash-flow series overstates fiscal-2025 and fiscal-2026 relative to fiscal-2023 and fiscal-2024 by roughly $350-380 million a year, and any growth rate computed across the break is wrong. On the vendor's numbers FY2023→FY2026 free cash flow compounds at 13.0%; on a consistent basis it compounds at ≤10.4%. Neither number is disastrous and the direction of the business is the same either way — but a screen ranking on free-cash-flow growth would rank ADP roughly 2.6 percentage points too high, and this is the single most consequential defect in the file.

The consistent-basis fiscal-2026 figure is stated as an upper bound and labelled as one, because the fiscal-2026 10-K has not been filed and only the nine-month intangibles figure ($352.4M) is available. The full-year number will be higher, so consistent-basis FY2026 free cash flow is at most $4,892.2M, a 4.52% yield on the $108.14B market capitalisation, against the vendor's implied 4.85%.

Capital intensity is genuinely low and the company says so. From the March 10-Q: "Capital expenditures for the nine months ended March 31, 2026 were $124.6 million... We expect capital expenditures in fiscal 2026 to be between $160.0 million and $180.0 million, as compared to $176.8 million in fiscal 2025." On $21.9 billion of revenue that is 0.8%. The real investment is the ~$350-380 million a year of software capitalisation, which the 10-Q attributes to "investments in internally developed software primarily for our next-gen products, and purchased software."

Capital return, from the filings: nine-month fiscal-2026 repurchases of $1,463.2 million against $956.5 million a year earlier, and 5.7 million shares retired against 3.4 million; full-year repurchases of $2,083.3 million per the vendor cash-flow block. Nine-month dividends paid were $1,943.1 million. Trailing dividend per share is $6.64, a 2.45% yield, at a 59.5% payout ratio. Total shareholder yield is therefore approximately 1.9% buyback plus 2.45% dividend ≈ 4.3%.

4. Balance sheet — and the $46 billion that is in the wrong place

ADP's balance sheet is almost entirely client money, and the payload puts the asset and the matching liability in different places.

From the 10-Q at 2026-03-31 ($M):

And from the payload at fiscal-year end ($M):

Look at the fiscal-2026 column against fiscal-2025. Total current assets collapse from $43,252.2M to $8,603.3M while total non-current assets expand from $10,117.1M to $54,590.0M — a $34.6 billion migration with no corresponding event. Meanwhile total current liabilities stay at $50,024.9M, because the client funds obligations that match those assets remain classified as current. The filing is unambiguous: ADP reports funds held for clients as a CURRENT asset and client funds obligations as a CURRENT liability, and they match almost exactly ($46,413.1M against $46,774.5M at 2026-03-31).

The damage this does to the derived ratios is total. The payload reports currentRatioTTM of 0.172, quickRatioTTM of 0.172, workingCapitalTTM of −$41,421,600,000, netCurrentAssetValueTTM of −$48,558.8M, investedCapitalTTM of −$35,454,000,000 and workingCapitalTurnoverRatioTTM of −1.12. Taken at face value, that describes a company facing an imminent liquidity crisis. ADP generated $5.4 billion of operating cash flow last year, carries $1.0 billion of net debt, and has $9.2 billion of undrawn committed credit facilities. Every liquidity and capital-efficiency ratio in this file is rejected, including returnOnInvestedCapitalTTM of 23.85%, which is computed on a negative invested-capital base and is therefore not merely imprecise but sign-incoherent.

What survives and is usable: net debt of $1,040.7M (total debt $5,270.8M less cash $4,230.1M — internally consistent, and consistent with the filing's long-term debt plus the $1.0B of May-2026 notes); enterprise value of $109,188.7M, which reconciles exactly to market capitalisation plus net debt with no unexplained residual; return on equity of 70.2%; and return on assets of 6.98%, which is meaningless because 74% of the assets are other people's money. Note that stockholders' equity of $6,031.2M against goodwill and intangibles of $4,937.7M leaves tangible book value of $1,093.5M — $2.74 a share. For a company with a 70% return on equity that is a feature, not a warning: ADP's book value is small because it distributes almost everything it earns.

5. Where the growth comes from, and the number that worries us

The revenue bridge the filings give, in the company's own words: growth comes "due to new business started from new business bookings, strong client retention, an increase in zero-margin benefits pass-throughs, the impact of foreign currency, an increase in pricing, and an increase in interest on funds held for clients."

Rank those by quality and the picture changes.

Two of the six named drivers — float interest and zero-margin pass-throughs — together account for $350.9 million of the nine-month revenue increase of $1,039.5 million, roughly a third, and neither carries the segment's normal economics. Add roughly a point of currency and the underlying franchise is compounding at about 5%. That is the honest organic rate, and it is what a 22.1x forward multiple has to be justified against.

Pays per control at +1% is the number that most constrains this. It is ADP's measure of employment growth among its own existing clients, on a same-store basis, and it is the cleanest volume signal any payroll processor produces. At +1%, ADP's revenue growth is coming almost entirely from price, new logos and rates — not from its installed base getting bigger. The discarded macro lane in Section 7 contains one independent channel reading ADP's own published employment series as showing "6 of last 7 months negative ADP job growth" for small businesses. That claim is about ADP's data product rather than ADP's stock, it carries no conviction weight here, and it is quoted anyway because it bears directly on the volume term above.

What would change the read: pays per control accelerating toward 2-3%, or a bookings number the company actually quantifies. The archive contains neither.

6. Valuation — priced in or room?

At $270.525 (market cap $108.14B, 399,734,282 shares):

Estimate coverage is adequate and thins fast. 11 analysts on FY2027 EPS, 9 on FY2028, and 4 on FY2029, which is excluded from every conclusion. The FY2027 EPS range is $12.093 to $12.333 — a 2.0% spread, the tightest forward distribution in this batch and a direct consequence of how predictable this company is. That tightness is informative in both directions: it means the consensus number is unlikely to be badly wrong, and it means there is very little dispersion for our view to exploit.

est.ebitdaAvg and est.ebitAvg are REJECTED for the fixed-ratio fabrication signature. In every forward year ebitdaAvg is exactly 31.73% of revenueAvg and ebitAvg is exactly 30.64% — FY2026, FY2027, FY2028 and FY2029 alike, to four decimal places. The historical rows carry different ratios (FY2023: 24.3% and 18.5%), which confirms the forward rows are generated rather than collected. They are also wrong in level: ebitdaAvg for FY2026 is $6,952.2M against a reported $6,353.3M (9.4% too high), and ebitAvg for FY2026 is $6,712.6M against actual pre-tax earnings of $5,730.3M (17.1% too high). All forward valuation in this dive uses epsAvg only.

A note on ADP's own terminology, because it is unusual and matters. ADP labels its pre-tax earnings line "EBIT" and reports an Adjusted EBIT margin of 27.3% for the nine months — but its interest expense sits inside total expenses, so ADP's "EBIT" is arithmetically earnings before income taxes, not earnings before interest and taxes. Any comparison of ADP's stated 27-30% "EBIT margin" against a peer's conventional operating margin is not like-for-like. The vendor compounds this from the other direction: operatingProfitMarginTTM of 18.58% and ebitMarginTTM of 26.71% are both in the file, differing by 813 basis points, and the derived operatingIncome field in inc_q reports −$122.5 million for the September 2025 quarter against $1,013.0 million of net income in the same quarter — an arithmetic impossibility. The whole operating-income family is rejected; pre-tax earnings and net earnings are used.

Peer context is unusable and worth naming. The vendor peer set is Deere, General Dynamics, Honeywell, Lockheed Martin, 3M, Northrop Grumman, Paychex, Parker-Hannifin, Trane Technologies and UPS — nine industrial and defence manufacturers and one actual peer. For a payroll software company this is not a comparison universe; it is a market-capitalisation bucket. Only PAYX ($42.2B) belongs, and one peer is not a set. No peer-multiple comparison is drawn.

6a. What today's price assumes (the inversion)

At $270.525 — 22.1x FY2027 consensus EPS, 20.2x FY2028, 24.3x trailing — the price embeds:

• Revenue compounds 5.8% in FY2027 and 5.7% in FY2028, reaching $24.5 billion. (Consensus; 12 analysts each.) Against a fiscal-2026 organic constant-currency rate of 6% and pays per control of +1%, this requires pricing, bookings and float balances to keep doing the work that headcount is not.
• Adjusted EPS reaches $12.253 in FY2027 and $13.390 in FY2028. (Consensus; 11 and 9 analysts.) That is +10.2% then +9.3% on 5.7% revenue growth, so roughly four points a year of the earnings growth must come from margin plus buyback. The margin record supports it — five consecutive years of expansion from 17.3% to 20.1% — but the buyback contribution is only about 1.9% a year at the fiscal-2026 pace.
• The earned yield on client float holds near 3.3%. (Our derivation; the filing supplies the rate and the balance.) This is the most fragile assumption in the price. Not because a rate decline is certain, but because it is the assumption with the largest coefficient and the least management control: $999.4 million of nine-month revenue at close to a 100% incremental margin is 22.3% of nine-month pre-tax earnings, and a 100-basis-point decline in the earned yield removes roughly 7% of pre-tax profit — a full year of consensus EPS growth, from a variable ADP does not set. The five-and-ten-year laddered portfolio spreads the impact over years rather than quarters, which is exactly why this is a fragility rather than a thesis-breaker.
• The market keeps paying roughly 22x forward earnings. (Our number.) At 20x FY2027E the stock is $245; at 25x it is $306. The multiple band is the entire question here — the earnings estimate has a 2.0% dispersion and the plausible multiple range is ±13%.
• PEO Services stops compressing. (Our number.) A third of revenue with margin down 110 basis points over nine months and pre-tax earnings down 2%. Consensus revenue growth of 5.7% with EPS growth of 9.3% is not consistent with a third of the business continuing to de-margin.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+10.2%, from ~$11.12 to FY2027E $12.253) + multiple drift (COMPRESSION of roughly 9%, from today's 22.1x on the forward year toward about 20.2x) + shareholder yield (+4.3%) ≈ +5%.

Our base case explicitly assumes the multiple compresses, and the reason is not a view about ADP — it is the starting point. The stock is 13.9% above its 50-day and 14.7% above its 200-day moving average after a +27.8% quarter; the street rates it Hold with a target 1.1% below spot; and the organic constant-currency growth rate is 6% with the same-store employment term at +1%. A 22.1x multiple on ~5% underlying growth is not obviously wrong, but it leaves nothing for us.

Note what that makes the expected return: essentially all of it is the dividend and the buyback. Earnings growth of 10% arrives and the multiple gives back roughly 9% of it. That is the definition of a fairly valued security, and it is the whole basis of the Hold.

If the multiple held at 22.1x FY2028E the price would be $296 (+9.4%) a year from now. If it compressed to 18x FY2027E, $221 (−18.4%). The stock spent part of the last twelve months at $188.79, which is 15.4x the FY2027 consensus — so the low end of this band is not hypothetical, it is where this security traded within the year.

6c. Variant perception (where we differ, what would surprise)

• We have essentially NO variant perception on the level, and our base is 1.6% above the street's. $272 against $267.67, with 8 of 36 analysts at buy and five at sell. We are not seeing anything the market is missing about ADP's earnings path, and the consensus estimate dispersion of 2.0% on FY2027 EPS tells you why: there is very little to disagree about. Under this programme's own rules, a no-edge name defaults toward Hold, and that is exactly what has happened here.
• Where we do differ is on the composition of the growth, and it is a negative divergence. The consensus revenue line treats 5.7% growth as a homogeneous quantity. It is not: roughly a third of the nine-month revenue increase came from float interest ($118.1M) and zero-margin PEO pass-throughs ($232.8M), neither of which carries normal segment economics, and roughly another point came from currency. Watchable number: organic constant-currency growth in the October print, which was 6% for the nine months. Below 5% and the 22x multiple has a problem the headline revenue line will not show.
• We think the float-interest line is under-discussed as a risk and over-credited as growth. It is 6.1% of revenue and 22.3% of pre-tax earnings, and it rose because the Federal Reserve held rates, not because ADP sold anything. The market is treating a rate-sensitive, near-100%-margin revenue stream as though it were recurring software revenue. It is recurring; it is not software.
• Positive surprise that would force a re-rate: pays per control accelerating to 2%+ alongside organic constant-currency revenue growth of 7%+, which would mean the volume term has turned and the 22x is being paid for real acceleration rather than for a rate environment; or a quantified disclosure of what next-generation products and WorkForce Software contribute, which would put a growth asset inside a company currently valued as a bond-with-upside.
• Negative surprise that would break the thesis: a fall in the earned yield on client funds below 3.0% with balances flat, which takes roughly 7% out of pre-tax earnings; PEO margin compression continuing past 150 basis points; pays per control turning negative; or simply the multiple reverting toward the 18-19x it traded at within the last twelve months, which costs 15-18% with no change in any estimate.

Synthos fair values

All three anchors are multiples of the FY2027 consensus adjusted EPS distribution (mean $12.253, low $12.093, high $12.333, 11 analysts), each cross-checked against FY2028.

• Bear ~$210 — 17.4x the FY2027 consensus LOW of $12.093, and 15.7x FY2028E. Cross-check: 11.2% above the 52-week low of $188.79; 18.9% below the 50-day average. The scenario: short rates fall and the float yield goes with them, PEO compression continues, pays per control turns negative, and the multiple returns to where this stock traded four months ago. −22.4%.
• Base ~$272 — 22.2x the FY2027 consensus MEAN of $12.253, and 20.3x FY2028E. Cross-check: 1.6% above the street's $267.67; 0.7% above the median target of $270. Sensitivity, stated openly: 20x gives $245 and 25x gives $306. The entire answer lives in a 20-25x band applied to an estimate whose own dispersion is 2%. The scenario: everything continues exactly as it is, and the multiple stays where it is. +0.5%.
• Bull ~$325 — 26.3x the FY2027 consensus HIGH of $12.333, and 24.3x FY2028E. Cross-check: 4.8% above the 52-week high of $309.03; 4.8% above the highest street target of $310. The scenario: pays per control accelerates, rates stay where they are, PEO stabilises, and the market pays a premium multiple for the most predictable large-cap earnings stream in the US market. +20.2%.

Base is 0.5% above spot; asymmetry is 0.90:1 to the DOWNSIDE (22.4% down, 20.2% up), before a shareholder yield of approximately 4.3%. A payoff ratio below one, a base case within a rounding error of the price, a street consensus below the price and a knowledge-base lane of zero is what a Hold looks like when it is arrived at rather than defaulted to.

7. Knowledge base — eighty-five hits, one survivor, and it is about a different company

Raw hits: 85. After a case-sensitive entity re-run: 1. Entity matches on ADP or Automatic Data Processing: 0. Name-level claims: 0. Discarded: 85.

The primary sweep ran entity terms ADP and Automatic Data Processing plus free text on ADP and payroll, across all 51,928 distilled claims. It returned 85 hits across 24 channels and it is a homograph sweep from top to bottom.

The discard, and why it is total. The word "payroll" in this knowledge base almost always means the US non-farm payrolls report. The channel distribution makes the point on its own: forward_guidance (12), jordi_visser (10), andreas_steno (8), real_vision (8), eurodollar_university (8), warren_pies (8), darius_dale (5), macrovoices (3) and sixteen others. Representative content: "Monthly non-farm payrolls are noise — a rounding error of ~100-200k net job changes" (compound_and_friends, 2025-03-21); "Annual payroll revisions have been negative multiple years running" (forward_guidance, 2025-08-13); "The GFC recovery was too slow because Congress withdrew fiscal support prematurely (ending the 2012 payroll tax cut)" (odd_lots, 2020-02-07). None of these is about a company.

The one claim that survives a case-sensitive entity match is about Equifax:

> 2023-11-15 · bullish · conviction 60 · horizon: thesis · entities: EFX · channel: business_breakdowns · speaker: Mo · role: independent · skill 1.0

> "Work number has a long record-growth runway — ~135M aggregated records today vs 220M Americans earning income; Equifax holds ~95M exclusive records plus exclusive payroll processor contracts."

It matched on the phrase "payroll processor" and it is a bull case for a competitor's data business, not a view on ADP. Discarded. We note in passing that the "exclusive payroll processor contracts" it describes are contracts with companies like ADP — so the only surviving claim in an 85-hit sweep treats ADP as a supplier to someone else's moat.

Two claims in the discarded pool name ADP as a data product, and they are quoted at zero conviction weight because they bear on the volume term in Section 5:

> 2025-12-05 · BEARISH · conviction 80 · horizon: thesis · no entities · channel: forward_guidance · no named speaker · skill 1.0

> "Small businesses are in full recession — 6 of last 7 months negative ADP job growth — while large-cap tech masks it."

> 2025-09-13 · neutral · conviction 75 · horizon: principle · no entities · channel: jordi_visser · no named speaker · skill 2.0

> "Non-farm payrolls and its revisions are near-useless; jobless claims (near record lows) and ADP are better real-time labor signals."

The first is the more useful. It is a claim about the small-business employment picture drawn from ADP's own published series, and small and mid-sized businesses are the core of both Employer Services' downmarket base and all of PEO Services. ADP's own disclosed pays per control of +1% is consistent with a weak but not contracting client base; this claim is more negative than that. It carries no conviction weight, it is eight months old, and it is not a view on the security — but it is the only outside read in our possession on the variable that most constrains ADP's organic growth, and it is unfavourable. The second is a compliment to the quality of ADP's data, not to the stock.

Attribution note. All 85 hits carry speaker: null except the single Equifax claim, which names "Mo" with speaker_role: independent. No claim carries a management speaker_role, and none is needed, because there are no name-level claims of any kind. No concentration sensitivity test is meaningful on a lane of zero — though it is worth recording that forward_guidance, jordi_visser, andreas_steno, real_vision, eurodollar_university and warren_pies together supplied 54 of 85 raw hits, 64%, which on a real lane would have triggered one.

Conclusion. Breadth 0, claim count 0, net conviction none. The Synthos knowledge base has no view on Automatic Data Processing. Fifty-one thousand nine hundred and twenty-eight distilled claims and not one of them is about this company. For a $108 billion S&P 100 constituent that is itself a finding: ADP is not a name the credible-thinker universe argues about, which is consistent with everything else in this dive — it is a predictable business that generates no disagreement, and disagreement is where this programme's edge comes from.

8. Data integrity — what we rejected and why

Seven findings. ADP's vendor file has one severe classification error, one definition change mid-series, and one arithmetically impossible field — against which the share count, the enterprise-value rebuild, the segment geography and the estimate basis all check out exactly.

1. $46 billion of funds held for clients is classified as NON-current in the FY2026 balance sheet while the matching client funds obligations remain CURRENT — REJECTED, with the entire liquidity and capital-efficiency ratio family. Between the FY2025 and FY2026 rows of bal_a, total current assets fall from $43,252.2M to $8,603.3M while total non-current assets rise from $10,117.1M to $54,590.0M, a $34.6 billion migration with no corresponding event. Total current liabilities remain $50,024.9M. The 10-Q at 2026-03-31 reports funds held for clients of $46,413.1M as a current asset and client funds obligations of $46,774.5M as a current liability — they offset. Consequential rejections: currentRatioTTM 0.172, quickRatioTTM 0.172, cashRatioTTM 0.085, workingCapitalTTM −$41,421,600,000, netCurrentAssetValueTTM −$48,558.8M, workingCapitalTurnoverRatioTTM −1.12, investedCapitalTTM −$35,454,000,000, and returnOnInvestedCapitalTTM of 23.85% — a positive return computed on a negative capital base, which is sign-incoherent. A screen using any of these would classify one of the least financially fragile companies in the S&P 500 as a liquidity risk.

2. capitalExpenditure changes definition between FY2024 and FY2025 — the free-cash-flow series is not comparable to itself. For FY2023 and FY2024 the vendor's figures ($571.6M and $563.4M) equal the filing's capital expenditures plus additions to intangibles, to the dollar ($206.3M + $365.3M; $208.4M + $355.0M). For FY2025 the vendor's $168.7M equals the filing's capital-expenditure line alone, excluding $378.3M of intangibles additions. FY2025 and FY2026 free cash flow is therefore overstated by roughly $350-380M relative to the FY2023 and FY2024 basis, inflating the three-year FCF growth rate from ≤10.4% to 13.0%. We use the filing's definitions and a consistent-basis series throughout, and state the FY2026 figure as an upper bound of $4,892.2M because the fiscal-2026 10-K has not been filed and only nine months of intangibles additions ($352.4M) are available. The capex sanity-check against D&A was run and this is what it found: $196.6M of "capex" against $585.9M of D&A is not a company under-investing, it is a company whose investment is being reported in a different line.

3. inc_q operatingIncome for the September 2025 quarter is −$122.5M against $1,013.0M of net income in the same quarter — arithmetically impossible, and the whole operating-income family is REJECTED. ADP's income statement has no operating-income line: total revenues less total costs of revenues less SG&A less interest expense equals earnings before income taxes. The vendor's derived figure has broken on at least one quarter, and the annual rows carry a related artefact — totalOtherIncomeExpensesNet for FY2026 is +$1,651.9M against −$101.8M in FY2025, a $1.75 billion swing that is a reclassification plug, not an event, and which makes the FY2026 "operating income" of $4,078.4M read as a 25% DECLINE from FY2025's $5,411.9M in a year when pre-tax earnings ROSE 7.9%. operatingProfitMarginTTM (18.58%) and ebitMarginTTM (26.71%) differ by 813 basis points in the same file. All of it is discarded; pre-tax earnings and net earnings are used, and the segment margins come from the 10-Q.

4. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature — REJECTED. ebitdaAvg is exactly 31.73% of revenueAvg and ebitAvg exactly 30.64% in every forward year (FY2026 through FY2029), while the historical rows carry different ratios (FY2023: 24.3% and 18.5%). They are also wrong in level: FY2026 ebitdaAvg of $6,952.2M against a reported $6,353.3M, and FY2026 ebitAvg of $6,712.6M against actual pre-tax earnings of $5,730.3M. All forward valuation uses epsAvg. The FY2029 row rests on 4 analysts and is excluded.

5. seg_prod silently drops the client-funds interest line in FY2025 — a $1.19 billion, 5.8% omission. The FY2024 and all earlier entries carry five categories including "Interest on Funds Held for Clients" ($1,024.7M in FY2024). The FY2025 entry carries only four — HCM $8,674.5M, HRO $3,782.3M, Global $2,626.0M, PEO $4,289.0M — summing to $19,371.8M against actual FY2025 revenue of $20,560.9M. The missing $1,189.1M is the float interest, which is the single most economically distinctive line in this company. There is also no FY2026 entry at all. Separately, these four categories are revenue-disaggregation groupings, NOT the two reportable segments (Employer Services and PEO Services) that the 10-Q uses and that carry the margin disclosure. All segment analysis in this dive comes from the 10-Q.

6. tech.max_dd_from_peak of −17.34% contradicts the same block's own 52-week range. tech reports hi52 of $309.03 and lo52 of $188.79 — a 38.9% peak-to-trough range — with ret_12m of −10.09% on a last price of $270.13, implying the stock stood near $300 a year ago. A maximum drawdown from peak of only 17.34% is not compatible with those three facts. We use the 52-week high and low and the twelve-month return, all of which agree with each other, and discard max_dd_from_peak. (Minor and recorded: quote.yearHigh/yearLow of $310.08/$188.16 sit 0.34% and 0.33% from tech.hi52/lo52; tech is used throughout for internal consistency.)

7. The peer set is a market-capitalisation bucket, not a comparison universe. peers returns DE, GD, HON, LMT, MMM, NOC, PAYX, PH, TT and UPS — three defence primes, four industrial manufacturers, a logistics company, an agricultural-equipment maker, and Paychex. For a payroll and HCM software company, nine of ten are irrelevant. No peer-multiple comparison is drawn. (We note without comment that GD and TT are separately covered in this same batch.)

Where vendor and filing AGREED — recorded, because verified-clean is a real finding:

• Share count is an essentially exact match. The 10-Q states 399,734,282 shares outstanding at 2026-04-27 against 399,734,001 implied by market capitalisation ÷ price — a difference of 281 shares, 0.0001%.
• The enterprise-value rebuild reconciles with no residual. enterpriseValueTTM $109,188.7M = market capitalisation $108,148.0M + net debt $1,040.7M, and net debt = total debt $5,270.8M − cash $4,230.1M. No ETN-class internal contradiction here. Long-term debt of $4,964.1M at FY2026 year-end is consistent with the filing's $3,977.3M at 2026-03-31 plus the $1.0 billion of 5.000% notes issued in May 2026 under the 8-K of 2026-05-07. (totalDebt includes $306.7M of capitalLeaseObligations; this is 5.8% of total debt and does not change any conclusion, but it is noted because lease double-counting is a confirmed defect class elsewhere in this programme.)
• seg_geo is complete and includes the home market. FY2025: United States $18,179.2M (88.4%), Europe $1,533.5M, Canada $489.3M, Other $358.9M — summing to $20,560.9M, exactly the reported FY2025 revenue. This is the defect class that has caught TJX, SYK and six others, and ADP passes it cleanly.
• The estimate basis matches the reported basis. est.revenueAvg for FY2026 is $21,910.6M against a reported $21,947.5M, 0.17% apart — no DE-class basis mismatch.
• acquisitionsNet for FY2025 of −$1,165.1M matches the filing's "Acquisitions of businesses, net of cash acquired" line exactly, and FY2026's −$22.8M is consistent with the nine-month figure of $22.9M.
• effectiveTaxRateTTM of 22.98% is consistent with the 10-Q's disclosed nine-month effective rate of 23.2%.

Vendor composite rating — accepted with a caveat. The rating block gives B+ / 3 overall, with 5/5 on return on equity, 4/5 on return on assets and discounted cash flow, and 1/5 on both debt-to-equity and price-to-book. The two 1s are artefacts, not findings: debt-to-equity of 0.874 and price-to-book of 17.9x are what a company with a 70% return on equity and a 59.5% payout ratio produces by construction, because it distributes rather than accumulates book value. The 4/5 discounted-cash-flow score rests on the free-cash-flow figure corrected in finding 2 and should be discounted accordingly.

Non-equity tripwire — checked and passed. ADP is common stock, $0.10 par value (voting), NASDAQ Global Select Market, per the 10-Q cover page. Price of $270.525 is not par-like; beta is 0.841; the dividend is a variable declared amount; volume was 1.59M shares (~$431M of turnover); the 52-week band of $188.79 to $309.03 is a 64% range. This is common equity.

9. Technicals

• Price $270.525. −12.6% from the 52-week high of $309.03; +43.1% above the 52-week low of $188.79. Position within the annual range: 68th percentile.
• Both moving averages are far below the price. +13.9% above a 50-day average of $237.46; +14.7% above a 200-day average of $235.74; the 50-day has crossed above the 200-day. These are the two widest price-to-average gaps in this batch, and they are the technical statement of the same fact the valuation section makes arithmetically.
• RSI 66.4 — strong, approaching but not at overbought. MACD +9.48, strongly positive.
• Relative performance, and the shape is unusual. 3-month +27.8% against SPY +7.6% and QQQ +7.7% — outperforming by 20 points in a single quarter; 6-month +9.1% against SPY +11.1%; 12-month −10.1% against SPY +24.3% and QQQ +30.8% — a 34-point deficit. The entire twelve-month gap was created in the first three quarters of the period and a large part of it has been closed in the last one. That is a round trip, and the practical consequence is that the easy part of the recovery is behind.
• Sentiment: 8 buy, 23 hold, 5 sell across 36 analysts; consensus target $267.67 — 1.1% BELOW spot — median $270 (−0.2%), high $310 (+14.6%), low $225 (−16.8%). This is the first name in this batch where the consensus target sits below the market price, and it is the single most direct piece of evidence for the verdict. Five sell ratings on a company of this quality is also unusual and worth recording.

Today's move and what it does to the entry

ADP closed 2026-08-04 at $270.525, up 0.28% or $0.745 from $269.78. It opened at $264.99, traded $263.36 to $270.73, and closed within 0.1% of the day high on 1.59M shares. No company filing is dated 2026-08-04; the last event was the fiscal-2026 results release on 2026-07-29, six days earlier.

The honest read: this is not an entry, it is the far side of one. The stock is 43.1% above its 52-week low, 14% above both moving averages, and has just outperformed the index by 20 points in ninety days. The point at which this becomes interesting is arithmetically specific and is stated below rather than left as a sentiment: at $232 the stock is 19x the FY2027 consensus, roughly on its 200-day average, and the 4.3% shareholder yield plus 10% earnings growth would not need any multiple help. That is a 14% decline from here, and it is inside the range this security traded in four months ago.

10. Insiders — eight transactions, and the file is dominated by mechanics

The reading, and it is close to signal-free.

Five of the eight transactions belong to one Corporate Vice President on a single day, and the pattern is the classic exercise-and-sell: an 887-share option exercised at $138.53 and 2,414 shares sold at $265.62 the same day, one day after the fiscal-year results. Total proceeds approximately $641,000. He retains 9,659.6 shares, a 20% reduction from 12,073.6.

The other three are identical 49.4306-share acquisitions at $223.94 on 2026-07-01 by the CEO and two officers, coded I-Discretionary. Identical fractional share amounts on the same date across three unrelated executives is a plan mechanic — almost certainly a deferred-compensation or benefit-plan crediting — not a decision. They are not purchases in any meaningful sense.

What the file does not contain: any transaction by the chief financial officer, any open-market purchase at market prices, and any sale by a named executive officer other than one Corporate VP. Net: nothing to read. We record only that the single genuine market sale in the file was executed at $265.62 on the day after the fiscal-year print, at 96% of a price the stock had not seen since February.

11. Verdict, kill-criteria and flip conditions

Hold.

This verdict is arrived at, not defaulted to, and the arithmetic that produces it is short.

The business. Fiscal 2026 revenue of $21,947.5M (+6.7%), net earnings of $4,413.5M (+8.2%), a 20.1% net margin after five consecutive years of expansion from 17.3%, a 70.2% return on equity, $5,441.2M of operating cash flow, net debt of $1,040.7M, a 0.841 beta and a total shareholder yield near 4.3%. Employer Services expanded its margin 130 basis points in the March quarter. There is nothing wrong with this company and we do not want to be read as saying there is.

The price. $270.525 is 22.1x the FY2027 consensus of $12.253, 20.2x FY2028's $13.390, 13.9% above the 50-day average and 14.7% above the 200-day, after a +27.8% three-month move against SPY's +7.6%. The street rates it Hold by 23 to 8 with five sells and its consensus target of $267.67 is below the last trade. Our base of $272 is 0.5% above spot with 0.90:1 asymmetry — the first sub-parity payoff ratio in this batch.

The composition. Organic constant-currency growth is 6%, roughly a point below the reported rate. Pays per control — the same-store employment measure inside ADP's own client base — grew 1%. Of the nine-month revenue increase of $1,039.5M, $118.1M was float interest and $232.8M was zero-margin PEO pass-throughs, and PEO's margin fell 110 basis points with its pre-tax earnings down 2% on 6% revenue growth. $999.4M of nine-month revenue, 6.1% of the total and 22.3% of pre-tax earnings, is the Federal Reserve.

The conviction. Zero. Eighty-five raw knowledge-base hits, one survivor after a case-sensitive run, and the survivor is about Equifax. Fifty-one thousand nine hundred and twenty-eight distilled claims contain no view on this company.

Put together: a very good business at a full price, with no variant perception, a negative payoff ratio, a street target below spot and an empty conviction lane. That is a Hold, and manufacturing anything else out of it would be dishonest.

Pre-registered KILL criteria — what would take this to Avoid:

• The earned yield on funds held for clients falling below 3.0% with average balances flat or lower, which removes roughly 7% of pre-tax earnings from a line with a near-100% incremental margin.
• Pays per control turning negative, which would mean the installed base is shrinking and every point of revenue growth has to come from price and new logos.
• PEO Services margin falling below 12.5% (13.4% for nine months, down 110 basis points), or PEO pre-tax earnings declining for a fourth consecutive quarter.
• Organic constant-currency revenue growth below 5% in any quarter, against 6% for the nine months.
• A break in the beat pattern — five consecutive quarters of 1.3-2.1% beats is the foundation of the multiple, and a miss on a company this predictable is a different kind of event than a miss elsewhere.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price near $232, which is 19x the FY2027 consensus and roughly the 200-day average, at which the 4.3% shareholder yield plus ~10% earnings growth requires no multiple expansion at all. This is the single most likely route to an upgrade and it requires nothing from the company.
• Pays per control accelerating to 2%+ with organic constant-currency revenue growth at 7%+, which would mean the volume term has turned and 22x is being paid for acceleration rather than for a rate environment.
• A quantified disclosure of next-generation product or WorkForce Software revenue. ADP has been capitalising roughly $350-380 million a year of internally developed software and has told us only about its amortisation. A revenue line attached to it would put a growth asset inside a company currently valued as a very high-quality bond.
• PEO margin stabilising for two consecutive quarters.
• Any name-level knowledge-base coverage at all. At present there is none.

Where ADP fits in the Synthos Framework Portfolio. No position today. It goes on the watch list at a specific number — $232, 19x FY2027 consensus — with a 2% initial sleeve at that level in the quality-compounder bucket. Sizing note: this is the kind of name where the mistake is not owning it, it is paying 22x for it after a 28% quarter and then having no room to add when it is 19x. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $270.525, with the fair-value anchors, the kill criteria and the $232 upgrade trigger all gradeable.

Single biggest risk: the multiple. There is no leverage risk here, no credit risk, no customer-concentration risk and no disclosed litigation of consequence. The risk is that a company growing revenue at 5-6% organic constant-currency, with 6.1% of revenue supplied by short-term interest rates and a same-store employment term at +1%, is being asked to hold 22x forward earnings from a starting point 14% above its own 200-day average, with its own analyst community rating it Hold and targeting a price below the market. The stock traded at $188.79 within the last twelve months, which is 15.4x the same FY2027 consensus. Nothing about the business changed between that price and this one.

Provenance & disclosures

• Traceability: 0 knowledge-base claims name Automatic Data Processing out of 51,928 distilled claims (raw hits 85, surviving a case-sensitive entity re-run 1, entity matches on ADP 0, name-level 0, discarded 85; breadth 0, net conviction none). The sweep is a pure homograph result: the free-text term "payroll" matches the US non-farm payrolls report, which is a standing topic across forward_guidance (12 hits), jordi_visser (10), andreas_steno (8), real_vision (8), eurodollar_university (8), warren_pies (8) and eighteen further channels. The single case-sensitive survivor carries entity EFX and is a bull case for Equifax's Work Number (business_breakdowns, 2023-11-15, speaker "Mo", speaker_role: independent, bullish 60) — it matched on the phrase "exclusive payroll processor contracts" and is discarded. Two discarded claims name the ADP employment report as a data product and are quoted in Section 7 at zero conviction weight because they bear on ADP's own volume driver: forward_guidance 2025-12-05 (bearish 80, "Small businesses are in full recession — 6 of last 7 months negative ADP job growth") and jordi_visser 2025-09-13 (neutral 75, skill 2.0, "jobless claims... and ADP are better real-time labor signals"). Neither is a view on the security and neither enters any conviction score. All 85 hits except the Equifax claim carry speaker: null and are channel-attributed. No claim carries a management speaker_role. No concentration sensitivity test is meaningful on a lane of zero, though six channels supplied 64% of the raw pool. All quotes are verbatim from the stored claim text.
• Data as-of: fiscal-2026 full-year and quarterly revenue, earnings, balance sheet and cash flow through 2026-06-30 from the vendor payload; all segment results, margins, client-float balances and yields, pays per control, capital-expenditure figures, share count and operating commentary through 2026-03-31 from the 10-Q filed 2026-04-30; annual figures for FY2025 and earlier from the 10-K filed 2025-08-06 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873552 = 2026-08-04T19:59:12Z ($270.525, +0.28%; 50-DMA $237.46; 200-DMA $235.74; RSI 66.4; MACD +9.48) · knowledge-base claims 2026-08-04. ADP's fiscal year ends 30 June; "FY2026" throughout means the year ended 2026-06-30. All figures come from the Synthos vendor data file for ADP or from the SEC filings in the ADP archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents, and one significant gap: 10-K filed 2025-08-06 (fiscal 2025); 10-Q filed 2026-01-29 (December 2025 quarter); 10-Q filed 2026-04-30 (March 2026 quarter — the source of nearly every operating figure in this dive); 8-K filed 2026-05-07 (the $1.0 billion 5.000% senior notes due 2036); 8-K filed 2026-06-26 (the $5.7 billion 364-day and $3.5 billion five-year credit facilities replacing prior $4.55 billion and $3.5 billion facilities); 8-K filed 2026-07-29 (fiscal-2026 fourth-quarter and full-year results). All carry preserved [TABLE] statement data. THE GAP: the fiscal-2026 10-K has not been filed, and the 2026-07-29 8-K contains only the cover page, Item 2.02 notice and exhibit index — the earnings release itself is Exhibit 99, which is not in the extracted text. So fiscal-2026 fourth-quarter and full-year segment detail, client-float balances, pays per control and management guidance are NOT available from a filing, and this dive does not estimate around them: every fourth-quarter and full-year figure used here comes from the vendor payload or the earnings calendar and is labelled as such, while every segment, margin and operating metric is a nine-month figure from the March 10-Q.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the capital-expenditure definition — the vendor's FY2023 and FY2024 figures ($571.6M, $563.4M) equal the filing's capital expenditures plus additions to intangibles to the dollar, while FY2025's $168.7M equals the capital-expenditure line alone and excludes $378.3M of intangibles, making the payload's own free-cash-flow series non-comparable across FY2024/FY2025 and inflating three-year FCF growth from ≤10.4% to 13.0% (we rebuild on a consistent basis and cap FY2026 at $4,892.2M, labelled an upper bound); the classification of funds held for clients, which the 10-Q reports as a current asset ($46,413.1M at 2026-03-31) matching a current liability ($46,774.5M) while the payload's FY2026 balance sheet moves $34.6 billion of assets to non-current and leaves the obligations current, invalidating currentRatioTTM (0.172), workingCapitalTTM (−$41.4B), investedCapitalTTM (−$35.5B) and returnOnInvestedCapitalTTM (23.85% on a negative base); inc_q operatingIncome of −$122.5M for the September 2025 quarter against $1,013.0M of net income in the same quarter, and the related FY2026 totalOtherIncomeExpensesNet plug of +$1,651.9M that makes "operating income" fall 25% in a year pre-tax earnings rose 7.9% — against which the 10-Q's disclosed nine-month pre-tax earnings of $4,472.5M and EBIT margin of 27.1% are used; seg_prod, whose FY2025 entry silently drops the "Interest on Funds Held for Clients" category present in every earlier year, understating FY2025 revenue by $1,189.1M (5.8%) and omitting the single most economically distinctive line in the company, with no FY2026 entry at all and with categories that are revenue disaggregations rather than the two reportable segments; est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio signature (exactly 31.73% and 30.64% of revenue in every forward year) and for overstating FY2026 EBITDA by 9.4% and FY2026 EBIT by 17.1% against reported figures; and tech.max_dd_from_peak of −17.34%, incompatible with the same block's 52-week range of $188.79–$309.03 and its −10.09% twelve-month return. Where vendor and filing AGREED — recorded: the share count is an essentially exact match (10-Q 399,734,282 at 2026-04-27 against 399,734,001 implied by market capitalisation ÷ price, 0.0001% apart); the enterprise-value rebuild reconciles with no residual ($109,188.7M = $108,148.0M market cap + $1,040.7M net debt, with net debt = $5,270.8M total debt − $4,230.1M cash, and long-term debt of $4,964.1M consistent with the filing's $3,977.3M at 2026-03-31 plus the $1.0B May-2026 notes); seg_geo is complete and includes the United States at 88.4% of FY2025 revenue, summing exactly to the reported total; the estimate basis matches the reported basis (est.revenueAvg FY2026 $21,910.6M against a reported $21,947.5M, 0.17%); acquisitionsNet matches the filing's acquisitions line exactly in both FY2025 (−$1,165.1M, WorkForce Software) and FY2026; and effectiveTaxRateTTM of 22.98% is consistent with the 10-Q's 23.2% nine-month rate.
• Basis note: analyst estimates are on adjusted diluted EPS. ADP's adjustments are small — nine-month fiscal-2026 GAAP diluted EPS was $8.49 against adjusted $8.48, a difference of 0.1%, and adjusted EBIT exceeded EBIT by $30.1M on $4,472.5M, 0.7%. This is one of the cleanest GAAP-to-adjusted bridges in the programme and no conclusion here turns on it. Adjusted figures "exclude interest income and interest expense that are not related to our client funds extended investment strategy, and net charges, including certain legal matters, broad-based optimization initiatives, and (gains)/losses on ADP Ventures' investments." Trailing fiscal-2026 adjusted EPS of ~$11.12 is the sum of the four quarterly earn_cal actuals ($2.49 + $2.62 + $3.37 + $2.64) and is labelled as such throughout; the vendor's inc_a EPS of $10.97 is the GAAP figure on basic-equivalent shares. Also note ADP's idiosyncratic use of "EBIT" to mean earnings before income taxes — its interest expense sits inside total expenses — so its stated 27-30% "EBIT margin" is not comparable to a conventional operating margin.
• Estimate coverage: 12 analysts on FY2027 and FY2028 revenue and 7 on FY2029; 11 on FY2027 EPS and 9 on FY2028 — the anchors for all three fair values — with an FY2027 EPS range of $12.093 to $12.333, a 2.0% spread, the tightest forward distribution in this batch; 4 on FY2029 EPS, which is excluded from every conclusion.
• Peer note: the vendor peer set — Deere, General Dynamics, Honeywell, Lockheed Martin, 3M, Northrop Grumman, Paychex, Parker-Hannifin, Trane Technologies and UPS — contains one actual peer (Paychex) and nine industrial, defence and logistics companies. No peer-multiple comparison is drawn.
• Fair-value caveat: the $210 / $272 / $325 anchors are multiples of the FY2027 consensus adjusted EPS distribution — 17.4x the low of $12.093, 22.2x the mean of $12.253, and 26.3x the high of $12.333 — each cross-checked against FY2028E (15.7x, 20.3x, 24.3x). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 20x gives $245 and 25x gives $306, so the answer lives entirely in the multiple, because the estimate dispersion is only 2%. The base case assumes the forward multiple COMPRESSES from 22.1x toward 20.2x over the next year, which is why a +10% earnings year produces a ~+5% total return.
• Rate-sensitivity note: interest on funds held for clients was $999.4 million of nine-month fiscal-2026 revenue (6.1% of the total, 22.3% of nine-month pre-tax earnings), earned at a 3.3% average yield on $40.2 billion of average balances, and it carries close to a 100% incremental margin. A 100-basis-point fall in the earned yield is roughly $400 million of revenue and pre-tax earnings, about 7% of the fiscal-2026 pre-tax total of $5,730.3 million. The portfolio is laddered "out to five years... and out to ten years" per the 10-K, so the impact accrues over years rather than quarters. This sensitivity is our derivation from the filing's disclosed balances and yields; it is not a company disclosure.
• Timing: fiscal-2026 fourth-quarter and full-year results were released 2026-07-29, six days before this dive, and beat the consensus adjusted EPS estimate by 1.9% ($2.64 against $2.59) — the fifth consecutive beat, all between 1.3% and 2.1%. The next print is 2026-10-28, 85 days away (consensus adjusted EPS $2.74, revenue $5,488M). The most recent insider filings are dated 2026-07-31. 2026-08-04 carried no company-specific filing.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.