Every vendor charging per employee is selling against a number that is no longer rising, and the market has already priced it.
01
02
The per-employee-per-month licence was never a software pricing model. It was a claim on the customer's headcount growth, and for two decades that claim paid. It has stopped.
ADP guides pays per control, employment measured inside its own client base, to flat-to-1%. That is the cleanest read available on whether mid-market payroll seats are growing, because it is not a survey or an index but the largest player counting the employees it actually bills for. On one day in August, three vendors serving three overlapping segments disclosed the same shape independently: flat per-employee volume against weekly hours at a five-year high. Employers are holding headcount and extracting more hours from the heads they have. The labour is growing. The thing the software charges for is not.
What follows is arithmetic rather than strategy. Paylocity guided recurring growth down from 12.2% into roughly 8%. Paycom guided up into the same band. Two vendors arriving at one number from opposite directions is not a coincidence of go-to-market; it is what happens when the underlying unit is flat and every seller is dividing the same pie. There is no configuration of sales motion that beats the sector rate when the denominator has stopped moving.
The consequence is that adjacency expansion stops being a strategy choice and becomes a forced move. When ADP pushed into IT provisioning it was not diversifying for its own sake, it was reaching for revenue that does not depend on a headcount multiple. Capital priced this before the vendors said it out loud: core HR, payroll, benefits and financial wellness is the only work-tech category whose funding peak is 2025-26 rather than 2021, at $6.5B across two years, precisely because payroll carries the highest switching cost and the widest adjacent surface.
The sharpest evidence is what the market did to a good quarter. ADP beat on both lines, raised guidance, and drew six analyst price-target raises with no downgrades. The stock fell 3.48%. Analysts moved their numbers up and the equity moved down, which is a multiple compressing rather than an earnings problem. What is being repriced is the durability of an installed base the company itself has guided to flat-to-1% growth.
What would change our mind. The frontline is the stated exception and it is a real one: headcount there genuinely still grows and churns, which is why seat-based agents have compounding volume in that segment. And one month of hours data is one month. If pays per control returns above 2% for two consecutive quarters while mid-market guides re-accelerate past 10%, the ceiling was cyclical and this is wrong.
Transformed 18 September. The claim first said growth above the band comes from attach or acquisition. The evidence that arrived kept saying something narrower: the first response to a flat billing unit is price. Recruit grew revenue per posting while postings fell, ADP said AI is paid for through general price rises, and Intuit named pricing beside attach while the workers it pays stayed at 18 million. The claim now orders the levers, price first and attach second, which means a vendor growing above the band on attach with no price contribution now counts against it.
03
Employment inside the installed base has stopped growing while labour input rises. ADP guides pays per control to flat-to-1% - the largest player measuring employment inside its own client base, quarterly and public. Three vendors across three overlapping segments disclosed the same picture on one day: flat per-employee volume against weekly hours at a five-year high. The forced response is price per unit first and attach second; adjacency expansion follows rather than leads.
04
Every vendor whose ACV is a headcount multiple converges on the sector growth rate, because no go-to-market beats a flat unit - already visible as Paylocity guiding 12.2% down into ~8% and Paycom guiding 7-8% up into it. Differentiation moves off growth and onto attach, which is why capital concentrated on the adjacent surface: Core HR + payroll + benefits + financial wellness is the only work-tech category whose funding peak is 2025-26 rather than 2021, at $6.5B across 2024-25. Vendors without an adjacent surface get repriced as utilities whatever their AI story - ADP beat, raised, took six target raises and zero downgrades, and fell 3.48%.
05
Scoped to non-frontline. The frontline is the exception - the one large segment where headcount genuinely still grows and churns, which is why seat-based agents have compounding volume there. See seat-growth exception in frontline-data-control-plane.
06
Given equal weight to the claim. Hiding the counter is how a prediction becomes an article of faith.
07
08
Specific public numbers we check on a schedule, so this claim can be tested without taking our word for it. Not checked means nobody looked. That counts as nothing, never as agreement.
| What we check | Status | Latest reading | Last checked |
|---|---|---|---|
| US pays per control, ADP's own installed base ADP quarterly earnings · quarterly | not moving | Pays per control grew ~1% in FY26 (CFO, Citi TMT); a 1% move is ~25-30bps of ES revenue; CFO ranks bookings, retention and price (100-150bps, 'similar in FY 2027') above it | 2026-09-15 |
| Small-business weekly hours worked index Paychex Small Business Employment Watch · quarterly | changing | hours at five-year high vs flat jobs index | 2026-08-05 |
| Recurring-revenue growth guides Paylocity / Paycom earnings · quarterly | changing | both converged on ~8% from opposite directions | 2026-08-05 |
09
Cumulative linked signals, by direction. A line that only climbs in green is being read generously.
8 counted: 3 diverges · 5 supports · 6 predate the claim and are not counted
| Date | Signal | Bearing | On which part of the mechanism |
|---|---|---|---|
| 2026-08-04 | signal-payroll-heads-versus-hours | Before the claim (not counted) | Employment inside the installed base has stopped growing while labour input rises. |
| 2026-08-05 | signal-midmarket-payroll-eight-percent | Before the claim (not counted) | Mid-market recurring-revenue growth converges on the ~8% band because no go-to-market beats a flat unit. |
| 2026-08-04 | signal-paylocity-fy27-guide | Before the claim (not counted) | Mid-market recurring-revenue growth converges on the ~8% band because no go-to-market beats a flat unit. |
| 2026-08-03 | signal-paychex-wise-copilot | Before the claim (not counted) | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-08-13 | signal-recruit-arpj-postings-fall | diverges | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-06-12 | signal-workday-var-hybrid-channel | Before the claim (not counted) | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-08-27 | signal-workday-ai-arr-adoption-gap | supports | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-04-08 | signal-silae-kanoon-legal-payroll | Before the claim (not counted) | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-09-01 | signal-gusto-smb-hiring-decelerates | supports | Employment inside the installed base has stopped growing while labour input rises. |
| 2026-09-02 | signal-adp-ner-august-midsize-squeeze | supports | Employment inside the installed base has stopped growing while labour input rises. |
| 2026-09-04 | signal-bls-flat-workforce-pepm-ceiling | supports | Employment inside the installed base has stopped growing while labour input rises. |
| 2026-09-10 | signal-cegid-silae-merger | diverges | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-09-09 | signal-adp-ai-priced-through-general-increases | diverges | The forced response to a flat unit is price per unit first and attach second, identifiable in the vendor's own disclosure; adjacency expansion follows rather than leads. |
| 2026-09-17 | signal-intuit-payroll-line-becomes-workforce | supports | Employment inside the installed base has stopped growing while labour input rises. |
10
The market moved along a different path than the one drawn. The prediction keeps its identity and its history. A narrative that changed is itself evidence.
Was: No mid-market HR&P vendor sustains recurring-revenue growth materially above the ~8% band on seat expansion alone. Growth above it comes from attach or acquisition and is identifiable as such in the vendor's own disclosure.
Became: No mid-market HR&P vendor sustains recurring-revenue growth materially above the ~8% band on seat expansion alone. Growth above it comes from price per unit first and attach second, and is identifiable as such in the vendor's own disclosure.
Why: Six of fourteen entries diverged on the same part, and they diverged the same way: the response to a flat billing unit was PRICE, not adjacency. Recruit grew revenue per job posting while postings fell (13 Aug); ADP's CFO said AI is paid for through general price increases of 100-150bps a year (9 Sept); Intuit listed 'pricing for value' as an ARPC driver beside attach while workers paid stayed at 18M (17 Sept). The original text said growth above the band comes from 'attach or acquisition', so every price-led vendor scored as divergence and the claim could never be confirmed or contradicted on the evidence that actually arrived. DIRECTION, stated honestly: naming price admits evidence the old text scored as divergence. In exchange the claim now ORDERS the levers, so a vendor whose above-band growth is disclosed as attach- or acquisition-led with no price contribution is evidence against it, which the old text could not produce. The kill is unchanged.
10 signals carried forward; 4 dropped as no longer bearing on the claim: Cegid and Silae merging is consolidation, not a pricing or attach response to a flat unit; the flat-seat driver was absent from the deal rationale. It does not test the re-specified part.; Pre-formation (3 Aug): provenance, never scored.; Pre-formation (12 Jun): provenance, never scored.; Pre-formation (8 Apr): provenance, never scored.
11
Re-derives the retired trial `us-volume-flat`, with an instrument the trial lacked. Declared and dated 2026-08-12; NOT backdated to the trial.
This prediction carries no resolution date. We cannot predict when evidence will arrive, so the review cadence attaches to the instruments above rather than to the claim. It runs until the market proves it, moves it, or twelve months pass with no material signal against it.
How a signal becomes a prediction →